Copyright TechTarget - All rights reserved https://cyber.law.harvard.edu/rss/rss.html Techtarget Feed Generator en Wed, 14 Oct 2026 21:12:10 GMT https://www.techtarget.com/searchenterprisedesktop editor@techtarget.com <p>GlobalFoundries will operate two new semiconductor fabs in Malta, N.Y., under a U.S.-Japan investment initiative worth up to $21 billion. The project adds to U.S. efforts to expand domestic chip manufacturing under the CHIPS and Science Act, while giving U.S. and Japanese interests ownership of GlobalFoundries' facilities.</p> <p>For data center operators and IT infrastructure buyers, the investment could expand U.S. production of chips used in networking, power management and connected systems. A separate $955 million reallocation of GlobalFoundries' federal funding would speed investments in silicon photonics, silicon-germanium (SiGe) and advanced packaging, technologies relevant to <a href="https://www.techtarget.com/it-infrastructure/news/366628409/Broadcoms-Jericho4-boosts-data-center-AI-networking">high-speed data movement</a> and AI infrastructure. The new fabs, however, will not begin delivering additional capacity immediately.</p> <p>Japan's initial investment of approximately $10 billion will fund the first 300-millimeter wafer fab, which GlobalFoundries expects to bring online in 2029. A second investment of approximately $10 billion would fund another fab at the site.</p> <p>The first facility is expected to add 900,000 wafers of annual manufacturing capacity when fully operational. A second would add a similar amount. The project is expected to create about 27,000 construction and ecosystem jobs and more than 1,000 high-skilled manufacturing jobs in New York's Capital Region.</p> <p>The deal targets foundational semiconductor manufacturing used in automotive, industrial and connected infrastructure markets. These chips support functions such as power management, connectivity and embedded computing.</p> <p>"This is one of the largest foundational-node capacity commitments since the <a href="https://www.techtarget.com/it-infrastructure/news/252523061/CHIPS-Act-takes-step-forward-on-long-road-to-production">CHIPS Act</a>, and the first funded this way," said Stephen Sopko, semiconductor and deep tech practice lead at HyperFrame Research.</p> <p>The production timeline and conditional second fab leave execution risks. The first facility is still years from production, and the second depends on additional investment.</p> <p>"Look at the customer list: Renesas, GM, Apple, Cirrus Logic," Sopko said. "This is demand-backed capacity, not a build-it-and-they-will-come fab."</p> <section class="section main-article-chapter" data-menu-title="Ownership and capacity allocation remain open questions"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Ownership and capacity allocation remain open questions</h2> <p>The arrangement separates ownership of the fabs from operation of the facilities. Reuters <a href="https://www.reuters.com/business/retail-consumer/japan-invest-up-21-billion-two-us-chip-factories-be-run-by-globalfoundries-2026-10-14/" target="_blank" rel="noopener">reported</a> that GlobalFoundries will earn fees from construction, operations and technology licensing. CEO Tim Breen told Reuters that GlobalFoundries will manage customer relationships and that the company intends to prioritize U.S. and Japanese customers.</p> <p>The U.S. government will also acquire an additional 4% stake in GlobalFoundries itself. Breen told Reuters that the transaction would bring total U.S. government ownership of the chipmaker to approximately 7%.</p> <p>That stake is separate from the government's interest in the new fabs. GlobalFoundries remains majority-owned by Mubadala Investment Company, Abu Dhabi's sovereign investment company.</p> <p>The public announcement did not specify the ownership percentages in the new facilities or explain how the U.S. and Japanese interests will divide governance rights. Capacity allocation could become a point of contention if customers in both countries compete for production during a shortage.</p> <p>The arrangement also leaves questions about how investment returns will be divided and how durable the financing model would be if U.S.-Japan trade terms change. Those details will help determine how much control the two governments exercise over the manufacturing assets and how much operating discretion GlobalFoundries retains.</p> </section> <section class="section main-article-chapter" data-menu-title="Federal funding targets data center technologies"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Federal funding targets data center technologies</h2> <p>Separate from the fab investment, GlobalFoundries agreed with the U.S. Department of Commerce to reallocate $955 million from its existing CHIPS and Science Act award to accelerate investments in silicon photonics, SiGe and advanced packaging at facilities in Malta and Burlington, Vt.</p> <p>These investments target technologies used in communications infrastructure and data centers. Silicon photonics uses optical technology to transmit data, while SiGe supports high-speed electronic applications, including networking. Advanced packaging integrates semiconductor components into more complex systems.</p> <p>Sopko said the two funding streams serve different purposes.</p> <p>"The new fabs serve foundational demand. The reallocated CHIPS money points GF's own investment at AI data center infrastructure," he said. "Japan's capital covers the volume business, and federal dollars go to the higher-value technologies where GF can differentiate."</p> <p>GlobalFoundries has not publicly specified which process technologies the new fabs will use. Sopko expects the facilities could support fully depleted silicon-on-insulator, bipolar-CMOS-DMOS power technologies and embedded nonvolatile memory. Those platforms remain unconfirmed for the project.</p> </section> <section class="section main-article-chapter" data-menu-title="Workforce could constrain the build-out"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Workforce could constrain the build-out</h2> <p>GlobalFoundries expects to break ground on the first fab later this year. The company said its Malta site is permitted and ready for construction, with existing infrastructure expected to accelerate development.</p> <p>Staffing the facilities could prove more difficult than securing capital. The project would add thousands of construction jobs and more than 1,000 high-skilled manufacturing positions in a region already building its semiconductor workforce.</p> <p>"Capital is the easy part. People are the constraint," Sopko said.</p> <p>He pointed to GlobalFoundries' partnership with Rensselaer Polytechnic Institute, formalized in December 2025 after four years of collaboration. About 150 RPI alumni already work at GlobalFoundries, he said.</p> <p>"The real test of this deal isn't the groundbreaking," Sopko said. "It's whether Malta can staff two more fabs on schedule without starving Fab 8 [GlobalFoundries' existing manufacturing facility in Malta]."</p> <p>The first fab's expected 2029 start date leaves several years before the new capacity becomes available. GlobalFoundries must still translate the investment commitment into a staffed manufacturing operation, while the second facility's funding and schedule remain conditional.</p> <p><i>Shane Snider is a senior news writer at </i>TechTarget<i>, covering AI infrastructure, hyperscale data centers, cloud platforms, and the power and energy systems driving modern compute expansion. You can reach Shane at </i><a href="mailto:shane.snider@informatechtarget.com"><i>shane.snider@informatechtarget.com</i></a><i> or on </i><a href="https://www.linkedin.com/in/shane-snider-1727755/" target="_blank" rel="noopener"><i>LinkedIn</i></a><i>.</i></p> </section> Japan's investment could add 1.8 million wafers of annual capacity in New York, but staffing two fabs and allocating production will test the plan's execution. https://cdn.ttgtmedia.com/rms/onlineimages/toolGearArrow_g103332398.jpg https://www.techtarget.com/it-infrastructure/news/366651991/GlobalFoundries-to-run-21B-US-Japan-chip-fab-project Wed, 14 Oct 2026 20:30:00 GMT <p>Gartner predicts that AI infrastructure providers are unlikely to recoup the trillions they plan to spend, adding another variable to an already complex cost calculus for enterprise AI.</p> <p>Published this week, Gartner's new forecast estimated that some $3.5 trillion in <a href="https://www.techtarget.com/it-infrastructure/news/366651336/Report-Anthropic-locks-in-518B-AI-infrastructure-commitments">AI infrastructure investments</a> will ultimately be written off, triggering extreme market consolidation by 2030. This is because AI providers are expected to spend up to $7 trillion on infrastructure by 2030, which would require $8.5 trillion in revenues to cover announced capital spending, financing costs and a 12% return on invested capital. But AI providers are expected to generate only $3.5 trillion in AI revenue over the same period, according to the Gartner's calculations.</p> <p>Will Sommer, senior director analyst at Gartner, said the companies most heavily <a target="_blank" href="https://www.gartner.com/en/newsroom/press-releases/2026-10-14-gartner-says-over-us-dollars-three-and-a-half-trillion-of-artificial-intelligence-infrastructure-spending-could-be-exposed-to-impairment-by-the-end-of-2030" rel="noopener">affected by these trends</a>, should they come to pass, will be a "Cambrian explosion" of neocloud providers, which are providers that arose specifically to handle AI workloads.</p> <p>"We [take the] position that 80% of neoclouds will cease independent operations by 2030," Sommer said. "That doesn't mean that compute assets are just going to disappear -- ceasing independent operations may include acquisition, it may include partnerships, or it may include outright failure. Not every single company that disappears is going to fail."</p> <p>Still, the disappearance of a familiar vendor into another raises significant considerations for CIOs. Weighing the potential longevity of neocloud vendors, along with how they are capitalized, could also sway an ongoing 'build vs buy' dilemma enterprises face when they weigh private AI security, data gravity and potentially lower token costs vs the complexity and costs of maintaining their own AI infrastructure in-house.</p> <p>"For the first time in really a generation, the math is starting to move toward 'build' and away from 'buy' or 'partner.'" Sommer said. "It doesn't mean that 'build' is preferred across the map, but it is starting to make a lot more sense."</p> <section class="section main-article-chapter" data-menu-title="Build, buy – where's the ROI?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Build, buy – where's the ROI?</h2> <p>At the same time, compute resources and data storage are scarce and costly -- and will remain so for at least the near future. One of the ways some organizations split the difference is to use edge devices -- even Mac Minis and laptops with GPU hardware -- for at least some AI workloads, Sommer said.</p> <p>"Running relatively smaller models on local compute, whether that be a laptop that has a relatively souped-up graphics processing unit and CPU or an on-premises server or some combination of those, it's going to be use-case dependent. But certainly we see more opportunity for those types of [deployments], given the nature of AI and the inherent scarcity and cost pressure [of infrastructure]," he said.</p> <p>But evaluating whether to use local hardware isn't just a matter of comparing the cost of equipment with the cost of using a cloud service. There's also the assessment of likely utilization of an on-premises hardware investment and whether lighter-weight models will adequately support a specific application, according to one IT practitioner.</p> <p>Frontier models are better at generating fast responses than the kind of smaller, lighter-weight models that are practical to run on-premises. "The only way to get a good ROI on local models run on your infrastructure is saturating it around the clock," said Steve Koelpin, an AI observability engineer who runs AI models on his own hardware.</p> <p>"So, this approach makes sense when you're not directly in front of the model chatting with it but rather letting it run long independent agentic workflows that don't require babysitting," Koelpin said.</p> <p>Attaining <a href="https://www.techtarget.com/it-strategy/opinion/Enterprise-AI-got-better-but-the-ROI-didnt">real ROI for AI applications</a> and infrastructure remains highly elusive as enterprise AI matures, according to recent market research from other analyst firms.</p> <p>For example, a <a target="_blank" href="https://hyperframeresearch.com/hyperframe-lens-research-3/" rel="noopener">third-quarter survey</a> of more than 500 enterprise and technology business leaders conducted by HyperFrame Research found that 48% of AI and ML projects reached production, but only 27% also met their original business or ROI objectives. That represents an increase compared to the first quarter's survey, which saw just 23% of projects meet their original business or ROI objectives.</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> We have a long way to go before organizations figure out they're spending too much money. </figure> <figcaption> <strong>Rob Strechay,</strong>Lead analyst, VB Intelligence </figcaption> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>In another third-quarter survey of 375 companies by VentureBeat Research, a similar number, 40%, of respondents said they'd tracked AI ROI rigorously. Of that number, only about 10% said they had their finance department's approval and buy-in on that ROI number in a smaller, September portion of the survey with 102 respondents.</p> <p>Only 21.6% of respondents to VentureBeat's September infrastructure survey reported rigorously tracking AI infrastructure ROI. Of that group, just 45.5% had finance sign-off, which was equivalent to only 9.8% of all respondents, said Rob Strechay, lead analyst at VB Intelligence, who conducted the survey.</p> <p>"That's pretty telling that nobody's at that level of maturity yet," Strechay said. "But also, we have a long way to go before organizations figure out they're spending too much money."</p> </section> <section class="section main-article-chapter" data-menu-title="Market research differs on enterprise neocloud interest"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Market research differs on enterprise neocloud interest</h2> <p>Further muddying the investment outlook for CIOs is the fact that market studies sharply differ on how widely used neoclouds are <a href="https://www.computerweekly.com/feature/Weighing-up-the-enterprise-risks-of-neocloud-providers">among enterprise companies</a>. HyperFrame's survey found that neoclouds were in use by just 2% of respondents, with a projected growth to 8%; by comparison, AWS was in use by 59% of respondents and Azure by 51%.</p> <p>VentureBeat's study, however, found that 42.1% of respondents were interested in doing more with neoclouds.</p> <p>Neoclouds, which sprang up to meet outsized demand for AI infrastructure, play an important role in cases where <a href="https://www.techtarget.com/cybersecurity/feature/The-push-for-digital-sovereignty-What-CISOs-need-to-know">digital sovereignty</a> is a priority for enterprise buyers. Even if the industry dramatically consolidates by 2030, that leaves plenty of short-term opportunities for enterprises to source AI infrastructure, Strechay said.</p> <p>"It's a question CIOs should be asking of their neoclouds: how are they capitalized, what does that mean over the long term?" Strechay said. "I think most of these companies that are going to the neoclouds, it's such a two- or three-year deal kind of thing. … It's only 4 years away, but 2030 feels like it's 10 years away in AI terms."</p> <p>Hyperscalers, meanwhile, are much more diversified than neocloud vendors. Still, they are not immune to potential write-downs on huge infrastructure investments, Sommer said.</p> <p>"Whether or not all of the hyperscalers end up winning this market, I think it is still TBD," he said. "It may be that some of these hyperscalers or near-hyperscalers compete, don't win sufficient market share to remain a GPU-as-a-service provider, and pivot to something else."</p> <p>While the big three clouds can absorb AI infrastructure losses, some second-tier cloud infrastructure vendors might also be subject to consolidation, Strechay said.</p> <p>"Some of them are private-equity-backed as well, and so their problem is how much money they have to go after and take [another cloud] private," Strechay said. "As the price of these AI data centers comes down, there has to be consolidation. There's just no way all of these will last."</p> <p><em>Beth Pariseau, senior news writer for Informa TechTarget, is an award-winning veteran of IT journalism. Have a tip?&nbsp;</em><a target="_blank" href="mailto:bpariseau@techtarget.com?subject=News%20tip" rel="noopener"><em>Email her</em></a><em> or connect on </em><a target="_blank" href="https://www.linkedin.com/in/bethpariseau" rel="noopener"><em>LinkedIn</em></a><em>.</em></p> </section> Gartner forecasts that up to 80% of neoclouds will cease independent operations by 2030, as projected AI infrastructure costs exceed projected revenue. https://cdn.ttgtmedia.com/rms/onlineimages/cloud_g1213812869.jpg https://www.techtarget.com/it-infrastructure/news/366652181/Gartner-forecasts-massive-AI-infrastructure-write-downs Wed, 14 Oct 2026 17:33:00 GMT <p>LONDON -- Agentic AI is changing business operations, but the pressure to keep up with rapid advancements means many are using tools without knowing where they will lead.</p> <p>That was a central theme at IBM’s Think on Tour EMEA 2026 conference on Wednesday, where executives argued that trust, governance and oversight are now as important to business outcomes as transformation itself.</p> <p>In her keynote, Ana Paula Assis, senior vice president and chair for EMEA and APAC at IBM, said 80% of executives expect AI to drive revenue growth, yet only 20% can explain exactly how that will happen.</p> <p>The stakes are rising as businesses move beyond chatbots toward <a href="https://www.techtarget.com/ai/news/366651790/Meta-joins-enterprise-agent-race-with-Muse-Wonder-facing-a-trust-gap">agents capable of carrying out tasks</a> and making decisions with less human intervention. IBM research presented at the event found that three-quarters of surveyed executives across EMEA expected an agent to execute a complete business task and make a decision without human intervention by 2027. Meanwhile, 80% said at least one <a href="https://www.techtarget.com/ai/news/366651355/While-agents-cause-trouble-calls-for-AI-regulation-mount-from-Anthropic-lawmakers">AI-related problem</a> had reached the boardroom or C-suite.</p> <p>“The question now is: Can we operate it at scale safely, reliably?” Assis said. “That is not a technology problem. That is an operating problem, and it requires a fundamentally different, new AI operating model.”