What Big Tech's AI spending means for your IT budget

The scale of Big Tech's AI spending

Hyperscaler infrastructure spending nearly tripled from early 2018 to $142 billion during the third quarter of 2025 alone, according to Synergy Research Group. It's a trend that is likely to grow in the coming years.

"By the time we get to 2027, there will have been a trillion dollars invested just in servers and the network infrastructure to run AI by hyperscalers," John Lovelock, distinguished VP analyst, chief of research at Gartner, said. "That doesn't include the data center buildout, the power supply requirements, the staffing, the Opex to run the data centers, but a trillion dollars in Capex on IT."

Capital intensity has surged to levels previously unthinkable for technology companies, with some hyperscalers now spending more than half of their revenue on Capex. These ratios are more like those of industrial or utility companies than those of traditional tech firms.

The comparison to utilities is apt, according to Vamsi Duvvuri, EY Americas technology, media and telecommunications AI leader. This scenario is similar to when electricity and railroads became capital-intensive utilities, and enterprises needed to update machinery and upskill labor. Today, companies must update hardware, software and workforce capabilities.

The infrastructure itself is different from traditional IT spending. GPU clusters cost orders of magnitude more than conventional compute. Power and cooling requirements can exceed hardware expenses, and technology refresh cycles are much faster.

"These companies are engaging in a land grab where they are securing infrastructure and AI model access for the next decade," Scott Bickley, advisory fellow at Info-Tech Research Group, commented.

Why Alphabet's bond sale matters to IT executives

A century bond reflects expectations that AI infrastructure will deliver value across decades, not quarters. This reframes how enterprises should think about their own AI investment horizons.

Balance sheet flexibility. Even cash-rich companies are choosing debt over depleting reserves.

"Alphabet is treating AI infrastructure as a long-term asset to preserve balance-sheet flexibility because it believes AI will become foundational to its platform," Savio Lobo, CIO at Ensono, said.

Investor confidence gap. While investors seem willing to fund century bonds for hyperscaler AI buildouts, most CIOs must justify AI spending quarterly, while boards expect comparable transformation. This gap creates pressure that many IT leaders struggle to navigate.

The growing budget gap for enterprise IT

The disconnect between hyperscaler spending, board expectations and IT budget realities is widening.

Jim Rowan, managing director at Deloitte, noted that his firm's research has found that nearly three in four companies plan to deploy agentic AI within two years. Only one in five report having a mature governance model for autonomous agents.

"Boards will always desire clear ROI metrics from any implementation effort, but with most organizations still early in their AI adoption journey, this can be difficult to achieve, especially as budgets often focus on the technology alone," Rowan said.

Where AI costs hide

CIOs should view AI as an ecosystem, not just technology, according to Duvvuri. This means tracing costs through process redesign, data pipelines, infrastructure, security, risk and compliance, people and vendors.

Real costs show up as:

• Over-investments in duplicated models, shadow AI tools and untracked experimentation cycles.
• Under-investments in scalable processes for design, build and tracking that encompass more than just technical considerations.

Why hyperscaler strategies don't translate

"Enterprise CIOs should not use Alphabet's or other hyperscalers' approach to AI acquisition as a guidepost," Bickley said.

The tension between innovation pressure and fiscal discipline intensifies as organizations realize they cannot replicate hyperscaler approaches.

"Big Tech is building the grid. Enterprises should focus on the factories and workflows that use that electricity efficiently, safely and at scale," Chirag Mehta, vice president and principal analyst at Constellation Research, commented.

Rethinking how AI fits into the IT budget

Given that enterprises cannot replicate hyperscaler spending models, IT leaders need a different budget-planning framework.

"You can't even say, 'Are you investing in AI?' Because everyone is," Lovelock said. "This year, there will be more money spent on software with AI in it than on software without."

For most CIOs, Lovelock noted, this represents a rebranding of dollars already being spent. For example, CRM software without AI simply becomes CRM software with AI.

"AI is a technology that sprinkles on top of everything that we do, like salt. And like salt, it's going to make everything taste better," Lovelock added.

Capex vs. Opex considerations. The traditional question of whether to classify AI as Capex or Opex misses a fundamental shift.

"The biggest mismatch occurs when boards fail to distinguish between optimization, core product or growth," said Jerry Shu, co-founder and CTO at Daylit.

AI for operational efficiency should be Opex. Infrastructure for new revenue streams represents Capex. Capex covers reusable core components such as data platforms and governance frameworks, according to Lobo. Variable compute and licensing remain Opex.

Treating AI as a portfolio. Mehta recommends allocating 70% to scale proven workflows, 20% for adjacent expansions and 10% to frontier bets.

Phasing investments to match adoption. Many organizations commit too much too early.

