Tenstorrent raises $1.5B to challenge Nvidia in AI servers and chip IP
AI chipmaker reaches $6.2B valuation as it expands server production and licenses processor IP, but it remains to be seen whether it can offer a viable Nvidia alternative.
Tenstorrent has raised $1.5 billion at a $6.2 billion post-money valuation, giving the AI chipmaker fresh capital to scale production of its Galaxy servers and expand its processor licensing business as it competes for workloads dominated by Nvidia.
The Series E round, which Tenstorrent said Monday it had closed, was led by Trace Capital and Innovation X, with participation from Fidelity Management & Research Company, T. Rowe Price, SanDisk, ARK Invest, Morgan Stanley and existing investors.
Tenstorrent plans to use the funding to expand Galaxy server manufacturing and commercial deployments, develop new generations of its AI and RISC-V CPU intellectual property, and hire globally. The company did not disclose how it would allocate the capital among those priorities or provide a revenue forecast.
For enterprise IT buyers, the central question is whether Tenstorrent can turn its alternative AI architecture into a dependable source of production compute. The company has named several commercial deployments and partners but has not disclosed revenue or a comprehensive breakdown of system shipments.
Virtu Financial has deployed Galaxy servers on-premises for AI agents supporting trading and operations. Equinix offers Galaxy systems through its Distributed AI Hub, while Smallest.ai runs its Lightning V2 real time text-to-speech model on Galaxy Blackhole servers for organizations that want to operate voice AI on-premises. In Japan, ai& has deployed more than 120 Galaxy systems as part of its sovereign AI infrastructure, according to Tenstorrent.
The new valuation is more than twice the roughly $2.7 billion implied post-money value of Tenstorrent's December 2024 financing, which raised more than $693 million at a $2 billion pre-money valuation.
"The $6.2 billion post-money figure is ambitious but not crazy in today's AI infrastructure market," said Stephen Sopko, practice lead for semiconductor and deep tech at HyperFrame Research.
"Think of it as paying a premium for a proven architect before the foundation is fully poured," Sopko said, pointing to CEO Jim Keller's track record and investor appetite for alternatives to Nvidia’s hardware and software stack. The valuation, he said, leaves room for execution risk because the company still needs to demonstrate that commercial deployments can scale predictably.
Inference offers a path to compete
Tenstorrent builds AI processors around its Tensix AI cores and develops RISC-V CPU technology. It sells compute systems ranging from individual cards and workstations to Galaxy servers and larger superclusters, while licensing its processor designs to companies developing their own silicon.
The company is positioning its systems for inference -- running trained models to generate responses, code, images and video -- rather than trying to displace Nvidia across the entire AI market. Its open software stack and use of standard Ethernet are intended to give customers more control over their infrastructure, including organizations that want to run open-weight models on-premises or build sovereign AI systems.
Tenstorrent claims its Galaxy systems deliver leading performance on some inference workloads. Artificial Analysis' live provider benchmark for the LTX-2.3 Fast video model lists Tenstorrent among four providers and shows recent generation times of roughly 9 to 12 seconds for a six-second, 1080p video. The other listed providers took longer in those examples. Results vary by prompt and test, so the figures should not be treated as a universal ranking across AI workloads.
Sopko said the commercial opportunity is clearest where customers prioritize inference performance, cost and control over a single vendor's software ecosystem. But a benchmark win does not establish total cost of ownership across a production deployment, including utilization, networking, software engineering, support and the availability of replacement hardware.
Chip licensing broadens the business
Tenstorrent's licensing business gives it another route to market beyond selling complete systems. The company licenses its TT-Ascalon RISC-V CPU and Tensix AI cores to chip designers, including Samsung and LG, according to its company materials.
The systems and IP businesses could reinforce each other. Galaxy deployments give Tenstorrent a way to demonstrate its architecture in complete systems, while licensing can extend its technology into customer-designed chips. But the two businesses also require different kinds of execution: systems manufacturing and customer support on one side, and design enablement and software support for external silicon developers on the other.
"The open software stack helps both sides, but execution risk rises if the company tries to be excellent at full systems, chiplet IP, and developer tooling at once," Sopko said.
Tenstorrent must also compete with the software ecosystem surrounding Nvidia GPUs. An open stack can give customers flexibility, but buyers still need to evaluate model support, the effort required to port workloads, production reliability and the support available when problems arise.
Production scale is the next test
Tenstorrent says its Galaxy Blackhole systems are in production and shipping in volume. The company has also described superclusters built from dozens of Galaxy systems. Those statements indicate progress beyond laboratory demonstrations, but they do not establish how quickly the company can meet larger orders or sustain deliveries across multiple customers.
Building a few hundred servers is one thing; predictable multi-thousand unit ramps are another.
Stephen SopkoPractice lead, HyperFrame Research
Manufacturing capacity and component supply will be central to that effort. So will evidence that customers are expanding deployments after initial installations, rather than treating Galaxy as a limited alternative for selected workloads.
"Building a few hundred servers is one thing; predictable multi-thousand unit ramps are another," Sopko said.
Enterprise buyers should look for independent benchmarks under production-like conditions, repeat orders from named customers, model breadth, transparent delivery commitments and credible multi-year product road maps, Sopko said. Evidence that chip-licensing customers are bringing their own designs to market would also help establish whether Tenstorrent's IP business can grow alongside its systems operation.
The $1.5 billion round gives Tenstorrent more capital to pursue both strategies. Its next challenge is converting that funding and its technical claims into sustained deployments, dependable supply and a business that can support its valuation.
Shane Snider is a senior news writer at TechTarget, covering AI infrastructure, hyperscale data centers, cloud platforms, and the power and energy systems driving modern compute expansion. You can reach Shane at [email protected] or on LinkedIn.