</p> <section class="section main-article-chapter" data-menu-title="Applying AI without losing control"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Applying AI without losing control</h2> <p>Michael Nimhauser, global head of sovereignty at <a target="_blank" href="https://www.google.com/url?sa=i&amp;source=web&amp;rct=j&amp;url=https://atos.net/en/blog/digital-sovereignty-for-local-authorities&amp;ved=2ahUKEwje9pfXprqXAxVpM1kFHcRdK_MQy_kOegoIAggACAEIDhAE&amp;opi=89978449&amp;cd&amp;psig=AOvVaw2tvF4jUHtG-3CL6vSNv-xQ&amp;ust=1792094578236000" rel="noopener">Atos</a>, a France-based IT services and consulting firm, said organizations were attracted by AI’s potential but worried about losing control as adoption accelerated.</p> <p>“With all of the stars in their eyes and the potential, they also have a lot of fear of loss of control,” he said.</p> <p>That concern extends beyond where data is stored to include control over infrastructure, models and the systems coordinating AI agents. To combat this, Nimhauser (alongside other speakers at the conference) argued that <a href="https://www.techtarget.com/ai/news/366651997/Governments-prepare-for-the-next-wave-of-AI">governance</a> must be built into the architecture from the outset.</p> <p>“It needs to be by design,” he said. “It cannot be an afterthought.”</p> <p>Celerity, an IBM partner that uses IBM’s <a href="https://aibusiness.com/agentic-ai/bob-ibm-ai-coding-assistant">AI-powered software development tool Bob internally</a>, has taken a parallel approach to AI adoption and governance. The company assesses applications for their potential business value and ROI, while a separate governance review considers risk, data and access.</p> <p>Craig Aston, CEO of the U.K.-based managed services provider, said businesses need to encourage employees to identify opportunities to use AI, without allowing enthusiasm to outpace appropriate controls.</p> <p>“Having governance run alongside the adoption for me is the most important point,” he told <i>TechTarget</i>. “It’s really important that we don’t quell the energy of innovation, and we really do want our teams to come to us with ideas. But it does ultimately come down to risk and, if you get it wrong, it could bring a company down.”</p> </section> <section class="section main-article-chapter" data-menu-title="Keeping humans involved"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Keeping humans involved</h2> <p>U.K. insurer Markerstudy has also been deploying Bob, using it to help assess the risks and options involved in modernizing legacy technology.</p> <p>The company used Bob to analyze six million lines of code in a core insurance system it built over roughly 30 years. The review took four weeks and produced a heat map to help identify which parts of the system were most relevant to its future needs.</p> <p>Previously, Markerstudy could examine only sections of the code and extrapolate from them. The wider analysis helped inform a targeted <a href="https://www.techtarget.com/ai/news/366585946/New-IBM-Watsonx-GenAI-focuses-on-enterprises-governance">modernization program</a>, rather than a wholesale replacement. CTO Dan Fiehn noted, however, that the process did not entirely place decisions in the hands of AI.</p> <p>“We kept our team in the loop throughout the process,” Fiehn told <i>TechTarget</i>. “It wasn’t a ‘fire-and-forget.’ We had people validating output and training the models. But what Bob did do here is move something that was quite a subjective decision to make it quite objective.”</p> <p>For Fiehn, the objective was not simply to deploy more AI, but to use it to improve decision-making. “We don't think it's about who deploys the most AI wins,” he said. “We feel it's more about delivering measured outcomes that give better decisions for our customers and our colleagues.”</p> </section> <section class="section main-article-chapter" data-menu-title="IBM’s answer: a new operating model"> <h2 class="section-title"><i class="icon" data-icon="1"></i>IBM’s answer: a new operating model</h2> <p>IBM is positioning its hybrid cloud and AI offerings as part of its response. Assis argued that AI initiatives can struggle when data is isolated and infrastructure fragmented.</p> <p>“There is no good AI without <a href="https://www.techtarget.com/ai/news/366651628/The-context-factor-for-AI-agents">good data</a>,” she said.</p> <p>IBM has introduced Sovereign Core, an open source platform intended to give organizations greater control over the deployment and operation of AI environments. The company is positioning it as one option for businesses seeking to manage data, operations and governance as AI adoption grows.</p> <p>Regardless of the platform they choose, businesses will need to embed governance and controls into their AI strategies from the outset, conference attendees said.</p> <p>“AI is no longer just supporting the business,” Assis said. “It has become the business model. The organizations pulling ahead are not just adding AI, but rethinking how systems, decisions and processes work together with AI.”</p> <p><i>Scarlett Evans is a freelance writer with a focus on robotics and emerging technologies. Previously, she was assistant editor at IoT World Today, where she specialized in robotics and smart city technologies. Scarlett also has a background in the mining and resources sector, with experience at Mine Australia, Mine Technology and Power Technology.</i></p> </section> As AI agents move from answering questions to making decisions and acting, businesses face a growing challenge of keeping accountability in step with the pace of change. https://cdn.ttgtmedia.com/rms/onlineimages/ai_a264431831.jpg https://www.techtarget.com/ai/news/366651989/IBM-Agentic-AI-is-reshaping-governance-controls-and-trust Wed, 14 Oct 2026 16:26:00 GMT <p>Identifying underused capacity does not provide data center operators with guidance on how to use it. The answer depends on why the capacity is sitting unused.</p> <p>Two facilities can show similar utilization levels for completely different reasons. One could have resources reserved beyond what its applications need, while another could have enough compute, but a workload that cannot use it efficiently.</p> <p>Siraj Aziz, analyst for cloud and data center at Omdia, a division of Informa TechTarget, pointed to workload optimization, predictive automation and better visibility as ways organizations can improve capacity use. But those approaches only help when the underlying capacity can be recovered.</p> <p>For infrastructure leaders, the question is where capacity can be put to productive use with the infrastructure already in place, and where capacity requires a larger investment or a different use.</p> <section class="section main-article-chapter" data-menu-title="Find out why the capacity is underused"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Find out why the capacity is underused</h2> <p>Aziz said one challenge is that infrastructure visibility remains fragmented.</p> <p>Infrastructure teams use different systems to monitor networks, AI workloads, cloud resources and applications. However, <a href="https://www.techtarget.com/it-infrastructure/tip/AI-capacity-planning-Balancing-flexibility-performance-and-risk">physical infrastructure capacity</a> is tracked separately. The individual systems show what is happening within their own layer without explaining why the specific infrastructure space remains underused.</p> <p>For instance, a platform team can see an idle CPU. At the same time, an application is unable to scale further, or facilities teams can know that a row has reached its power limit even though a capacity dashboard shows empty rack space.</p> <p>The first step is to connect those discrete views to the workload. Operators need to understand what resources the application requires, what it consumes and what prevents additional work from being placed. This includes compute, memory, storage and networking, as well as available rack power and cooling, all of which affect placement.</p> <p>Low utilization can then be broken down into different problems. Capacity can be overprovisioned, poorly allocated or limited by the application itself. For example, a physical constraint, such as limited space, can prevent additional deployment. The available infrastructure can also be technically functional but unsuitable for the target workload.</p> <p>Operators also must distinguish between unused capacity and capacity kept in reserve. Some infrastructure is intentionally left available for failover, maintenance or predictable workload peaks. If the capacity is reclaimed, it can reduce resilience in the target application.</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> The objective is to identify which resources can be recovered without creating another constraint. </figure> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>Application architecture can make the distinction harder. Aziz noted that <a href="https://www.techtarget.com/it-infrastructure/tutorial/How-to-set-up-autoscaling-in-Azure-Kubernetes-Service">horizontal pod autoscaling</a> depends on the application being sized and designed so that additional replicas can increase useful work. Predefined hardware profiles can also restrict changes to resource allocation and power states.</p> <p>The objective is to identify which resources can be recovered without creating another constraint.</p> </section> <section class="section main-article-chapter" data-menu-title="Optimize and rightsize existing capacity"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Optimize and rightsize existing capacity</h2> <p>Once operators understand why capacity is underused, they can determine how much additional workload the existing infrastructure can support.</p> <p>Aziz described predictive automation, proactive scaling and <a href="https://www.techtarget.com/it-infrastructure/feature/The-great-workload-reshuffle-Choices-for-AI-and-analytics">workload optimization</a> as important characteristics organizations must have to manage capacity effectively. Operators can adjust capacity as workload demand changes and identify resources that have been allocated beyond what the application requires.</p> <p>Rightsizing is one way to do that, in which a workload can reserve more CPU than it consumes, leaving those resources unavailable to other applications even when the hardware is lightly used. Adjusting those allocations can free up capacity, but the sizing decision must consider workload peaks, latency requirements and failover conditions.</p> <p>Autoscaling extends the same idea to changing demand. Horizontal pod autoscaling can add or reduce application replicas, while <a href="https://www.techtarget.com/it-infrastructure/tip/Kubernetes-automation-Use-cases-and-tools-to-know">cluster scaling</a> can change the worker capacity available to those applications. But scaling creates additional useful capacity when the application can scale with it.</p> <p>The location of resources can also strain capacity. For example, a cluster can have sufficient total CPU or accelerator capacity but still be unable to place another workload because resources are spread across different hosts. Operators must schedule and consolidate to restore those resources to a usable state.</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> The cheapest source of new capacity can sometimes be capacity that the organization has already paid for. </figure> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>Google <a href="https://research.google/pubs/gass-gpu-automated-sharing-at-scale/" target="_blank" rel="noopener">found</a> that automated GPU sharing increased aggregate GPU throughput by 50% and enabled the evaluated cluster to support 19% more GPU jobs. A GPU assigned to one workload can have execution time available for another compatible workload.</p> <p>The cheapest source of new capacity can sometimes be capacity that the organization has already paid for.</p> </section> <section class="section main-article-chapter" data-menu-title="Decide whether to modernize, repurpose or walk away"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Decide whether to modernize, repurpose or walk away</h2> <p>Optimization has limits. Capacity that remains underused after rightsizing, scheduling and other operational changes may require a different decision, as the existing facility no longer matches the workload.</p> <p>Bruce Bateman, chief analyst for semiconductors at Omdia, argued that this decision should begin with demand. When discussing an operator that already has unused capacity, he summarized the issue as "It's demand, it's not compute." Bateman said that adding newer infrastructure does not create the workload demand. The operator must determine who will consume the additional capacity and what that workload requires.</p> <p>The next critical question is how much of the existing facility must change to support it.</p> <p>In the case of a single constraint, operators can justify a targeted upgrade. But the economics change when one upgrade triggers several others. Bateman explained that newer compute can require additional storage, higher-capacity routers and internet connectivity. Higher-density systems will introduce new power and cooling requirements.</p> <p>Leaders need to calculate the complete investment required to make the target workload usable. If a compute upgrade requires modifications to electrical distribution, cooling, networking, storage and building systems, the operator must decide whether to convert the facility to a different workload class.</p> <p>Bateman said that workload conversion is not always necessary. He added that a facility already supporting enterprise applications, databases, hosting or private cloud workloads can continue serving those customers if the economics remain attractive. Older infrastructure does not lose its value because it <a href="https://www.datacenterknowledge.com/cooling/for-high-density-ai-available-data-center-space-may-not-be-usable">cannot support the highest-density AI systems</a>.</p> <p>But in supporting the boom in AI systems, Bateman said inference is one possibility for operators who are not positioned to support large training environments. Private AI, regional inference and other local workloads can fit some facilities where customers value connectivity, locality, privacy and control.</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> The decision is to determine which workload offers the strongest business case for the existing infrastructure and how much additional investment that workload deserves. </figure> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>Building inference capacity because the broader AI market is growing creates the same risk as any other speculative upgrade.</p> <p>Bateman said that the broader idea is to "stay in your lane." In this context, the phrase does not suggest avoiding modernization, but points to understanding what the facility can support economically before trying to turn it into something fundamentally different.</p> <p>The decision is to determine which workload offers the strongest business case for the existing infrastructure and how much additional investment that workload deserves.</p> <p><i>Abhishek Jadhav is a technology journalist covering AI infrastructure, semiconductors and advanced computing systems.</i></p> </section> Data center capacity often goes underused due to fragmented visibility. Understanding resource needs and optimizing workloads can enhance efficiency and reclaim valuable capacity. https://cdn.ttgtmedia.com/rms/onlineimages/container_g922017556.jpg https://www.techtarget.com/it-infrastructure/feature/How-to-gain-value-from-underutilized-data-center-capacity Wed, 14 Oct 2026 15:57:00 GMT <p>Enterprise security teams have long assumed that the more widely deployed and heavily scrutinized an open-source library is, the less likely it is to harbor fundamental flaws. The conclusion of a yearlong AI code audit of Curl shatters that assumption.</p> <p>On Oct. 14, Curl maintainers released <a href="https://daniel.haxx.se/blog/2026/10/14/curl-8-23-0/">version 8.23.0</a>, fixing 24 total security vulnerabilities -- 18 of which stem from an extended AI-powered audit by security firm Aisle. Across the entire engagement, Aisle's models unearthed 35 CVEs and 160 bugs.</p> <p>Chief among the security firm's discoveries was <a href="https://nvd.nist.gov/vuln/detail/CVE-2026-8932" target="_blank" rel="noopener">CVE-2026-8932</a>, a 25-year-old authentication bypass lurking in Curl's connection pool since February 2001 that had survived decades of manual code reviews, static analysis and bug bounties.</p> <p>For enterprise security teams, the release is an immediate reminder to check whether affected versions of Curl are running across their environments. However, it's also a stark reminder that as AI drastically lowers the cost of vulnerability discovery, deep open-source dependencies long assumed to be "battle-tested" can no longer hide behind their age.</p> <section class="section main-article-chapter" data-menu-title="Hidden debt in the network stack"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Hidden debt in the network stack</h2> <p>Curl develops and secures software that runs on more than 20 billion devices. Curl and its underlying Libcurl library handle data transfers across networks and are embedded in software ranging from operating systems and cloud services to network appliances, making vulnerabilities in the project a potential concern across the enterprise <a href="https://www.techtarget.com/it-infrastructure/tip/Managing-open-source-components-in-the-software-supply-chain">software supply chain</a>.</p> <p>In its yearlong audit of Curl, the most surprising discovery was a vulnerability hidden in the software for 25 years. Introduced in February 2001, the flaw (CVE-2026-8932) stems from how Libcurl reused cached connections when an application changed client credentials or mTLS configurations. Under specific conditions, an attacker could hijack an existing connection pool entry to bypass authentication controls and execute requests under another user's identity.</p> <p>Beyond authentication bypasses, Aisle's audit revealed flaws that could allow server impersonation, cause transfers to inadvertently inherit security settings from prior sessions and allow requests to persist using revoked credentials.</p> <p>Curl founder and CEO Daniel Stenberg, who <a href="https://www.techtarget.com/cybersecurity/news/366652109/Bug-bounty-reckoning-AI-slop-forces-enterprise-strategy-shift">suspended the project's bug bounty program</a> at the beginning of 2026, had grown increasingly skeptical of the quality of reports he was getting. The program had paid out more than $100,000 for 87 confirmed vulnerabilities over the past seven years.</p> <p>However, the validation of Aisle’s audit findings marked a turning point.</p> <p>"The key takeaway [of the report] is that modern AI-powered code analyzers find flaws in ways nothing did previously," he told <i>TechTarget</i>. "The challenge now is how to deal with all of the findings and fix them."