"We see horror stories where companies sign large contracts, but actual internal adoption is incredibly low," Shu noted.

Practical budget strategies for AI adoption

Getting AI projects funded requires reframing IT cost management conversations from technology to value. The most effective CIOs present unit economics that finance teams can pressure test.

Framing AI spend in financial terms. Bickley advises framing AI use cases in terms of:

• Revenue expansion.
• Cost reduction.
• Working capital improvement.
• Risk mitigation.
• Customer satisfaction.

Starting with measurable workflows. Duvvuri recommends breaking down AI adoption into four levels:

• AI for people (Copilot) drives personal productivity.
• AI on platforms (CRM, ERP) delivers functional improvements.
• AI for processes (lead-to-cash) requires many teams working together.
• AI native products represent a true transformation.

Proving ROI incrementally. Eric Helmer, global CTO at Rimini Street, sees concrete ROI when applying AI to ERP core functions like vendor onboarding and inventory management. "If done right, each AI project should deliver enough ROI to justify and fund the next," he said.

Lobo advises being disciplined with initial investments and prioritizing early wins. Ensono invested in specific business cases that improved time-to-diagnose and incident prediction before making larger, long-term capability investments.

Using cloud and consumption-based pricing. Baron Fung, senior research director at Dell'Oro Group, recommends starting in public cloud to gain experience before evaluating on-premises deployment. Looking at specific tactics, Mehta recommends the following:

• Commit capacity only when utilization is predictable.
• Use stage gates that de-risk spending.
• Design for model churn so you can swap without rebuilding workflows.

Building compelling business cases. The strongest AI business cases tie investments directly to outcomes that matter to the business. Helmer recommends that CIOs focus on concrete applications such as vendor onboarding, inventory management and procurement, where AI delivers measurable returns.

"In every case, the best tactic is to do everything with a measurable business outcome in mind," Helmer said. "Faster does not automatically deliver value and ROI."

Look beyond immediate financial metrics. Rather than focusing solely on financial outcomes, examine interim metrics tied to effective AI integration, such as adoption rates, productivity gains and risk mitigation, according to Rowan.

Creating shared accountability. AI funding decisions work best when the C-suite is involved from the start.

• Lobo recommends securing CEO alignment around a clear AI investment thesis developed with the CFO.
• Duvvuri recommends establishing an AI FinOps team where every AI model has a lifecycle budget, owner and success metric.

What IT Leaders can learn from Big Tech's AI spending

Alphabet's bond sale offers lessons for enterprise IT, but not the obvious ones. The takeaway isn't to borrow for AI infrastructure; it's to adopt the strategic discipline that justifies such long-term commitments.

Borrow the discipline, not the dollars

Split spending into durable foundations versus fast-changing bets, then finance them differently, advises Mehta. Fund durable foundations, such as data readiness, security guardrails and evaluation frameworks, with multi-year commitments. Fund fast-changing elements, such as model choices, in short tranches of 90 to 180 days tied to measurable outcomes.

Use stage-gated budgeting tied to adoption and unit economics. Commit capacity only when utilization is predictable. Match AI funding models to your organization's risk tolerance and operational realities, not what hyperscalers can afford.

Think like venture capitalists, not private equity

Leaders need to evolve their thinking from a private equity lens to a venture capital lens, according to Duvvuri. Private equity expects stable returns over two to four years, with short-term returns covering implementation costs. Venture capital accepts that long-term returns outweigh the cost of experiments, with staggered funding as value gets proven.

Build an integrated strategy, not isolated pilots

The biggest mistake is treating AI initiatives as isolated pilots rather than part of an integrated value strategy, according to Rowan. Organizations struggle to move from experimentation to production because they lack clarity about measurable outcomes.

"Stop random acts of AI," Mehta said. "Pick three to five workflows where you can prove impact, then industrialize that pattern."

Duvvuri recommends avoiding the trap of just funding use cases. Scaling AI from a strategic lens cannot be done through use cases alone; they are the unit of execution. Instead, fund capabilities, both experimental and operational, that lead to outcomes.

The long-term competitive edge

Alphabet's bond sale reminds us that AI payoff operates on long timelines. Success depends on pacing, prioritization and financial clarity.

AI isn't a short-term phenomenon; it's a broader transformation.

"The biggest thing CIOs have to wrap their head around is that we are shifting super cycles," Lovelock said. "The last 20 years have been about digital technology, digitizing business. The next 20 years will be about putting intelligence into everything we do. This is not a fad you can get out of the way of."

Sean Michael Kerner is an IT consultant, technology enthusiast and tinkerer. He has pulled Token Ring, configured NetWare and been known to compile his own Linux kernel. He consults with industry and media organizations on technology issues.