Tenstorrent Inc., the AI chip startup led by semiconductor veteran Jim Keller, said it closed a $1.5 billion late-stage funding round that values the company at $6.2 billion. The Series E was co-led by Trace Capital and Innovation X, with new and existing investors including Protagonist, Morgan Stanley, Fidelity Management & Research Co., ARK Invest and T. Rowe Price. The Toronto-based company is using the capital to challenge Nvidia Corp.'s dominance in AI computing hardware.
Tenstorrent began as a chip designer but now sells a full lineup of AI compute systems. Its products include Galaxy servers, TT-QuietBox workstations and Blackhole cards for laptops, all built on its Tensix Neo accelerators and Ascalon CPUs, which use the open-standard RISC-V instruction set. The company says the designs scale efficiently and deliver strong AI performance without excessive power use.
The Galaxy servers are Tenstorrent's top sellers, positioned as enterprise systems that can be linked into large superclusters for training and running frontier models. Virtu Financial Inc. uses them for on-premises AI trading agents, and data center colocation provider Equinix Inc. offers them to enterprises. For local workloads, the liquid-cooled TT-QuietBox desktop workstation runs advanced models with minimal setup, powered by scalable Blackhole PCIe cards with RISC-V cores and GDDR6 memory.
Tenstorrent also develops the software its hardware depends on, including an open-source stack meant to let developers run any AI workload on its chips. The stack comprises TT-Forge, a compiler linking frameworks such as PyTorch to its RISC-V silicon; TT-Metalium, an SDK for hardware optimization and custom kernel development; and TT-LLK for low-level kernel work. Keller said the company built a single architecture and sells it at every size, describing the future of compute as open and scalable.
Constellation Research analyst Holger Mueller said Tenstorrent is among a growing group of startups pursuing more efficient AI computing platforms, and that Keller's track record makes it one of the most promising. He said the funding is good news for enterprises because Nvidia is no longer the only option, and more competition can make AI workloads cheaper and more flexible.
The new funds will support a global ramp-up of Galaxy server production for developers running large language models and video generation workloads, and will back research into the RISC-V ecosystem and global hiring. Protagonist co-founder and Managing Partner George Bousis said few companies design both the chip and the software that runs on it, and that Tenstorrent does both, with video generation results that beat industry benchmarks ahead of expectations.
More company and startup news from TechManNews.