</p> </section> <section class="section main-article-chapter" data-menu-title="The visibility blind spot"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The visibility blind spot</h2> <p>For enterprise security teams, the challenge goes beyond keeping up with newly discovered vulnerabilities. They also need to know where the affected utilities are running across their environment.</p> <p>James Morrison, security sales strategist at Dataprise and a former CISO and FBI cyber investigator, says organizations often lack that visibility due to <a href="https://www.techtarget.com/it-strategy/tip/17-potential-costs-of-shadow-IT">shadow IT</a>, when employees introduce unsanctioned software or developers incorporate open-source code into larger projects.</p> <p>"Because of the nature of shadow IT, if the corporate security team does not have visibility across the enterprise, the unknown software and its vulnerability are only found when the vulnerability is exploited," Morrison said.</p> <p>The problem doesn't end when an audit identifies a vulnerability. Morrison noted that organizations often put off patching less serious vulnerabilities, then fail to revisit those decisions as new threats emerge. He recommends taking a fresh look at vulnerabilities with each scan rather than assuming flaws previously considered minor remain low risk.</p> <p>"I have found it best to delete all preconceptions from previous periodic scans and treat every vulnerability found as if it was a brand-new vulnerability," Morrison said.</p> <p><a href="https://www.techtarget.com/cybersecurity/video/AI-kill-switch-Can-we-really-shut-down-artificial-intelligence">Artificial intelligence</a> can help by connecting newly disclosed vulnerabilities with current threat information to identify flaws that might pose a greater risk than companies initially realized, he said.</p> </section> <section class="section main-article-chapter" data-menu-title="Race to remediation"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Race to remediation</h2> <p>Detecting the connection pool vulnerability required analyzing dynamic behavior that traditional static analysis tools routinely miss. Stan Fort, founder, chief scientist and CTO at Aisle, says the bug only appeared when the software was used in a particular way, making it more difficult for conventional testing to catch it.</p> <p>To filter out AI-generated false positives, Aisle tested each potential vulnerability against the current code and assessed its impact. Then the company had a security researcher review the findings. After that, Curl's own security team independently reproduced the reports and decided which warranted CVEs.</p> <p>As AI models scale vulnerability research, the fundamental challenge for software maintainers and enterprise defenders is <a href="https://www.techtarget.com/cybersecurity/opinion/Top-vulnerability-management-challenges-for-organizations">shifting from finding bugs to triaging them</a>.</p> <p>"As discovery becomes cheaper, the main constraint will be validating and fixing findings quickly enough," Fort said.</p> <p>For Stenberg, the yearlong audit marks a permanent shift in software maintenance and open-source security governance.</p> <p>"The tables have turned," he said. "In the past, finding each flaw was a slow and hard process only done by a few skilled humans."</p> </section> An AI audit of Curl unearthed a 25-year-old authentication flaw in code powering 20 billion devices, proving heavily scrutinized legacy libraries still harbor hidden risks. https://cdn.ttgtmedia.com/rms/onlineimages/toolGearArrow_g1159416182.jpg https://www.techtarget.com/cybersecurity/news/366652233/Historic-Curl-flaw-shows-AI-risk-to-legacy-supply-chain Wed, 14 Oct 2026 15:52:00 GMT <p>The CMS wants to expand site-neutral payments in Medicare, but just how the agency sets those rates could matter more than it currently does, according to a recent <a href="https://jamanetwork.com/journals/jama-health-forum/fullarticle/2854819">research brief</a> from Johns Hopkins University.</p> <p>The research published in <i>JAMA Health Forum</i> suggests a true site-neutral payment policy that pays the same rates regardless of where patients receive care.</p> <p>In a bid to cut costs, Medicare currently Medicare pays hospital outpatient departments a reduced rate for certain services, such as clinic visits and drug administrations, when they can be safely performed in a doctor's office. That rate is generally about 40% of the standard hospital outpatient price.</p> <p>Th research brief looked at how well Medicare's current method for setting these payments works. The researchers used real 2024 Medicare claims data and 2026 payment rates to compare two methods for calculating site-neutral payments.</p> <p>The current method would generate the Medicare savings <a href="https://www.techtarget.com/revcyclemanagement/news/366634660/Site-neutral-pay-340B-clawback-to-offset-new-26-OPPS-boost">that the CMS intends.</a> The Johns Hopkins researchers found potential savings of $6.6 billion this year across 57 procedure groups studied, including those that are already paid at a site-neutral rate and those the CMS wants to make site-neutral next year.</p> <p>These savings would be similar to the $6.5 billion in savings if Medicare truly aligned rates between hospital outpatient departments and physician offices, the study showed.</p> <p>Even though the methods for cutting costs yield similar overall totals, researchers found that they don't align at the level of individual procedures.</p> <p>For example, paying hospitals at the physician's rate would result in lower payments for 13 of the 23 procedure categories, although the research did not specific which procedure categories would be affected. The true site-neutral payment would exceed the 40%-rate amount in four procedure categories.</p> <p>Researchers also noted that hospitals would actually get paid more and Medicare would save less with truly aligned payment rates compared to the 40% rate for clinic visits, drug administration, and imaging without contrast. Medicare already pays hospitals at the reduced rate for clinic visits and drug administration, while <a href="https://www.techtarget.com/revcyclemanagement/news/366645250/CMS-takes-aim-at-340B-site-neural-payments-in-OPPS-proposal">the CMS has proposed extending those rates to the latter category next year</a>.</p> <p>Site-neutrality has gained momentum since its humble start in 2015, when policies limited the number of services and hospital outpatient departments subject to the altered rates. Lawmakers are also pushing legislation like the Same Care, Lower Cost Act (<a href="https://www.congress.gov/bill/119th-congress/senate-bill/1629">S. 1629</a>), introduced by Sen. John Kennedy (R-LA), to more broadly align Medicare rates.</p> <p>Commercial insurers, like <a href="https://www.healthcaredive.com/news/elevance-cracks-down-on-hospital-billing-for-off-campus-care/831802/">Elevance Health</a>, have also recently implemented policies to curb some payments to hospital outpatient departments.</p> <p>However, hospital lobbyists have opposed site-neutrality. They argue that the <a href="https://www.aha.org/system/files/media/file/2024/06/Site-Neutral-Myth-vs-Fact-Infographic.pdf">reimbursement cuts could force hospitals to stop some outpatient services</a>, especially since they have higher fixed and operating costs, treat sicker and more vulnerable populations and face stricter compliance requirements.</p> <p>Still, the findings underscore the savings associated with aligning hospital outpatient and physician rates. The question that remains is: How should Medicare proceed with site-neutrality expansions?</p> <p>Johns Hopkins researchers suggested it's time for a new approach, considering the implementation challenges of reconciling the different rates. They recommended site-specific rates based on the Medicare Physician Fee Schedule, which would unmask those service-level variations for a more true site-neutral policy.</p> <p>The CMS has acknowledged these issues, <a href="https://www.techtarget.com/revcyclemanagement/news/366645846/Physician-pay-cuts-loom-as-CMS-unveils-major-Medicare-reforms">seeking feedback earlier this year</a> on ways for the fee schedule to better reflect resource costs across settings. Researchers said this signals "a longer-term effort to harmonize Medicare payment for ambulatory care."</p> <p><em>Jacqueline LaPointe is an Executive Editor at Xtelligent Healthcare Media, covering revenue cycle management, healthcare payers, health policy, and health IT since 2016.&nbsp;</em></p> Site-neutrality is now a question of when, not if, and research shows that how Medicare calculates the rates could affect some service lines more, despite similar overall savings. https://cdn.ttgtmedia.com/rms/onlineimages/medical%20revenue_g1366221395.jpg https://www.techtarget.com/revcyclemanagement/news/366652138/How-Medicare-calculates-site-neutral-rates-matters-research-says Wed, 14 Oct 2026 15:34:00 GMT <p><iframe title="Musing about Muse’s new supporting AI models, unruly DeepSeek models" allowtransparency="true" height="150" width="100%" style="border: none; min-width: min(100%, 430px); height: 150px;" scrolling="no" data-name="pb-iframe-player" src="https://www.podbean.com/player-v2/?i=vz9ta-1b84ac1-pb&amp;from=pb6admin&amp;share=1&amp;download=1&amp;rtl=0&amp;fonts=Arial&amp;skin=f6f6f6&amp;font-color=auto&amp;logo_link=episode_page&amp;btn-skin=2baf9e" loading="lazy"></iframe></p> <p>On <i>Targeting AI in the Moment</i> (<i>AIM</i>), the weekly news podcast from <a href="https://www.techtarget.com/ai/news/366651355/While-agents-cause-trouble-calls-for-AI-regulation-mount-from-Anthropic-lawmakers"><i>TechTarget.com</i></a><i>, </i>this week’s co-hosts chat about the latest developments in the turbulent world of generative AI.</p> <p>AI news writer <a href="https://aibusiness.com/author/esther-shittu">Esther Shittu</a> and her editor, Shaun Sutner, report the news on the sites and offer more analysis, perspective and context on recent AI developments in this short-form news podcast series.</p> <p>Check out these new stories:</p> <ul class="default-list"> <li><a href="https://www.techtarget.com/ai/news/366651790/Meta-joins-enterprise-agent-race-with-Muse-Wonder-facing-a-trust-gap">Meta joins enterprise agent race with Muse Wonder, facing a trust gap</a></li> <li><a href="https://www.techtarget.com/ai/news/366652106/DeepSeek-and-what-enterprises-should-learn-from-AI-agent-exploits">DeepSeek and what enterprises should learn from AI agent exploits</a></li> </ul> <p><i>Shaun Sutner is senior news director and lead editor for </i>TechTarget<i>’s AI &amp; Emerging Tech news coverage. He is a veteran journalist with more than 30 years of news experience. Esther Shittu is a </i>TechTarget<i> news writer and podcast producer and host who covers AI software and systems. Together, they host the </i>Targeting AI <i>podcast.</i></p> The social media giant tries to capitalize on, monetize its successful personal agent and maybe crack the business market. Chinese models go astray, like their U.S. counterparts. https://cdn.ttgtmedia.com/rms/onlineimages/ai_a279596285.jpg https://www.techtarget.com/ai/news/366652216/Musing-about-Muses-new-supporting-AI-models-unruly-DeepSeek-models Wed, 14 Oct 2026 15:12:00 GMT <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Executive summary</h3> <ul class="default-list"> <li>Anthropic's expected November IPO could surpass SpaceX's $1.77 trillion valuation, becoming the largest-ever and largest pure-play AI IPO.</li> <li>OpenAI is delaying its own IPO, citing safety concerns -- a move analysts read as either a wait-and-watch strategy or a sign of weaker confidence in its financials.</li> <li>Rising infrastructure costs and investor pressure could push AI pricing up, though strong competition and falling unit costs may limit broad price hikes.</li> <li>Enterprises should focus on controllable fundamentals -- business rules, data management and workforce upskilling -- to stay resilient regardless of which AI vendor leads the market.</li> </ul> </div> </div> <p>Anthropic's pending IPO could change the market's competitive dynamics, intensify the financial spotlight and influence AI model pricing.</p> <p>The expected November debut will be a watershed event for AI and the broader technology sector. Anthropic could surpass SpaceX as the largest IPO ever, if the $2 trillion valuation reported in various media outlets comes to pass. SpaceX went public in June at a valuation of $1.77 trillion. Anthropic would likely mark the largest pure-play AI IPO. SpaceX's offerings include reusable launch vehicles, connectivity services and a social media platform -- in addition to AI.</p> <p>Speculation around Anthropic's IPO comes at a time when AI's ability to translate massive investment into meaningful returns comes under increased scrutiny. <a href="https://www.techtarget.com/ai/news/366649605/The-AI-safety-crunch-and-how-enterprises-should-deal-with-it">Concerns about the technology's safety</a> have also grown.</p> <section class="section main-article-chapter" data-menu-title="Assessing competitive dynamics: Anthropic vs. OpenAI"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Assessing competitive dynamics: Anthropic vs. OpenAI</h2> <p>While Anthropic seems willing to launch in this environment, one of its top rivals, OpenAI, will wait it out for now. Sam Altman, CEO at OpenAI, told Fortune that the company won't proceed with an IPO this year, citing "everything happening with safety."</p> <p>Dean Lyulkin, a capital markets expert and CEO at Cardiff, a small business lender, said investors are likely to read the IPO timing in two ways: one that flatters OpenAI and one that doesn't.</p> <p>"The flattering version has precedent in payments," he said. "Mastercard went public ahead of Visa and helped establish a valuation for payment networks. Visa came later into a market that already got its economics."</p> <p>The first version points to the benefits of having a competitor warm up and educate the market. The other version focuses on the willingness to undergo heightened financial scrutiny.</p> <p>"OpenAI may simply not have enough confidence in its own story and economics in relative terms," Lyulkin said. "An IPO would give investors much greater visibility into compute costs, margins, capital requirements and customer growth. When your closest competitor is apparently prepared to subject itself to that scrutiny and you aren't, it's reasonable for investors to ask serious questions."</p> <p>That scrutiny has already begun. According to a leaked prospectus, Anthropic had 2025 revenue of $4.6 billion, but a net loss of $42 billion and an operating loss of $8.06 billion.</p> <p>Jacob Krell, senior director, secure AI solutions and cybersecurity, at Suzu Labs, a cybersecurity and AI firm, cited other data points from the prospectus: $7.33 billion in compute and infrastructure consumption in 2025 -- more than half of operating expenses -- and <a href="https://www.techtarget.com/it-infrastructure/news/366651336/Report-Anthropic-locks-in-518B-AI-infrastructure-commitments">$518 billion in future commitments</a> for cloud, computing and infrastructure. He said those figures describe a company that's scaling its cost base before showing the capability to produce enough margin to pay for the systems behind them.&nbsp;</p> <p>That Anthropic's projected valuation surpasses that of SpaceX suggests the market's emphasis on AI models, Krell said, noting SpaceX's vertically integrated data centers, its Starship program, Starlink and X social media applications.</p> <p>"I guess Wall Street is kind of predicting that the models are going to be worth so much more than the actual infrastructure," Krell said. "But that's not really how it's panning out, because so much of this spend is going to the infrastructure. There is a lot to model development, but it's dwarfed by the actual infrastructure spend."</p> <p>Damon Small, board member at Xcape, a cybersecurity and managed services provider, said he sees the AI investment climate tracking with the 1990s dot-com bubble. The enthusiasm and investment surrounding the dot-com era produced a wave of companies whose business models fell short of investor expectations, he said. Many failed. But the technology proved valuable and ultimately transformed the economy, he added.</p> <p>Small believes <a href="https://www.techtarget.com/it-strategy/feature/The-AI-hype-bubble-might-parallel-the-dot-com-era-bust">AI will follow a similar trajectory</a>. AI providers confront the same challenge as the dot-coms: how to parlay the potential of a world-changing technology into sustainable, profitable businesses, he noted.</p> <p>"If you have net losses that are an order of magnitude greater than your revenue, when are the investors going to get that money back?" Small asked. "I think that's put us in a bit of a precarious situation."</p> </section> <section class="section main-article-chapter" data-menu-title="Weighing the effects on enterprise AI customers"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Weighing the effects on enterprise AI customers</h2> <p>Spotlighting the frontier model providers' financial performance will eventually affect enterprise AI customers. The pressure to meet investor expectations is likely to alter providers' pricing, product lineups and other elements of business strategy, industry executives said.</p> <p>Model providers must deal with pressure from a couple of angles. Their supply-side costs could increase as they push to purchase the same infrastructure components. At the same time, they will need to optimize shareholder value, Krell noted.</p> <p>Those dual pincers will "have them try to extract as much as they can from the customers," Krell said. "We could potentially see substantial price increases."</p> <p>Krell pointed to one particular concern: AI startups that base their entire business model around the frontier AI supply chain.</p> <p>"They're basically wrapping the frontier AI calls in their software," he said. "If their [frontier AI companies] prices rise, they'll have to pass their prices to consumers."</p> <p>Chris Ebert, senior vice president of go-to-market at SCG, a company that advises businesses on AI use and cost-control strategies, questioned the prospect of price hikes.</p> <p>"I wouldn't assume a public listing automatically means higher prices," he said. "Competition among AI providers is strong, and unit costs have generally come down."</p> <p>Ebert said the bigger issue for most companies is the total <a href="https://www.techtarget.com/it-strategy/feature/The-hidden-costs-of-AI-What-leaders-must-budget">cost of AI</a> as usage grows, rather than the price of an individual API call.</p> <p>Small said he would advise AI companies to keep their entry price points as low as possible and offer an expanded suite of products or services that would demand premium pricing. Model vendors already offer pro-level tiers, with OpenAI launching a $500 per month Pro 500 plan in September.