Jim Keller’s Tenstorrent raises $1.5B at $6.2B valuation to expand open AI compute platform
Tenstorrent raised $1.5B at a $6.2B valuation to expand its open RISC-V AI compute platform and challenge Nvidia.
The Series E round was jointly led by Trace Capital and Innovation X, with participation from new and existing backers including Protagonist, Morgan Stanley, Fidelity Management & Research Co., ARK Invest and T. Rowe Price, SiliconANGLE reported. Tenstorrent said the capital will be used to ramp production of its Galaxy servers for global deployment to support developers running large language models and video generation workloads, fund research in the RISC-V ecosystem and expand its workforce.
Keller, the company’s chief executive, is widely regarded as a legend in the semiconductor industry, having previously helped design processors for Intel Corp., Advanced Micro Devices Inc., Apple Inc. and Tesla Inc. Tenstorrent started as a chip designer but pivoted to sell a full portfolio of AI compute platforms, including Galaxy servers, TT-QuietBox workstations and Blackhole cards. Those systems are powered by its Tensix Neo accelerators and Ascalon central processing units, both based on the open-standard RISC-V instruction set, and are designed to be scalable and energy-efficient for AI workloads.
The company’s best-selling products are the Galaxy Tenstorrent servers, enterprise-class systems that can be linked into large superclusters for training and running frontier models. Tenstorrent said Virtu Financial Inc. uses the servers to power on-premises AI trading agents, while data center colocation provider Equinix Inc. makes them available to enterprises. The TT-QuietBox is a liquid-cooled desktop AI workstation that the company says can test and run cutting-edge models with minimal setup. It is powered by Blackhole cards, scalable PCIe boards with RISC-V cores and GDDR6 memory; multiple cards and workstations can be networked to pool memory and compute.
Tenstorrent also develops an open-source software stack for its hardware, including TT-Forge, a compiler that bridges machine learning frameworks such as PyTorch to its RISC-V chips; TT-Metalium, a software development kit for hardware optimization and custom kernel development; and TT-LLK for low-level kernel optimizations. Keller said the platform approach lets Tenstorrent scale from a single workstation to large cloud-based clusters. “We’ve built one architecture and we sell it at every size,” he said in the SiliconANGLE report. “The future of compute is open, scalable and owned by the people building on it.”
Holger Mueller, an analyst at Constellation Research, told SiliconANGLE that Tenstorrent is among a growing number of startups trying to build more efficient computing platforms for AI, and that Keller’s track record makes it one of the most promising. “Tenstorrent gets the funding it needs to keep innovating, and this is good news for enterprises,” Mueller said. “Nvidia is not the only option anymore. More competition and deployment options for AI compute can only be beneficial for enterprises, making AI workloads cheaper and more flexible.”
Protagonist co-founder and Managing Partner George Bousis said Tenstorrent’s full platform approach sets it apart from startups seeking to provide alternatives to Nvidia’s GPU ecosystem. “Few companies design the chip and the software that runs on it, but Tenstorrent does both, and its video generation results beat industry benchmarks ahead of our expectations,” Bousis said in the report.