</p> <p>Newly public AI companies, however, might simultaneously employ multiple approaches -- price increases, a broader array of product tiers and model options that encourage cost-effective consumption. Vladimir Lukić, global leader for Boston Consulting Group's Tech and Digital Advantage practice, said AI providers will experiment with such measures and determine which gives them the greatest yield.</p> <p>In one tactic, AI providers already offer native services that help customers choose different models based on the task to be performed -- summarization versus deep research, for instance. Increasingly sophisticated users can evaluate token prices across various prompt types and make cost-conscious decisions.</p> <p>"The LLM providers are now starting to learn and say, it's in my interest for you to do the effective things and be prudent on the cost side," Lukić said. "Because if I do it well, the consumption overall will go up."</p> <p>Ebert also cited the importance of matching the model to the task.</p> <p>"The question companies should be asking isn't which model is the cheapest, but which approach delivers the best business outcome at an acceptable total cost," he said.</p> </section> <section class="section main-article-chapter" data-menu-title="Dealing with uncertainty"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Dealing with uncertainty</h2> <p>The prospect of bellwether AI IPOs signals a new financial landscape that will shape model economics and potentially trigger a vendor shakeout. Against that backdrop, what can enterprise leaders do to prepare as they scale AI agents and pursue other strategic deployments?</p> <p>Lukić said businesses must attend to the essentials. In the case of agentic AI, the <a href="https://www.techtarget.com/data-technologies/news/366646252/Context-is-king-as-agents-evolve-but-problems-persist">context layer is a key point of control</a>. That's where an enterprise's business rules reside to harden the models or direct the agents, he noted.</p> <p>"Keep focusing on that," Lukić said, noting the context layer transcends the specific frontier model in use.</p> <p>Indeed, BCG's Applied AI Index 2026, published in September, said companies should maintain their "enterprise cortex" -- which includes business rules and operating context -- under their exclusive control, emphasizing flexibility across models and providers.</p> <p>Lukić also highlighted the importance of managing data, managing the linkage between core systems and storage systems, and differentiating between systems of record and analytics layers. Upskilling the workforce to build, manage and evolve agents is also a key step, he said.</p> <p>"Those are no-regrets moves that, regardless of which of the LLMs is around two or three years from now, you will need to do in any scenario," Lukić said.</p> <p><em>John Moore is a freelance writer who has covered business and technology topics for 40 years. He focuses on enterprise IT strategy, AI adoption, data management and partner ecosystems.</em></p> </section> Anthropic's pending move intensifies the financial scrutiny on AI companies' business models, which will have downstream effects on enterprise AI adoption. https://cdn.ttgtmedia.com/rms/onlineimages/money_g1250581414.jpg https://www.techtarget.com/ai/feature/Anthropic-IPO-Market-forces-and-enterprise-AI-effects Wed, 14 Oct 2026 15:07:00 GMT <p><i>Within the next 12 months, 84% of IT and network operations (NetOps) leaders expect their organizations to have an AI-led operating model, according to </i><a href="https://newsroom.cisco.com/c/r/newsroom/en/us/a/y2026/m09/cisco-ai-research-agenticops-scaling-quickly-in-the-enterprise.html" target="_blank" rel="noopener"><i>a new survey from Cisco</i></a><i>. Already, enterprises are scaling AgenticOps and trusting agents with high levels of and even complete autonomy.&nbsp; </i></p> <p><i>The survey, <a href="https://www.techtarget.com/it-infrastructure/opinion/Splunk-conf26-Context-control-for-Ciscos-AI-agents-at-scale">conducted independently by Omdia,</a> a division of Informa TechTarget, gathered insight from more than 1,000 IT and NetOPs leaders working at organizations with at least 500 employees. The results offer insight into the operational complexity, creating the need for AgenticOPs, and the guardrails enterprises view as necessary to continue scaling agentic AI for NetOps.&nbsp; </i></p> <p><i>In this Q&amp;A, Joe Vaccaro, senior vice president/general manager, network platform and Assurance at Cisco, answers questions about the way <a href="https://www.techtarget.com/it-infrastructure/news/366645026/AgenticOps-for-infrastructure-faces-a-deterministic-dilemma">AgenticOPs is taking shape</a> in enterprises today and how it will drive the future of NetOPs. </i></p> <p><b>Editor's note</b>: The following interview was edited for length and clarity.</p> <h3><b>How is operational complexity putting pressure on traditional network operations in enterprises today?</b></h3> <p><b>Joe Vaccaro</b>: AI workloads are now becoming an increasing part of enterprise infrastructure. It's changing network traffic patterns and adding to operational complexity.&nbsp;</p> <p>When we think about a network, we often oversimplify it to a single connection between a client and an application. But the reality is that a network is a highly choreographed orchestra that connects a client to the application across many segments of the network or across multiple networks.</p> <p>Customers have seen a number of shifts related to that connectivity: the rise of cloud infrastructure replacing where workloads ran, the rise of internet-centric network connectivity replacing traditional corporate networks, and the rise of SaaS applications.</p> <div class="imagecaption alignLeft"> <img src="https://cdn.ttgtmedia.com/rms/onlineimages/joe_vaccaro.jpg" alt="headshot of Joe Vaccaro">Joe Vaccaro </div> <p>Now, AI is becoming a changing paradigm for how connections matter. As this complexity increases, <a href="https://www.techtarget.com/it-infrastructure/news/366643108/Why-AI-in-network-operations-requires-human-judgment">traditional AIOps and human-centered operating models</a> are reaching their practical limits in maintaining resilient network connectivity, which is paramount to delivering a great digital experience.</p> <h3><b>The survey found that 95% of enterprises say existing AIOPs tools can't keep up. Why are these tools not enough?</b></h3> <p><b>Vaccaro</b>: When degradation happens or performance challenges arise, very rarely do they stay within a single domain. We saw that 92% of the organizations surveyed identified common issues spanning networking, cloud security and applications.</p> <p>Customers today who are using different AIOps tools are often focused on a particular segment of the network or a particular domain. But connectivity issues can arise across many domains or segments of the network.</p> <p>At the same time, reliance on the network is increasing. People talk about accessing their AI tools. Well, AI depends upon the network.</p> <p>If your legacy AIOps tools are simply giving you recommendations but not actually able to drive action, then you're going to constantly be in a reactive position, trying to keep up with issues and trying to solve issues, versus: How do I spend time as an operations team trying to improve the network itself?</p> <h3><b>Enterprises are quickly moving to agentic-driven operations: AgenticOps. What does an AI-led operating model look like?</b></h3> <p><b>Vaccaro</b>: In many ways, the math of a human team no longer works. In the survey, we found that the average organization receives over 4,000 monitoring alerts per day, 51% of which are networking-related. If you have a practitioner who can clear 21 alerts a day, you would need a team of 100 people just to clear your backlog.</p> <p>The reality is that we know that they don't have that many people. A large majority of alerts go completely ignored or dismissed without finding the root cause. Now, some of that might be noise within the system, but there are issues that are going unaddressed that are creating frailty within the network itself.</p> <p>What we see is that by using agentic AIOPs and thinking of the agents as teammates. These teammates, when provided with cross-domain context, can help offload and augment humans. They can go through this agentic loop and help you continuously sense your environment across domains. Then, they can reason across cross-domain telemetry in context to identify problems in real time, not just suggest remediation steps, but put them into practice with a single click.</p> <p>In many ways, networking and network operations are <a href="https://www.techtarget.com/ai/feature/Is-your-business-ready-for-an-agentic-AI-team">the proving ground for agentic AI</a> in many organizations. Because at the end of the day, running a resilient network that delivers performance is critical to how businesses and organizations need to operate.</p> <h3><b>What kind of NetOPs tasks are AI agents taking on in enterprises today?</b></h3> <p><b>Vaccaro</b>: What we do see is that NetOps leaders are now more open to autonomy, but they're also very clear about the conditions required to scale it.</p> <p>The leading organizations we work with are identifying specific use cases and recurring issues across their networks that have a very strong signature in terms of the root cause and a very deterministic path for how you'd resolve them. Then, as agents can identify those issues, they can apply that level of determinism for how they go about solving it.</p> <p>The key thing is we talk about trust. How do you build trust? In the survey, we found that 69% of organizations require detailed explainability for an agent-driven action. That might even be underrepresented in terms of importance. The reality is that to offload and delegate, you need to ensure the agent finds the root cause you would expect. You do that by showing your work. You need to be able to show the reasoning for how you got to that conclusion.</p> <p>Over time, we will continue to build that level of trust and delegate more and more use cases to agents to scale the impact of network operation schemes.</p> <h3><b>The survey results indicate a high level of comfort with agent autonomy in NetOPs. Where are enterprise leaders comfortable reducing or eliminating human oversight?</b></h3> <p><b>Vaccaro</b>: To drive autonomy, the guardrails are not an optional step. The survey noted nearly 100% of organizations <a href="https://www.techtarget.com/cybersecurity/tip/How-to-build-AI-security-guardrails-without-blocking-innovation">require at least one guardrail</a> to trust before AI acts.</p> <p>What's an example of a guardrail? In general, agents relying upon an LLM give a probabilistic response. By nature, that's how LLMs operate. What we see is a probabilistic system to help you understand context and derive insights through sensing, reasoning and diagnosing, but you pair that with deterministic execution.</p> <p>When you come to this conclusion, instead of probabilistically making up how you would solve it, I'm going to dictate: here's how you solve it based on our organizational standard operating procedures or runbooks. Then it becomes a system that provides an additional level of trust. That's a key point, both in the survey findings and in what we see in practice as organizations adopt an AgenticOps mindset.</p> <h3><b>There are many reports indicating that AI adoption is outpacing governance and guardrails. What are the risks if that is the case for enterprises granting AI autonomy in NetOps?</b></h3> <p><b>Vaccaro</b>: It's important not to wipe autonomy across the entirety of network operations all at once. You need to identify specific use cases where you can understand how it's going to work in practice and then evolve with it.</p> <p><a href="https://www.techtarget.com/ai/news/366651175/AI-Safety-is-Falling-Behind-the-Pace-of-Development">Governance is important for understanding</a> how these agents will respond and for showing their work before they move to the next step, to ensure that there is potentially human oversight or approval before they move into action.</p> <p>But over time, as a level of trust and governance is established, we can begin to use the agents across a wider range of network operations tasks. That helps a team scale. I think we will continue to have humans at the center of network operations. That doesn't change. But now, as a team, they're able to focus their time on where they want to apply judgment rather than on keeping up with the flood of alerts that potentially hit them daily.</p> <h3><b>Scaling agentic AI requires trust, visibility and control. How can enterprises achieve those things as the adoption of AgenticOps accelerates?</b></h3> <p><b>Vaccaro</b>: It starts with that visibility. If you only have visibility into, say, the wireless segment of your network, an agent will always think that's the root of the problem. The reality, though, is that an issue might surface in how people connect to your wireless network, but the root cause might be upstream within your DHCP service. Or it might be upstream in your WAN connectivity. Or it might be the application itself. Do you have the right level of cross-domain telemetry and an understanding of the system as a whole to give agents the broadest set of context to reason across?</p> <p>The second is control -- the ability to have that level of deep intelligence.</p> <p>And then, it's about trust. When an action has surfaced, do you have the ability to see the evidence chain for how they drew that conclusion, to understand the confidence and risk score, and to know, within these guardrails, which issues need to be elevated to human oversight and which can be taken fully autonomously? If the confidence or risk score is elevated, you need to surface it for a human-led investigation.</p> <h3><b>What can enterprise leaders expect for the future of NetOPs? </b></h3> <p><b>Vaccaro</b>: This world is moving fast. At a practitioner level, their teams are already moving in this direction. In many ways, they're moving in this direction because they have no other choice. As a network operations team, their job is to operate the day-to-day heartbeat of their company across their corporate networks. And as business leaders, they need to empower their teams to leverage agentic operations to scale while maintaining governance and trust, ensuring a positive impact on their overall business and objectives.</p> <p>One of the things that came out in the survey is that we're seeing AI traffic across networks doubling every six months. How do we use AI to be able to augment network operations? But we also need to think about how we augment networks as agentic traffic rapidly increases across the environment.</p> <p>That's going to mean how do we think about agent security? How do we think about AI defense and model security? How do we think about identity across these agents? These will all be important in building confidence and trust in the systems. It's not just about how we operate our network, but also about <a href="https://www.techtarget.com/ai/news/366649952/AI-agent-security-How-reliable-is-enterprise-AI-testing">how we secure and trust the agents</a> operating alongside us.</p> <p><i>Carrie Pallardy is a freelance journalist with experience writing in cybersecurity, technology and healthcare. She currently covers a wide range of issues relevant to today's CIOs and IT leaders.</i></p> <p><i>Omdia is a division of Informa TechTarget. Its analysts have business relationships with technology vendors.&nbsp;</i></p> <p>&nbsp;</p> Cisco's Joe Vaccaro explains how AgenticOps is reshaping NetOps, as a new survey finds 84% of IT leaders expect an AI-led model within 12 months. https://cdn.ttgtmedia.com/rms/onlineimages/ai_a373894778.jpg https://www.techtarget.com/it-strategy/feature/Survey-finds-why-enterprises-are-adopting-AgenticOps Wed, 14 Oct 2026 14:53:00 GMT <p><span style="font-weight: 400">IFS has launched </span><a href="https://www.ifs.com/en/products/ai/ifs-intelligence/"><span style="font-weight: 400">IFS Intelligence</span></a><span style="font-weight: 400">, a new product that collates data from across your industrial operation (from assets, systems, processes) and turns it into meaningful action in minutes – creating value through new revenue streams and impactful operational efficiencies.</span></p> <p><span style="font-weight: 400">IFS Intelligence performs work directly in the systems that run physical operations.&nbsp;</span></p> <p><span style="font-weight: 400">Built around everything the operation already knows – its assets, work orders, schedules and history – it understands how that operation runs, working from live data across IFS and third-party systems, and acting only within the rules the business sets.</span></p> <p><span style="font-weight: 400">Industrial organisations have learned to operate with lost capacity: unplanned downtime has a target, missed first-time fixes are built into service plans, and scrap has an allowance. Many of those losses trace back to the technology they run on: systems that do not talk to each other, screens that take months to learn, implementations that take years, and the spreadsheets and workarounds people build to bridge the gaps.</span></p> <section class="section main-article-chapter" data-menu-title="Consumer-grade simplicity"> <h2 class="section-title"><i class="icon" data-icon="1"></i><span style="font-weight: 400">Consumer-grade simplicity&nbsp;</span></h2> <p><span style="font-weight: 400">The knowledge to cut those losses already exists, but getting at it means knowing which system to open, which screen to find, and who to ask. IFS Intelligence is designed to close the gap between what an operation knows and what it can get done. While deeply rooted in complex workflows and industry-specific context, consumer-grade simplicity is built into the user experience, with governance and guardrails embedded.</span></p> <blockquote> <p><span style="font-weight: 400">Nick Patience, VP &amp; practice lead for AI group, The Futurum Group, said: “Industrial organisations face a genuine dilemma: their frontline teams generate insights faster than enterprise systems can act on them. There is a growing need for solutions that address that gap and are built from industrial workflows. What matters to enterprise customers is that any advanced AI solution understands the operational constraints unique to these sectors, for the recommendations and actions it surfaces to be accurate, executable, and trusted. The market opportunity is real, but success requires that kind of industry-specific grounding.”</span></p> </blockquote> <p><span style="font-weight: 400">IFS Intelligence brings Industrial AI to the point of work: people in the field, on the plant floor, on a rig or in the office can ask questions in plain language through a single conversational interface. IFS Intelligence reasons across live operational data to explain what is happening and plan what should happen next, with the sources shown. It can then act on that plan, governed by IFS’s Industrial AI Harness.</span></p> </section> <section class="section main-article-chapter" data-menu-title="Business-critical scenarios"> <h2 class="section-title"><i class="icon" data-icon="1"></i><span style="font-weight: 400">Business-critical scenarios</span></h2> <p><span style="font-weight: 400">Customers are already using IFS Intelligence across a number of business-critical scenarios. One recent use case was experienced by a production manager who learned that a large order due next quarter was at risk. Finding out why, and what to do about it, meant connecting inventory, production capacity, supplier commitments, logistics constraints and the financial impact of delay.&nbsp;</span></p> <p><span style="font-weight: 400">With IFS Intelligence, one question can identify the risk, surface likely causes and recommend the next action, which the manager can approve and set in motion. What used to take several reports, systems and specialists, or might not have even been possible, becomes one governed flow from question to action.</span></p> <blockquote> <p><span style="font-weight: 400">Curtis Taylor, chief business operations officer, Sims Crane &amp; Equipment, said: “I was literally building skills and working out how to make our operational data actually accessible to our teams, piece by piece, when I saw IFS Intelligence. That changed the conversation entirely. Out of the box, all the data in our systems can be surfaced to our people. They can ask questions about our fleet, which cranes would be best to meet a particular customer requirement, for example, maximising our fleet efficiency.”</span><span style="font-weight: 400"><br> </span></p> </blockquote> <p><span style="font-weight: 400">IFS says that IFS Intelligence gives users three ways to put Industrial AI to work:</span></p> <ul class="default-list"> <li style="font-weight: 400"><b>Ask and act with Cowork:</b> <span style="font-weight: 400">question live operational data in plain language, investigate problems and prepare actions that run only once a person provides approval. Any answer can become a Lens, a live view that keeps updating as the operation moves, and work that succeeds once can be saved as a Skill for anyone to repeat. Cowork also reaches connected systems such as Microsoft 365 through MCP.</span><span style="font-weight: 400"><br> </span></li> <li style="font-weight: 400"><b>Delegate with Loops:</b><span style="font-weight: 400"> hand-defined workflows to digital workers that run in the background and escalate when judgment or approval is needed. Teams build their own digital workers in Studio, so capacity can grow without waiting on hiring.</span><span style="font-weight: 400"><br> </span></li> <li style="font-weight: 400"><b>Extend with Runway:</b><span style="font-weight: 400"> use plain-language requirements to speed up the migration, </span> <span style="font-weight: 400">configuration and extension of IFS, with less implementation effort and less risk.</span><span style="font-weight: 400"><br> </span></li> <li style="font-weight: 400"><b>Trust and Safety: </b><span style="font-weight: 400">Underneath all three, the IFS Industrial AI Harness applies the context and controls that industrial operations require.</span><span style="font-weight: 400"><br> </span></li> </ul> <blockquote> <p><b>“</b><span style="font-weight: 400">We often see industrial organisations operating at around 60% of their potential. The missing 40% is already inside the business. It sits in the expertise of the control room, the knowledge of the night shift, and the data spread across every system. IFS Intelligence brings all of it together and puts it to work. People can ask a question, take action, or hand a task to a digital worker, all within the rules, permissions and approvals that govern their business. That is how industry closes the gap,” said Mark Moffat, CEO, IFS.</span></p> </blockquote> <p><span style="font-weight: 400">IFS Intelligence is available now. It is offered as a universal enterprise-wide license. The package unlocks IFS Intelligence capabilities across the full spectrum of in-scope software that the customer has licensed from IFS. Customers receive full access to the IFS Intelligence capabilities with no feature-gating or tiering. In addition, customers will also receive an entitlement of AI Executions. IFS says that this is not a credit model but a fair usage cap.</span></p> <p><span style="font-weight: 400"><img loading="lazy" decoding="async" class="wp-align alignnone wp-image-12784" src="https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-192254.png" alt="" width="580" height="287" srcset="https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-192254.png 1143w, https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-192254-300x148.png 300w, https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-192254-1024x506.png 1024w, https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-192254-768x380.png 768w" sizes="(max-width: 580px) 100vw, 580px"> <img loading="lazy" decoding="async" class="wp-align alignnone wp-image-12785" src="https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-192326.png" alt="" width="580" height="290" srcset="https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-192326.png 1162w, https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-192326-300x150.png 300w, https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-192326-1024x511.png 1024w, https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-192326-768x383.png 768w" sizes="(max-width: 580px) 100vw, 580px"></span></p> <p>&nbsp;</p> <p>&nbsp;</p> </section> IFS has launched IFS Intelligence, a new product that collates data from across your industrial operation (from assets, systems, processes) and turns it into meaningful action in minutes – creating ... https://www.computerweekly.com/blog/CW-Developer-Network/IFS-Unleashed-Day-2-IFS-Intelligence-a-route-to-solve-applied-Industrial-AI-challenges Wed, 14 Oct 2026 14:35:54 GMT <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Key takeaways</h3> <ul class="default-list"> <li><b>Lower token prices don't guarantee lower bills. </b>More requests, longer context, reasoning and repeated attempts by AI agents can eat up the savings.</li> <li><b>One request can trigger an entire workflow.</b> Enterprise AI agents retrieve information, reconcile records, call tools, check results and recover from failures. Each LLM interaction adds costs.</li> <li><b>Measure successful execution to control costs.</b> Map the workflow, test its economics and enforce usage limits. The useful metric is cost per correctly completed business task.</li> </ul> </div> </div> <p>For enterprise technology leaders, falling AI prices sound like good news. In practice, the total bill often tells a different story.</p> <p>Recently released models, such as GPT-5.6 and <a href="https://www.techtarget.com/ai/news/366650804/Anthropic-unveils-Opus-55-powerful-performance-still-premium-price">Claude Opus 5.5</a>, can write and debug code, analyze documents and use tools to complete multi-step tasks, all at substantially lower token prices. A year ago, Claude Opus 4.1 cost $15 per million input tokens and $75 per million output tokens; Opus 5.5 costs $4 and $20 per million, respectively -- a reduction of approximately 73%. OpenAI cut the price of GPT-5.6 Luna, its fastest and most affordable model, by 80% on July 30, and the price of GPT-5.6 Terra by 20%, leaving Luna at $0.20 per million input tokens and $1.20 per million output tokens.</p> <p>Yet, lower prices have not <a href="https://www.techtarget.com/it-strategy/feature/How-CIOs-can-avoid-the-AI-surprise-bill">eliminated budget surprises</a>. Uber reportedly exhausted its full-year AI budget by April as engineering teams accelerated adoption of AI coding tools. The price of a computation unit is falling, but the amount of computation behind a business request can grow considerably. The apparent contradiction becomes easier to understand when we look at what enterprises are asking AI to do now.</p> <section class="section main-article-chapter" data-menu-title="Cheaper tokens, more work behind every request"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Cheaper tokens, more work behind every request</h2> <p>A simple chatbot interaction might involve sending a question and receiving an answer. An enterprise agent handling a business process may need to make repeated model calls, retrieve records, inspect results, revise its approach and verify completion.</p> <p>That many steps change the economics. Consider an illustrative example: If the <a href="https://www.techtarget.com/it-strategy/feature/Tokenmaxxing-How-CIOs-can-extract-maximum-value-from-AI-tokens">effective token price</a> falls by 50%, but a workflow consumes four times as many tokens, its token cost doubles. If more employees run that workflow more frequently, spending multiplies across the entire head count.</p> </section> <section class="section main-article-chapter" data-menu-title="The cost multiplier: From answering to executing"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The cost multiplier: From answering to executing</h2> <p>Consider a familiar business request: "Which completed jobs haven't been invoiced?" The sentence is short. Executing it may require the agent to do the following:</p> <ul class="default-list"> <li>Retrieve completed jobs from a field-service application.</li> <li>Identify matching customers across operational and billing records.</li> <li>Check whether invoices exist and whether they cover the completed work.</li> <li>Investigate inconsistent names, missing identifiers and partial invoices.</li> <li>Recover from failed calls or incomplete responses.</li> <li>Produce a verified list of exceptions.</li> </ul> <p>If the request includes creating the missing invoices, then the workflow also needs approval, controlled writeback and confirmation that the transactions succeeded.</p> <p>Each additional model call <a href="https://www.techtarget.com/ai/tip/How-to-apply-FinOps-to-optimize-agentic-AI-costs">processes billable input and output tokens</a>. Retries add calls, and records returned by tools can become input to later calls. Cached input is cheaper but still billed. Reasoning models also generate "thinking" tokens before they answer, and those are billed as output tokens -- the more expensive side of the price list -- so a harder question can cost far more than its short answer suggests.</p> <p>Clio, a legal software company, <a target="_blank" href="https://www.clio.com/enterprise/blog/the-release-report-april-2026/" rel="noopener">reported</a> 38% fewer prompt tokens with no quality loss in multi-step document analysis tests when code handled intermediate tool operations. Fewer tokens came from changing how the workflow ran.</p> </section> <section class="section main-article-chapter" data-menu-title="Managing the cost of execution"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Managing the cost of execution</h2> <p>Routing easy steps to smaller models, caching repeated context and using batch discounts all help manage costs. They can reduce token consumption or token charges, but they don't show how much a finished task costs, so CIOs must still measure the entire workflow.</p> <p>Follow these steps to help manage costs:</p> <h3>1. Map the workflow before you price it</h3> <p>For example, accounts payable teams follow an invoice from receipt through purchase-order matching, delivery verification, any exceptions review and payment approval. Count the model calls and include what happens when records are missing. Use code to check totals and match purchase-order numbers, and use the model to interpret and explain discrepancies. Budget for the <a href="https://www.techtarget.com/it-strategy/news/366651902/Exception-handling-in-AI-The-silent-killer-of-enterprise-deployments">exception path</a> and the clean record match.</p> <h3>2. Test the cost of a correct result</h3> <p>For customer returns, test requests involving split shipments, expired return windows, missing receipts and unavailable order records. Compare models on whether they apply the return policy correctly, what each completed case costs and how much employee correction it requires. Run cases repeatedly to expose expensive retries and repeat the tests when the model or return policy changes in the future.</p> <h3>3. Set limits that stop unproductive runs</h3> <p>For an enterprise coding agent, track spending per accepted change, including failed attempts. If the agent repeatedly runs the same failing test or cannot access a dependency, stop it and notify an engineer of the blocker. Set spending, retry and runtime limits before rollout. Measure changes that pass review and testing. Token consumption alone tells you how much activity you paid for.</p> </section> <section class="section main-article-chapter" data-menu-title="The bottom line"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The bottom line</h2> <p>CIOs must reset AI budgets around completed workflows now. The 2026 <a target="_blank" href="https://www.researchgate.net/publication/403154371_The_Price_Reversal_Phenomenon_When_Cheaper_Reasoning_Models_End_Up_Costing_More" rel="noopener">research preprint</a> <i>The Price Reversal Phenomenon: When Cheaper Reasoning Models Cost More</i> tested eight models across 12 tasks. In 32% of model-pair comparisons, the lower-priced model incurred higher total inference costs. Listed token prices are not a reliable basis for a budget forecast.</p> <p>Before expanding deployment, CIOs should measure cost per correctly completed task, including failed attempts, infrastructure and human review. Next, CIOs need to set spending and retry limits, and track completion rates and correction time against the existing process. Finally, CIOs can only scale when the results justify the expense.</p> </section> Falling token prices are lowering AI prices. But as enterprises shift from chatbot questions to multi-step, agent-enabled workflows, costs can skyrocket unless managed properly. https://cdn.ttgtmedia.com/rms/onlineimages/money_g1021600178.jpg https://www.techtarget.com/it-strategy/opinion/AI-tokens-got-cheaper-but-the-deployment-bill-didnt Wed, 14 Oct 2026 14:23:00 GMT <p><span style="font-weight: 400">A new annual index from IFS suggests a growing execution divide as industrial frontline teams tap operational data to solve critical challenges and achieve cost savings with AI while leadership catches up on governance, AI safety, and scale</span></p> <p><span style="font-weight: 400">IFS has launched new research revealing a significant divide between frontline execution and C-suite strategy in industrial AI usage. According to the </span><i><span style="font-weight: 400">IFS Industrial Index 2026</span></i><span style="font-weight: 400">, which surveyed over 2,200 senior decision-makers from across 14 countries, operational workers are driving real-world AI implementation and delivering tangible, commercial value.&nbsp;</span></p> <p><span style="font-weight: 400">Meanwhile, formal AI governance frameworks and data access structures are not keeping pace.</span></p> <section class="section main-article-chapter" data-menu-title="The operational impact"> <h2 class="section-title"><i class="icon" data-icon="1"></i><span style="font-weight: 400">The operational impact</span></h2> <p><span style="font-weight: 400">When frontline teams – engineers, plant managers, service leads, logistics managers – connect their domain expertise and operational data with Industrial AI, the business impact is immediate. Rapid adoption on the floor is delivering direct commercial returns in savings across key operational areas, despite solutions not currently being used to the fullest.&nbsp;</span></p> <p><span style="font-weight: 400">Over 90% of global organisations report significant gains in reliability and performance:</span></p> <ul class="default-list"> <li style="font-weight: 400"><span style="font-weight: 400">93% </span> <span style="font-weight: 400">improvement in first-time-right performance</span></li> <li style="font-weight: 400"><span style="font-weight: 400">90% </span> <span style="font-weight: 400">increase in asset availability&nbsp;</span></li> <li style="font-weight: 400"><span style="font-weight: 400">88% </span> <span style="font-weight: 400">improvement in on-time delivery schedule adherence</span></li> </ul> <p><span style="font-weight: 400">Removing operational friction also frees employee capacity. With over 50% of employee time saved across key workflow areas like production throughput and quality assurance, companies are repurposing workforce capacity toward higher-value initiatives:<br> </span></p> <ul class="default-list"> <li style="font-weight: 400"><span style="font-weight: 400">36% prioritising workforce safety, training and operational compliance </span></li> <li style="font-weight: 400"><span style="font-weight: 400">35% optimising asset performance and reliability </span></li> <li style="font-weight: 400"><span style="font-weight: 400">26% accelerating continuous process improvement </span></li> <li style="font-weight: 400"><span style="font-weight: 400">25% growing overall market share</span><span style="font-weight: 400"><br> </span></li> </ul> </section> <section class="section main-article-chapter" data-menu-title="The knowledge impact"> <h2 class="section-title"><i class="icon" data-icon="1"></i><span style="font-weight: 400">The knowledge impact</span></h2> <p><span style="font-weight: 400">Factory floor teams are pushing ahead as they know the operational problems that need to be solved, recognising that the answers already lie in their institutional knowledge and operational data.</span></p> <p><span style="font-weight: 400">The data reveals that frontline workers are taking workforce transformation into their own hands:</span></p> <ul class="default-list"> <li style="font-weight: 400"><span style="font-weight: 400">43% have trained and introduced agentic digital workers into their day-to-day operations over the past 12 months.</span></li> <li style="font-weight: 400"><span style="font-weight: 400">37% have upskilled peers to use industrial AI tools in everyday operations.</span></li> <li style="font-weight: 400"><span style="font-weight: 400">36% have directly trained staff to collaborate with digital workers</span></li> </ul> <p><span style="font-weight: 400">In contrast, executive leadership’s primary AI effort over the same period was upskilling management to improve their own AI literacy (41%), directing attention inward rather than enabling the floor. While 27% of respondents view AI literacy as the first vital step for organisational readiness, ground-level execution is moving regardless, leaving leadership confidence (30%) as one of the biggest operational barriers to scaling success.</span></p> <blockquote> <p><span style="font-weight: 400">“The factory floor is moving at full speed. Workers see the operational insights, know their data better than anyone, and they know they need industrial-grade AI solutions to turn that expertise into action. The real challenge for leadership isn’t driving adoption; it’s having the insights to clear legacy roadblocks and delivering the secure industrial harness, data pipelines, and governance required to scale these wins safely and globally,” said Cathie Hall, chief product &amp; customer officer at IFS.</span></p> </blockquote> <p><span style="font-weight: 400">While floor-level initiative proves demand for intelligent tools, the disconnect between perceived readiness and actual operational safeguards creates potentially dangerous exposure.&nbsp;</span></p> </section> <section class="section main-article-chapter" data-menu-title="Critical controls are missing"> <h2 class="section-title"><i class="icon" data-icon="1"></i><span style="font-weight: 400">Critical controls are missing</span></h2> <p><span style="font-weight: 400">Despite over 70% of organisations claiming confidence in their safety and data policies, critical controls are missing where they matter most and only 11% have validated AI outputs for safety-, quality-, or mission-critical processes.</span></p> <ul class="default-list"> <li style="font-weight: 400"><span style="font-weight: 400">Only 11% have established control mechanisms for AI use with sensitive or export-controlled data.&nbsp;</span></li> <li style="font-weight: 400"><span style="font-weight: 400">Only 4% have implemented mandatory human oversight or approval requirements for AI-supported decisions.</span></li> <li style="font-weight: 400"><span style="font-weight: 400">Only 20% have embedded Industrial AI into supply chain management or </span> <span style="font-weight: 400">maintenance planning (MRO) – the functions where operational failures carry the highest consequences</span></li> </ul> <p><span style="font-weight: 400">In industrial operations, this gap between confidence and control is an operational liability. Agentic AI at scale without robust safety frameworks and decision oversight creates compounding risk: unvalidated outputs in mission-critical processes, uncontrolled data exposure, decisions without human accountability.</span></p> <p><span style="font-weight: 400">Hall concluded: “Industrial operations are defined by speed and consequences. Every decision, every asset, every second compounds. To compete, leadership must match the frontline’s pace. That means having Industrial AI at the heart of operations to support rapidly scaling agentic AI, enabling the safety, governance, and data controls that make rapid decision-making trustworthy and operationally sound. Companies that move now will compress decision cycles and extend competitive advantages. Leadership cannot afford to lag.”</span></p> <p><img loading="lazy" decoding="async" class="wp-align alignnone wp-image-12779" src="https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-191248.png" alt="" width="580" height="324" srcset="https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-191248.png 1171w, https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-191248-300x168.png 300w, https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-191248-1024x573.png 1024w, https://itknowledgeexchange.techtarget.com/cwdn/files/2026/10/Screenshot-2026-10-14-191248-768x430.png 768w" sizes="(max-width: 580px) 100vw, 580px"></p> </section> A new annual index from IFS suggests a growing execution divide as industrial frontline teams tap operational data to solve critical challenges and achieve cost savings with AI while leadership ... https://www.computerweekly.com/blog/CW-Developer-Network/IFS-Unleashed-Day-2-IFS-Industrial-Index-2026-shows-factory-floor-surging-in-AI Wed, 14 Oct 2026 14:17:28 GMT <p>US authorities have fined Bank of America (BoA) $1.2bn for failing to prevent and report national security risks including Chinese money laundering.</p> <p>The US Department of the Treasury’s <a href="https://home.treasury.gov/news/press-releases/sb0657">Financial Crimes Enforcement Network</a> (FinCEN) said the fine related to the bank’s failure to identify and report suspicious transactions.</p> <p>The punishment includes a $750m fine from the Office of the Comptroller of the Currency (OCC), related to deficient and ineffective anti-money laundering compliance.</p> <p>FinCEN said: “… Bank of America admits that it wilfully violated the Bank Secrecy Act (BSA) and failed to implement and maintain an <a href="https://www.techtarget.com/ai/feature/Using-AI-in-AML-fights-fraud-while-protecting-privacy?_gl=1*a7z70m*_ga*Mjg4MTcxMzU3LjE3Nzk5NjUxNTU.*_ga_TQKE4GS5P9*czE3OTE5ODQ5MDQkbzExMCRnMSR0MTc5MTk4NTAzMSRqNTIkbDAkaDA.">anti-money laundering</a> (AML) programme that met the minimum requirements.”</p> <p>The US Department of the Treasury said: “The central failures identified through FinCEN’s investigation involved deficiencies related to Bank of America’s suspicious transaction identification and reporting and customer due diligence processes. As these issues arose, Bank of America prioritised its revenues and minimised the impact on its required compliance resources.”</p> <p>Scott Bessent, secretary <span>at the US Department of the Treasury, said: “Financial institutions are on the front lines of protecting our economic and national security. When a bank of this size fails to meet its anti-money laundering and sanctions obligations, the consequences extend far beyond its own walls, exposing the entire US financial system to significant illicit finance risks.”</span></p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about anti-money laundering technology</h3> <ul class="default-list"> <li><a href="https://www.computerweekly.com/news/365530101/Estonian-anti-money-laundering-platform-used-to-fight-APP-fraud">A real-time data-sharing platform originally built to identify money laundering is also being applied to the fight against authorised push payment fraud</a>.</li> <li><a href="https://www.computerweekly.com/news/252513230/UK-second-in-money-laundering-hall-of-shame">Banks need to step up their anti-money laundering processes if billions of pounds’ worth of criminal activity is to be prevented</a>.</li> <li><a href="https://www.computerweekly.com/news/252507903/NatWest-admits-to-weaknesses-in-anti-money-laundering-systems">Bank pleads guilty to failures concerning the laundering of hundreds of millions of pounds, but says it has since improved its anti-money laundering systems</a>.</li> </ul> </div> </div> <p>The OCC said Bank of America is the second-largest bank in the US and the sixth-largest in the world, processing hundreds of trillions of dollars in transactions across the globe.</p> <p>“It has a high, and increasing, risk profile for money laundering, terrorist financing and other illicit activities,” said comptroller of the currency Jonathan Gould. “Almost two years ago, the bank signed an order with the OCC and committed to fixing its longstanding BSA deficiencies.</p> <p>“Despite its obligation to the American people as a national-chartered, global systemically important bank, the bank’s BSA/AML compliance programme continues to reflect widespread breakdowns in internal controls, which threatens US national and economic security.</p> <p>“The OCC expects all banks to have an effective BSA/AML compliance programme,” he said. “Those that do not can expect swift and significant action from the OCC.”</p> <section class="section main-article-chapter" data-menu-title="Over in Europe"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Over in Europe</h2> <p>There could be <a href="https://www.computerweekly.com/news/366642629/EU-finance-firms-urged-to-get-on-with-anti-money-laundering-compliance">trouble ahead for banks in Europe</a> in relation to AML regulations.</p> <p>In May this year, a survey carried out by PwC Luxembourg found that incomplete guidance and data quality challenges are making it difficult for financial services firms to meet the rules of the EU’s AML Package regulations.</p> <p>PwC found that only a third of the 500 firms questioned across 40 EMEA countries will be ready for the July 2027 <a href="https://www.techtarget.com/searchenterpriseai/feature/Using-AI-in-AML-fights-fraud-while-protecting-privacy">AML regulation</a> deadline.</p> <p>PwC found that companies are struggling because full guidelines on the regulations, known as regulatory technical standards, are not yet available and, from a technological point of view, they are finding it difficult to provide the data demanded by regulators.</p> <p>Data quality continues to be the most important barrier to advanced technology and artificial intelligence (AI) adoption, according to 89% of electronic and virtual payments firms, 64% of banks and insurance companies, and 52% of asset and wealth management (AWM) companies in the EMEA region.</p> <p>The PwC figures revealed that 30% of banks believe they will need a significant change in their overall data structure to meet regulations, with 46% of AWM firms facing the same challenges, along with 39% of insurers and 43% of e-payment service providers.</p> </section> Bank of America has been fined £1.2bn for its failings related to anti-money laundering https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/Dollars-Fotolia_78255734.jpg https://www.computerweekly.com/news/366651843/Bank-of-America-hit-with-12bn-fine-after-money-laundering-probe Wed, 14 Oct 2026 11:30:00 GMT <p>As generative AI tools flood vulnerability intake pipelines with plausible-looking "slop," tech giants are reconsidering the ROI of traditional, paid bug bounty programs. Between external triage exhaustion and the rise of internal automated AI auditing, industry heavyweights like Intel and Google are forcing a rewrite of vulnerability disclosure rules.</p> <p>What began as isolated adjustments could herald an industry-wide pivot. Earlier this year, Curl creator Daniel Stenberg shut down the project's bug bounty program after being overwhelmed by a flood of LLM-generated false positives.</p> <p>By mid-September, reports surfaced that Intel had quietly suspended cash rewards for its long-running bug bounty program. Weeks later, on October 1, Google <a href="https://x.com/GoogleVRP/status/2105689195180179605" target="_blank" rel="noopener">stopped accepting vulnerability reports</a> for its own Open-Source Software Vulnerability Reward Program (OSS VRP), explicitly citing an unsustainable influx of invalid, automated reports.</p> <p>Together, these moves mark a turning point in vulnerability management. Security teams are facing an asymmetric resource drain in which low-effort AI reports cost submitters almost nothing to generate but require hundreds of expensive engineering hours to manually review and debunk.</p> <section class="section main-article-chapter" data-menu-title="Triage under pressure"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Triage under pressure</h2> <p>Intel's sudden pause on paid bug bounties is one of the most prominent casualties of this shifting operational landscape.</p> <p>In a statement to <i>TechTarget</i>, the tech giant said: "As part of an ongoing review of how the <a href="https://www.computerweekly.com/opinion/Why-bug-bounty-programmes-are-essential-for-cyber-resilience">bug bounty</a> program can best support Intel's broader product security efforts, Intel has paused the program until further notice. Intel will continue to honor existing bug bounty commitments and accept security reports through its Vulnerability Disclosure Program."</p> <p>Intel did not respond to a follow-up question about whether the decision had been prompted by an increase in the use of AI tools to discover vulnerabilities. However, Kara Sprague, CEO of HackerOne -- the platform hosting Intel's rewards program -- confirmed that submission volumes across enterprise programs are surging.</p> <p>"More of them look credible at first glance but were never tested against the target or are outside the program's scope: a report can now be written in minutes," she said.</p> <p>"Validating it still takes what it always did: reproducing the finding, confirming the impact and deciding whether it needs a fix. That is the constraint on every program right now, and it is the work we have built our platform around."</p> <p>Industry data underscores the sheer scale of this vulnerability disclosure bottleneck. According to a recent <a href="https://cloud.google.com/blog/topics/threat-intelligence/vulnerability-discovery-and-exploitation-trends-in-the-ai-era" target="_blank" rel="noopener">report from Google</a>, the number of vulnerabilities disclosed per month doubled between January and July 2026, rising from 5,045 in January to 10,477 in July, before climbing further to 10,740 in August.</p> </section> <section class="section main-article-chapter" data-menu-title="The shift to internal auditing"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The shift to internal auditing</h2> <p>As the maintainers of Curl discovered, a substantial increase in submissions takes critical time away from the analysts and researchers who must review each report. As a result, the emphasis is shifting towards ensuring that submissions meet specific criteria and reducing the time spent assessing reports that do not identify genuine vulnerabilities.</p> <p>Reflecting on the collapse of Curl's bug bounty program in a post on his <a href="https://daniel.haxx.se/blog/2026/01/26/the-end-of-the-curl-bug-bounty/" target="_blank" rel="noopener">blog</a>, creator Daniel Stenberg tracked the clear correlation between the widespread adoption of generative AI tools like ChatGPT and the plummeting signal-to-noise ratio in disclosure queues.</p> <p>"Looking back, I think we can say that the downfall of the bug bounty program started slowly in the second half of 2024 but accelerated badly in 2025," Stenberg wrote.</p> <p>In Intel's case, the company confirmed that its program had been paused rather than scrapped altogether. Katie Moussouris, CEO of Luta Security, said she suspected that two complementary factors led Intel to stop paying for bugs: a flood of low-quality AI-generated reports and Intel's own teams deploying AI to find vulnerabilities more efficiently in-house.</p> <p>These high-profile pauses raise fundamental questions about the long-term viability of crowdsourced security. Rather than an outright extinction of bug bounties, however, experts view the current disruption as a permanent realignment of enterprise security budgets.</p> <p>Moussouris told <i>TechTarget</i> she did not believe bug bounties would return to their former pricing levels. While vulnerability disclosure remains welcome, she noted that the wider picture depends on how quickly AI can automate internal testing and patch creation.</p> <p>"The investment in software that is secure by design and using AI internally to find and fix vulnerabilities will take up more budget and time than what has been previously allocated to external testing of all kinds," she said. "AI tooling, agentic red teaming and bug bounties become too burdensome for most organizations to find value among the duplicates and false reports."</p> </section> <section class="section main-article-chapter" data-menu-title="An industry at a crossroads"> <h2 class="section-title"><i class="icon" data-icon="1"></i>An industry at a crossroads</h2> <p>Sprague said she expects more programs to change their rules, and some may pause rewards. "The problem many program owners face is familiar: more reports are coming in, and someone has to work out which ones warrant attention from engineering," she said.</p> <p>"There's another side to this, though: we're also seeing more valid findings than ever. Pausing a program may ease the volume, but it means losing the work of researchers who are finding real issues."</p> <p>The HackerOne CEO stressed that pausing rewards is a difficult choice because a program may still receive the same raw volume of unvetted reports while losing access to the skilled researchers capable of uncovering complex flaws.</p> <p>"If the problem is sorting through the volume, we need to get better at that work," Sprague said. "Those researchers are still worth listening to."</p> <p>The impact of Intel's pause will be felt across the industry: from researchers seeking payment for their work to the organizations that receive reports, validate vulnerabilities and develop patches.</p> <p>With major technology companies now scaling back aspects of their bug bounty programs, the question is whether others will follow suit -- and whether the sector can adapt to a world in which AI makes finding potential vulnerabilities easier, but validating them remains a significant challenge.</p> <p><em>Dan Raywood is a journalist with more than 20 years of professional writing experience, including more than 12 years covering cybersecurity.</em></p> </section> As AI fills bug bounty queues with low-quality report slop, tech giants like Intel and Google are rethinking paid rewards and shifting toward automated internal defenses. https://cdn.ttgtmedia.com/rms/onlineimages/security_g1192070289.jpg https://www.techtarget.com/cybersecurity/news/366652109/Bug-bounty-reckoning-AI-slop-forces-enterprise-strategy-shift Wed, 14 Oct 2026 11:14:00 GMT <p>After a wait of almost a month, the UEFA Champions League returned on 13 October. In addition to seeing strong performances by Barcelona, Arsenal, FC Bruges and Athletico Madrid on the pitch, there was a commensurate robust performance on global networks hosting streaming video of the games, with traffic hitting new highs according to data from DE-CIX.</p> <p>Major live events significantly increase demand for simultaneous access to digital content and, with it, the load on the underlying network infrastructure. The new record values coincided with numerous UEFA Champions League matches and showed how increasing numbers of football fans followed the games via streaming platforms.</p> <p>The <a href="https://www.computerweekly.com/news/366649631/DE-CIX-expands-global-AI-ready-connectivity-with-Microsoft-Azure-route">leading internet exchange provider</a> said that its exchange in Frankfurt hit 21.47 Tbps for the first time, just 25 days after it first crossed 20 Tbps, while global traffic across DE-CIX’s exchanges passed 32 Tbps. Both peaks came 20 minutes after kick-off in Tuesday’s UEFA Champions League matches, as fans streamed the games live. The data traffic in Frankfurt was equal to around 12 million TikTok videos streamed at once.</p> <p>Both records were measured at 21:20 Central European Summer Time, just 20 minutes after kick-off in numerous UEFA Champions League matches. In all, the first night of the Champions League Matchday saw seven games across Europe – from London, Madird, Istanbul, Milan, Leipzig, Stavanger and Villareal – kick off at 21:00 CET.</p> <p>Internet traffic has been growing steadily for years, however DE-CIX stressed that the pace of this development has now accelerated. These new traffic levels are said to have come just over three weeks after the last records were set, with peak traffic values having increased by around 7%.</p> <p>On 17-18 September 2026, DE-CIX Frankfurt traffic exceeded the 20-terabit mark for the first time, reaching 20.02 Tbps, while global traffic surpassed the 30 Tbps threshold. The spike was said to have been generated by fans watching UEFA Europa League and Europa Conference League competition matches at this time.</p> <p>DE-CIX said these new records underline the growing importance of high-performance digital infrastructure in an increasingly connected world. “Just three-and-a-half weeks ago, we crossed the 20 Tbps mark in Frankfurt for the first time. Now we are already at almost 21.5 terabits, and worldwide at more than 32 terabits. This momentum is remarkable and demonstrates the scale that digital traffic has now reached,” said CEO Ivo Ivanov.</p> <p>“Live sports make this development particularly visible when millions of people access digital content simultaneously. Behind this lies a long-term trend: streaming, cloud services, gaming and increasingly artificial intelligence are changing the requirements placed on our networks. High-performance, direct and resilient connections are becoming ever more important for the digital economy.”</p> <p><a href="https://www.de-cix.net/en/locations/frankfurt">DE-CIX Frankfurt</a> is one of the largest Internet Exchanges in the world and serves as a central data hub for Europe. Since the beginning of 2020, peak traffic at DE-CIX Frankfurt has increased by more than 160%. In 2025, nearly 48 exabytes of data were exchanged at DE-CIX Frankfurt, where almost 1,100 networks are connected.</p> <p>Overall, DE-CIX operates more than 60 internet, cloud and AI exchanges across North and South America, EMEA and Asia. Its global interconnection ecosystem connects more than 4,600 networks from over 100 countries.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about streaming video networks</h3> <ul class="default-list"> <li><a href="https://www.computerweekly.com/news/366649189/ITV-tames-World-Cup-streaming-spikes-with-Akamai-TrafficPeak">ITV tames World Cup streaming spikes with Akamai TrafficPeak</a>: UK’s leading commercial broadcaster deploys network management and real-time data platform for operational intelligence deployed to safeguard live-streaming architecture during the World Cup.</li> <li><a href="https://www.computerweekly.com/news/366573493/BT-Broadpeak-intro-multicast-tech-to-enhance-network-video-streaming">BT, Broadpeak intro multicast tech to enhance network video streaming</a>: UK’s leading network provider teams with content delivery network firm to enhance live video streaming quality and reliability using adaptive bit rate and multicast-assisted unicast delivery technology.</li> <li><a href="https://www.computerweekly.com/news/366630816/Orange-Synamedia-join-forces-to-expand-multi-CDN-reach">Orange, Synamedia join forces to expand multi-CDN reach</a>: Video software provider teams with global telco to improve network performance for streaming providers to deliver content more smoothly and reach more viewers, especially in the Middle East and Africa.</li> <li><a href="https://www.computerweekly.com/news/366650343/AI-improves-reliability-of-5G-multicast-broadcasting">AI improves reliability of 5G multicast broadcasting</a>: Japanese university study presents ‘lightweight’ approach for improving the reliability of mobile broadcasting to help prevent temporary video freezing caused by lost data during wireless transmission.</li> </ul> </div> </div> Return of top-level European football sees surge in online viewership traffic at Frankfurt data exchange and other locations reach new highs https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/sport-football-2-adobe.jpeg https://www.computerweekly.com/news/366652212/Internet-traffic-records-topple-as-Champions-League-returns Wed, 14 Oct 2026 11:06:00 GMT <p>When <a href="https://www.techtarget.com/it-infrastructure/news/366650261/AI-CEOs-call-for-slowing-AI-boom-as-billions-flow-to-data-centers">Dario Amodei</a> called for the brakes to be pumped on AI model development, it provoked reaction from not only his fellow frontier AI CEOs but no less than the Pope, the President and the King of the UK.&nbsp; But whether to apply the brakes is just one component of the "race" that's underway to <a href="https://www.techtarget.com/ai/feature/Controlling-AI-models-is-harder-than-building-them">develop and deploy AI safely</a> and at scale.</p> <p>"Clear rules of the road are important. But without people to put those rules into practice, rules are reduced to just words." I wrote that three years ago, when I argued that "the time to professionalize AI governance is now."</p> <p>Now, in 2026, it's even more plain to see just how important a scaled and professional workforce dedicated to AI governance is going to be if AI is to serve --and not damage -- society in positive and productive ways.</p> <p>History has shown that early investment in a professionalized workforce pays dividends later. Today, that investment is more urgent than it is early. When global capital markets expanded in the 20th century, stability and, ultimately, success, were not achieved simply by lawmakers passing statutes against corporate fraud. It required the invention of the modern accounting profession -- a trained, standardized workforce embedded within every enterprise to audit ledgers, verify claims, and operationalize financial controls. We are currently facing a parallel structural necessity with <a href="https://www.techtarget.com/it-strategy/feature/SMBs-face-AI-governance-reckoning">AI governance</a></p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Despite the heightened prioritization of AI governance, organizations are lagging in their response. </figure> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>This is an imperative for Main Street, not just a Silicon Valley problem. While existential risks loom large in the minds of frontier AI tech CEOs, hundreds of thousands of organizations across the economic spectrum are experiencing real-world rewards and risks today. On the latter, risks, such as privacy, hallucinations, cybersecurity, bias and <a href="https://www.techtarget.com/cybersecurity/tip/Shadow-AI-agents-What-CISOs-need-to-know">shadow AI</a>are not found only on the frontier, nor do they require hypothetical predictions as to their likelihood of occurring.<i>These risks are happening today inside many organizations.</i> In fact, more than two-thirds of organizations are already investing in people and processes to tackle these risks. But more work is needed still.</p> <p>Whether and how to regulate AI has seized lawmakers across the world. It animates all manner of policy debates, from national security and global competitiveness to human rights and even what it means to be sovereign. Attempts to legislate are as complex as they are varied. The EU has shown some signs of buyer's remorse with its much-heralded <a href="https://www.techtarget.com/ai/news/366646620/EU-AI-Act-compliance-deadline-is-here-What-to-watch">AI Act</a>, China has advanced exacting rules on generative AI, and lawmakers in the U.S. have proposed hundreds of bills at the state level, resulting in dozens of laws on the books targeting certain AI risks.</p> <p>Regardless of the rules policymakers ultimately decide are necessary, in addition to the many existing laws that already apply to AI, or whether industry can somehow coordinate to voluntarily pump the brakes on frontier model development, the rapid and responsible deployment of AI systems will rely on the professionals behind the wheel and the structures and ecosystem to empower their flourishing.</p> <p>Despite the heightened prioritization of AI governance, organizations are lagging in their response. An average organization anticipates needing 10 people to appropriately cover AI governance responsibilities. Yet, most of those same organizations doubt they have the budget or structures to attract and empower such a workforce. Hundreds of thousands of AI governance professionals will be demanded in short order.This is not a latent workforce, waiting to mobilize.&nbsp; We cannot tap into a vast talent pool of AI governance professionals. This is a workforce that needs to be built and scaled at a pace commensurate with the development and deployment of the technology itself.</p> <p>If an AI governance workforce can professionalize, at scale, it can play a powerful role in developing, implementing and distributing innovative and trusted AI systems. Until then, AI outcomes in law, technology and society will vary as widely as the nature and impact of the risks.</p> <p>The time to professionalize was three years ago. It's not too late to catch up.</p> <p><i>J. Trevor Hughes is the president and CEO of the <a href="https://iapp.org/" target="_blank" rel="noopener">IAPP</a>, the professional home for privacy, AI governance and digital responsibility globally. With over 90,000 members in more than 150 countries, the IAPP provides training, certification, publications, research, events and networking opportunities to respond to growing need for professionals to manage the intersections of data, technology and humanity.</i></p> <p><i>Trevor has contributed to media outlets such as The New York Times, TechCrunch and Wired and has testified on issues of privacy, surveillance and privacy-sensitive technologies before U.S. Congress, the U.S. Federal Trade Commission, British Parliament and more. </i></p> Rules alone can't make AI safe -- IAPP's Trevor Hughes argues it's time to build and scale a professionalized AI governance workforce before risks outpace oversight. https://cdn.ttgtmedia.com/rms/onlineimages/ai_a279596285.jpg https://www.techtarget.com/ai/opinion/We-need-responsible-people-to-drive-responsible-AI Wed, 14 Oct 2026 11:06:00 GMT <p>The tech firms ploughing money into artificial intelligence (AI) infrastructure may not see a return on their capital investment until 2032, according to analyst Gartner.</p> <p>Although enterprise and consumer spending on AI has grown tremendously, Gartner’s latest report, <em>Prepare for AI market consolidation: Capital write-downs could exceed $4 trillion by 2032</em>, warned that the AI market is still nascent and may prove to be insufficient. In particular, a survey from Gartner found that 74% of CIO polled report that their AI initiatives are losing money or only breaking even.</p> <p>The analyst firm’s cautious outlook adds to growing concerns that the AI bubble is similar to the dotcom boom of the late 1990s and will eventually burst.</p> <p>Earlier this week, the <a href="https://www.computerweekly.com/news/366652153/IMF-warns-of-cyber-threat-to-digitally-connected-financial-markets">International Monetary Fund (IMF)</a> warned of of the risks. The IMF analysed 14 public transactions from AI, comprising $300bn of equity investment and $650bn of debt in exchange for $1.4tn of future revenue commitments.</p> <p>“What makes this a little bit unusual compared to previous deals is the future revenue commitments part,” warned Athanasios Vamvakidis, resident representative of the IMF in Croatia.</p> <p>He noted that equity investments may deplete in value, the debt may be difficult to service and the revenue commitments in the future may not materialise. This could have a knock-on effect based on what the IMF sees as the circularity of deals in the AI sector.</p> <p>Vamvakidis added: “If there are some kind of issues with those entities, then the circularity would pose risks to the market.”</p> <p>As the IMF noted, the financial market appears to be very optimistic about earnings forecasts. When AI firms beat these expectations, the market upgrades the forecast even higher. But there is a big question, according to the IMF, about whether this can continue or whether revenue growth will become increasingly more difficult.</p> <section class="section main-article-chapter" data-menu-title="AI spending growth"> <h2 class="section-title"><i class="icon" data-icon="1"></i>AI spending growth</h2> <p>Gartner has forecast that if <a href="https://www.computerweekly.com/news/366648782/Gartner-Agentic-AI-wont-benefit-from-economies-of-scale">AI capital spending</a> continues at its current trajectory, there is likely to be multitrillion-dollar write-downs of AI infrastructure assets by 2032. It warned that AI-optimised infrastructure-as-a-service (IaaS) spending will be too low for most AI compute providers to achieve sustainable returns on their capital assets before the current generation of semiconductors is fully depreciated.</p> <p>Most AI models can pretty much do the same job. Gartner senior director analyst Will Sommer said: “AI models are commodities. A lot of what an AI model can do can be achieved with almost every other model out there right now.”&nbsp;</p> <p>The challenge for AI infrastructure firms, according to Sommer, is their ability to monetise their graphics processor unit (GPU) AI accelerator hardware. “What AI firms are betting on is that they are able to develop the most advanced models that can self-improve faster than any [other model] can, which gives them sufficient competitive advantage relative to other AI model providers,” he said.</p> <p>As AI technology becomes more commoditised, he said AI model providers are trying to push enterprise customers into new areas that use more AI tokens or require higher value tokens.</p> <p>When looking at the market dynamics of what AI infrastructure providers are actually spending their money on, according to Sommer, around 30% is spent on purchasing and deploying AI training infrastructure. While this cost will reduce over time, he sees it as research and development (R&amp;D) spend.</p> <p>Out of the remaining 70%, between 30 to 50% is AI inference, which the service provider hopes will be funded by its customers paying to use the AI infrastructure. However, Sommer said utilisation rates are relatively low. “AI infrastructure providers are having a hard time getting the utilisation rates above 50% on a really good day,” he added.</p> <p>Then there is the consumer market, which accounts for up to about 35% of AI inference. “No one’s paying for AI subscriptions. Yes, there’s advertising revenue, and we take account of ad revenue in our analysis here, but it’s really hard to say that the future of AI is ad revenue,” he added.</p> <p>According to Sommer, the enterprise market represents “a fraction of a fraction of a fraction of the total compute that has been deployed”.</p> <p>To meet Gartner’s market forecasts for AI, Sommer said an AI infrastructure provider would need to make twice as much money on <a href="https://www.techtarget.com/it-strategy/feature/FinOps-for-AI-How-CIOs-are-navigating-tokenomics">AI token usage</a> as the investment it has made to provide these tokens. “Every single token that can be monetised would need to earn a margin of over 100% between 2026 and 2030,” he said.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more Tokenomics stories</h3> <ul class="default-list"> <li>Rising token bills force rethink of <a href="https://www.computerweekly.com/news/366649643/Rising-token-bills-force-rethink-of-AI-cost-management">AI cost management</a>: APAC enterprises are consuming more tokens even as unit prices fall, and the advice on what to do about it runs from smarter caching and gated model access to renting their own infrastructure.</li> <li>Boomi CEO shares vision of <a href="https://www.computerweekly.com/news/366645085/Boomi-CEO-shares-vision-of-AI-cost-management">AI cost management</a>: Steve Lucas, CEO of Boomi, believes the answer is prompt routing, which sends queries to the LLM with the lowest token cost and caches responses.</li> </ul> </div> </div> </section> Is the artificial intelligence bubble about to burst? Analyst Gartner does not expect CIOs will ramp up the amount they pay for AI inference https://cdn.ttgtmedia.com/visuals/ComputerWeekly/HeroImages/money-pounds-banknotes-cash-RZ-adobe.jpg https://www.computerweekly.com/news/366652193/Why-many-AI-companies-are-set-to-fail Wed, 14 Oct 2026 10:51:00 GMT <p>The Federal Energy Regulatory Commission approved a registry to track large data centers and other electricity loads on the PJM grid, while rejecting a proposal to curtail certain new loads during power emergencies.</p> <p>The <a href="https://elibrary.ferc.gov/eLibrary/docinfo?accession_Number=20261009-3055" target="_blank" rel="noopener">Oct. 9 order</a> gives PJM Interconnection a mechanism to collect information about facilities drawing at least 50 megawatts (MW) at a single electrical site. It rejects the regional grid operator's proposed Interim Resource Adequacy Service (IRAS), including its bring-your-own-new-capacity provisions and related capacity-market changes.</p> <p>For grid operators, the decision leaves a central problem unresolved: how to <a href="https://www.techtarget.com/it-infrastructure/news/366650699/Power-grid-operator-faces-data-center-stress-test">plan for large electricity demands</a> that may arrive before sufficient generation and transmission infrastructure is available. For data center developers, the approved registry offers no guarantee of grid access or clarity on what new curtailment obligations PJM could impose under a revised proposal.</p> <section class="section main-article-chapter" data-menu-title="Registry tracks demand, not project readiness"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Registry tracks demand, not project readiness</h2> <p>PJM's registry will collect information about large loads, including their locations, projected in-service dates, ramp schedules, peak demand, supply arrangements and backup generation.</p> <p>The information can help PJM forecast when new demand may connect and how quickly facilities could ramp toward full operation. The forecasts, however, depend on the maturity of the underlying projects.</p> <p>"A reported 500 MW project is not necessarily 500 MW of executable, energized or persistent operating demand," said Neil Osnato, founder of Persistence Analytics Group.</p> <p>A project may have commercial agreements and substantial financial commitments while still depending on unfinished generation, transmission or other infrastructure. Osnato separates project development into five stages: requested demand, contractually committed demand, executable infrastructure capability, energized load and sustained operating demand.</p> <p>Each stage represents a different level of readiness. Treating projected demand as equivalent to executable capacity can distort grid planning and capital allocation, Osnato said.</p> <p>The registry can improve PJM's visibility into the project pipeline. It does not independently verify that proposed facilities have the infrastructure, financing and equipment needed to connect on schedule or sustain their projected consumption.</p> </section> <section class="section main-article-chapter" data-menu-title="FERC rejects PJM's curtailment framework"> <h2 class="section-title"><i class="icon" data-icon="1"></i>FERC rejects PJM's curtailment framework</h2> <p>PJM proposed IRAS to address concerns that large new loads could connect before sufficient generation capacity becomes available. Under the proposal, certain new loads without qualifying new capacity could face curtailment during emergencies.</p> <p><a href="https://elibrary.ferc.gov/eLibrary/filelist?accession_Number=20260813-5118" target="_blank" rel="noopener">The filing also included provisions</a> for customers bringing new capacity and for excluding eligible new large loads from part of PJM's capacity-market demand calculation beginning with the 2029/2030 delivery year.</p> <p>FERC rejected the framework, raising concerns about state jurisdiction over retail customers, the treatment of different customer groups and PJM's justification for the proposal. The commission also questioned whether the framework would impose retail obligations or compensation terms beyond its authority.</p> <p>The order leaves <a href="https://www.techtarget.com/it-infrastructure/news/366650455/PJM-proposes-ride-through-rules-for-large-data-center-loads">PJM's existing reliability responsibilities</a> and operational authorities intact. It does, however, prevent the grid operator from implementing the rejected IRAS framework as proposed.</p> <p>Elizabeth K. Whitney, managing principal at Meguire Whitney, said the ruling highlights the jurisdictional challenge of assigning curtailment obligations through a multistate grid operator's tariff.</p> <p>Whitney pointed to<span style="text-decoration: line-through;"> </span>a <ins datetime="2026-10-14T10:15" cite="mailto:Snider,%20Shane"><a href="https://www.spp.org/documents/75954/20260210_revisions%20to%20add%20the%20conditional%20high%20impact%20large%20load%20service_er26-1323-000.pdf" target="_blank" rel="noopener">potential alternative</a></ins>. FERC accepted the framework in June. It provides conditional, non-firm transmission service to certain large loads while they await the resources or network upgrades required for long-term firm service.</p> <p>"The functional difference is that conditional transmission service is interrupted when thermal or voltage limits are exceeded, whereas a curtailment mandate is triggered when there is insufficient generation to meet load," Whitney said.</p> <p>The SPP framework permits service interruptions when transmission conditions require them. Its maximum seven-year term is intended to encourage customers to secure resources or complete network upgrades. Whitney said FERC's show-cause orders pointed to SPP's large-load framework as a model for other grid operators.</p> <p>PJM must determine whether a revised approach can address the commission's concerns while accounting for its own tariff and jurisdictional constraints.</p> </section> <section class="section main-article-chapter" data-menu-title="PJM reviews its next steps"> <h2 class="section-title"><i class="icon" data-icon="1"></i>PJM reviews its next steps</h2> <p>PJM said it is reviewing the order to determine how to meet FERC's requirements while recognizing responsibilities the commission indicated belong to the states.</p> <p>The grid operator proposed IRAS amid electricity demand growth that is outpacing new generation. PJM argued that existing customers should not bear the resulting costs or reliability risks, tying that position to its Ratepayer Protection Pledge, under which data centers and other large loads committed to help maintain grid reliability and affordability.</p> <p>"FERC's acceptance of our Large Load Registry and definition of large loads is a step in that direction," said Jeff Shields, a PJM spokesman.</p> <p>PJM declined to elaborate on its next steps while it reviews the order.</p> </section> <section class="section main-article-chapter" data-menu-title="Developers weigh grid access against reliability"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Developers weigh grid access against reliability</h2> <p>Whitney said data center developers should consider on-site generation and battery storage to reduce exposure to grid constraints. Conditional transmission service and curtailment arrangements can make grid supply less predictable for large loads, she said.</p> <p>On-site power can also provide a facility with greater operating flexibility while retaining access to grid service for reliability support. Developers must weigh those benefits against the cost of building and maintaining generation and storage, as well as the conditions attached to their grid connections.</p> <p>Whitney also criticized PJM's capacity-market structure, arguing that it has contributed to supply-demand imbalances, higher consumer prices and greater scrutiny from policymakers. Those are her assessments of the market, rather than findings in FERC's order.</p> <p>PJM now must implement the registry and determine whether to submit a revised resource-adequacy proposal. Meanwhile, grid planners still need to establish which proposed loads are likely to materialize, when supporting infrastructure will be ready and what authority exists to manage demand during a shortage.</p> <p>The registry will give PJM more information about the projects in its pipeline. Whether those projects can connect and operate as forecast -- and how the grid operator can manage reliability when demand outpaces available resources -- remains unresolved.</p> <p><i>Shane Snider is a senior news writer at </i>TechTarget<i>, covering AI infrastructure, hyperscale data centers, cloud platforms, and the power and energy systems driving modern compute expansion. You can reach Shane at </i><a href="mailto:shane.snider@informatechtarget.com"><i>shane.snider@informatechtarget.com</i></a><i> or on </i><a href="https://www.linkedin.com/in/shane-snider-1727755/" target="_blank" rel="noopener"><i>LinkedIn</i></a><i>.</i></p> </section> FERC approved PJM's registry for large electricity loads but rejected proposed curtailment rules, leaving grid reliability and new-load obligations unresolved. https://cdn.ttgtmedia.com/rms/onlineimages/strategy_g1159086107.jpg https://www.techtarget.com/it-infrastructure/news/366652121/FERC-approves-PJM-load-registry-rejects-curtailment-plan Wed, 14 Oct 2026 10:46:00 GMT <p>The current communications industry has literally reached for the stars, with space becoming not a final but a financial frontier for networks. Looking to tap into the space between, NTT Docomo and Abside Networks have signed a memorandum of understanding (MoU) towards the development of a future infrastructure using high-altitude platform stations (HAPS) and realisation of next-generation networks taking advantage of artificial intelligence (AI).</p> <p>HAPS is an area generally defined by unmanned aerial platforms that operate in the stratosphere at an altitude of approximately 20 kilometres. By equipping the platforms with wireless communications equipment, HAPS is being studied as a means of providing coverage in areas that have traditionally been difficult to serve, including airspace and maritime regions, as well as underserved locations such as sparsely populated and mountainous regions where conventional coverage has not been economically viable.</p> <p>Moreover, <a href="https://www.computerweekly.com/news/366649923/Sceye-SoftBank-demo-high-altitude-platform-systems-connectivity">players in the industry</a> regard the stratosphere as offering an optimal vantage point: it is close enough to Earth for high-capacity connectivity and observation, yet the altitude is high enough for wide-area reach. Potentially importantly, the stratosphere offers space-like conditions without the cost of being in space and the disadvantages of being in orbit.</p> <p>At the background of the MoU are studies and demonstrations that <a href="https://www.docomo.ne.jp/english/">Docomo</a> has been conducting toward the realisation of wide-area communications services utilising HAPS. The operator believes that to deliver high-quality communications services via HAPS, it is important to develop radio access networks (RANs) and communications equipment suited to the stratospheric environment.</p> <p>As conditions in the stratosphere differ from those on the ground, communications equipment and network design must be adapted accordingly, requiring specialised knowledge and experience.</p> <p>Through <a href="https://x.company/projects/loon/">Project Loon</a> – a stratospheric communications project led by Google that sought to provide communications services using high-altitude balloons – <a href="https://www.abside-networks.com/AI3DN/">Abside</a> has already developed knowledge and experience in stratospheric communications and has been engaged in the development of HAPS communications equipment and RAN technologies. It is also developing <a href="https://www.computerweekly.com/news/366651684/Sateliot-puts-five-more-satellites-into-5G-NTN-IoT-constellation">non-terrestrial network (NTN</a>) technologies and AI-based network control technologies.</p> <p>As a result, the two companies believe that their respective visions and technologies may be complementary and signed the MoU to jointly explore potential future collaboration toward networks that integrate the ground, stratosphere and space.</p> <p>The two companies said that they will conduct joint studies toward future collaboration in areas including wide-area, high-capacity and massive-connectivity communications using HAPS, network control and data utilisation. Under their initiative, they will jointly study technical collaboration for the future use of HAPS.</p> <p>The principal areas of study are system architectures for wide-area, high-capacity and massive-connectivity communications using HAPS; application of AI to demand forecasting and the control of beams, capacity, spectrum and interference in stratospheric RANs; collection and utilisation of data relating to communications quality, network operations and sensing in the stratosphere; future integration with external applications and service platforms.</p> <p>In addition, the two companies will bring together their respective expertise and technologies and pursue joint studies toward the realisation of next-generation communications infrastructure utilising HAPS and AI.</p> <p>This will include unmanned aerial platforms that operate in the stratosphere at an altitude of approximately 20 kilometres. By equipping the platforms with wireless communications equipment, the two firms noted that HAPS is being studied as a means of providing coverage in areas that have traditionally been difficult to serve. These include airspace and maritime regions, as well as underserved locations such as sparsely populated and mountainous regions where conventional coverage has not been economically viable.</p> <p>Another key project is <a href="https://www.abside-networks.com/AI3DN/">AI 3D Networks</a>, AI3D, is a vision said to be central to 6G deployments of a single AI-driven network spanning terrestrial, stratospheric and space-based layers. It built on the foundation of Abside’s AI3DR RAN product lines.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about HAPS</h3> <ul class="default-list"> <li><a href="https://www.computerweekly.com/news/366649923/Sceye-SoftBank-demo-high-altitude-platform-systems-connectivity">Sceye, SoftBank demo high-altitude platform systems connectivity</a>: Test programme travels more than 15,000km in the stratosphere, backhauls into operator’s core network, verifies direct-to-device and edge computing capabilities and Haps-based communications.</li> <li><a href="https://www.computerweekly.com/news/366619933/MWC-2025-NTT-Docomo-preps-for-beyond-5G-future">NTT Docomo preps for beyond 5G future</a>: Leading Japanese operator unveils slew of announcements to boost advanced communications across domestic territory and Africa, encompassing the new era of 6G and aircraft-based systems.</li> <li><a href="https://www.computerweekly.com/news/365529874/BT-tests-out-high-altitude-flying-antenna">BT tests out high-altitude flying antenna</a>: The trial with Stratospheric Platforms may lead to hydrogen-powered drones being used to provide 4G and 5G rural coverage.</li> <li><a href="https://www.computerweekly.com/news/366559632/UK-companies-secure-contract-to-trial-sky-based-5G-infrastructure">UK companies secure contract to trial sky-based 5G infrastructure</a>: Aircraft manufacturer, communications firm and cooling systems company join forces to broaden and transform access to high-speed 5G communications through fleet of zero-emission turbo prop aircraft.</li> </ul> </div> </div> Comms equipment company and operator see complementary nature of respective visions and technologies and explore potential collaboration toward networks that integrate ground, stratosphere and space https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/Conceptual-image-aircraft-in-stratosphere-PR-hero.jpg https://www.computerweekly.com/news/366652057/Docomo-Abside-Networks-sign-high-altitude-comms-MoU Wed, 14 Oct 2026 10:28:00 GMT <p>As <a href="https://www.softcat.com/solutions">Softcat</a> delivers another solid set of full-year numbers, the channel player’s CEO has reflected on its approach to the market and adoption of artificial intelligence (AI), and mused on the prospects of further acquisitions.</p> <p>Investors already knew the business was going to deliver strong numbers, but woke up to hear that Softcat had reported gross invoiced income improving 30.5% and gross profit growth of 24.7%, amplifying the benefits of being able to provide a broad range of products and services.</p> <p>The results, which come a day after <a href="https://www.computerweekly.com/microscope/news/366652099/Bytes-adds-MSSP-player-and-shares-H1-update">Bytes updated the market</a> and in the wake of strong <a href="https://www.computerweekly.com/microscope/news/366650126/Computacenter-delivers-record-breaking-H1">H1 numbers from Computacenter</a>, have underlined the strength of the leading channel firms in what continues to be a challenging market.</p> <p>Graham Charlton, Softcat CEO, was upbeat about its performance, and believed the decisions it had made over the past few years and the moves it continues to make have put the business in a good position for the future.</p> <p>He said the business had been concentrating on delivering breadth and depth to customers, and that was one of the key reasons it had grown wallet share with existing users and attracted more into the fold.</p> <p>“The quality of that offering across the full range, that really is what pays dividends for us,” said Charlton. “You can’t shortcut to that. We’ve been building that by organic investment over 33 or 34 years now, so it gives us a sustainable and very hard-to-replicate source of advantage.”</p> <section class="section main-article-chapter" data-menu-title="Investing in AI"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Investing in AI</h2> <p>On the AI front, the business has been developing its own agents and incorporating it into more of its operations to unlock efficiencies. Investments have also been made in back-end systems to put the business on a firm footing as demands on its business processes increases.</p> <p>“We’ve been working on our own technology for five or six years now,” said Charlton. “We’ve implemented a lot of new core systems, a new finance system, a new sales system, a new HR system.</p> <p>“We’ve also been developing our own internal tools,” he added, stating that its CatNav tool enabled staff to navigate through its systems. “Even new account managers who are just learning the trade have got access to what is the equivalent of giving a satellite navigation system to someone driving a car. It’s a game changer in terms of being able to get where you want to go much quicker.</p> <p>“We’re only at the beginning of the opportunity that we can deliver now into our own operations through the upgrades and the modernisation we’ve done to our technology, and some of these new tools that we’re building,” said Charlton. “What we can do with the offering we take to customers and how we deploy that in the years ahead is really exciting, now that we’ve got these foundational building blocks in place.”</p> </section> <section class="section main-article-chapter" data-menu-title="Expanding the business"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Expanding the business</h2> <p>Charlton has also overseen the expansion of the business, notably moving into the US for the first time this summer with the <a href="https://www.computerweekly.com/microscope/news/366650632/Softcat-makes-1bn-move-for-GDT">acquisition of GDT</a>.</p> <p>“I think we’ve got three engines now,” he said. “We haven’t just got Softcat in the UK and Ireland. We haven’t just got GDT in the US. We’ve got the combination of those two things as well, which is something incremental on top of that.</p> <p>“The market and the opportunity in the UK and Ireland is still huge, and will afford us the chance to compound double-digit growth for the next 10 or 20 years, no doubt about it,” said Charlton. “GDT have the same opportunity in the US, and by operating together selectively, where it makes sense, we can offer something to large, complex international customers that neither business can do in isolation.”</p> <p>He said the key was to expand while retaining the approach it had been so successful with over the past three decades. “It’s [about] making sure that the UK and Ireland business and the job we do for our customers continues to grow in the way that it always has,” said Charlton.</p> <p>“GDT are on the same path in the US, and we can accelerate that with our know-how and patient investment, and we can create the right collaboration and centralisation, where it adds value for customers now or people between the two businesses.”</p> <p>There have been a couple of acquisitions in the past few years, but Charlton is keen to stress there is no rush to make further M&amp;A moves.</p> <p>“We’re always interested in the possibility,” he said. “We have a really clear strategy. We know what we want to build, and sometimes acquisitions can accelerate that. We’re always looking, but we’ve got a very high bar, and our most important avenue of growth will always be organic. But if there’s something that makes real sense to make us go quicker, then we have the power and the ambition to act on that as well.</p> <p>“I don’t know if we’ll make any more acquisitions in the next few years,” said Charlton. “We will continue to look at it because that research effort helps inform the organic investments that we’re making anyway.”</p> <p>The story of Softcat over the past few years has been one of calculated investments, in people, systems and M&amp;A, as the business ensures it has put down the foundations to support future growth.</p> <p>“It’s built into our cadence, and I sometimes refer to it as fixing the roof while the sun shines, because when you’re delivering 20-25% growth, if you’re not investing in the future, then you’re missing the most important point of how you run your business,” said Charlton. “We’ve got a great track record of transforming the offering, broadening it, deepening it. You can’t ever stop with that, and the faster you go, the more you can invest, the stronger your future, and the bigger your competitive advantage becomes. That is exactly the model we have, and we’ll continue to pursue.”</p> </section> As the channel player shares its FY26 numbers, it is a moment to reflect on a strategy that continues to deliver results https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/seedings-nature-growth-1-adobe.jpeg https://www.computerweekly.com/microscope/news/366651842/Softcat-CEO-talks-of-investing-in-future-growth Wed, 14 Oct 2026 10:00:00 GMT 60 webmaster@techtarget.com