Jim Keller's Tenstorrent More Than Doubles Its Valuation to $6.2 Billion on a $1.5 Billion Series E

Tenstorrent, the AI chip startup led by legendary chip architect Jim Keller, said on Monday it has raised $1.5 billion in a late-stage funding round that values the company at $6.2 billion post-money — more than double its valuation from December 2024. The Series E was led by Trace Capital and Innovation X, with participation from Cathie Wood's ARK Invest, Sandisk, Morgan Stanley, T. Rowe Price and Tether Investments.

The round is a sharp step up from the company's previous financing, which raised just over $693 million in December 2024 at a $2 billion pre-money valuation (about $2.6 billion post-money, as Bloomberg's framing of the new round notes), led by Samsung Securities and AFW Partners. The new $6.2 billion figure arrives after Bloomberg reported earlier this year that Tenstorrent had attracted takeover interest — interest the company evidently chose to answer with a growth round instead of a sale.

The Santa Clara, California-based company says it will use the capital to scale production of its Galaxy servers, rack systems that combine dozens of Tenstorrent's own AI chips to power applications including chatbots, coding assistants and video generators. Unlike nearly every other AI accelerator company, Tenstorrent builds both its chips and its software stack on RISC-V, the open-source instruction set architecture that lets companies design processors without license fees and without dependence on proprietary ecosystems like Arm's.

Keller is one of the most credentialed chip designers in the industry: he led Tesla's first self-driving chip program and held key design roles at Apple and AMD, among others. His pitch is that the AI buildout's bottleneck is shifting from raw FLOPS to cost per token, and that an open, license-free architecture with a cleaner software story can undercut the incumbent GPU ecosystem on economics rather than peak performance.

The raise lands in a crowded lane. NVIDIA still owns the vast majority of AI accelerator spending, AMD's Instinct line has landed hyperscaler design wins, and a wave of well-funded challengers — Cerebras, Groq, SambaNova and ASIC efforts inside Google, Amazon and Broadcom's custom-silicon business — are all chasing the same inference budget. Tenstorrent's differentiation is deliberate: it sells both chips and full servers, licenses its technology, and bets that open standards will win the developers who want to tune their own stacks.

What the round buys is time and manufacturing capacity. Galaxy servers are a systems business, which means inventory, supply chain and validation costs that a chip-licensing model avoids — and the company will need reference customers running production inference workloads, not benchmarks, to justify the new valuation. If AI infrastructure spending keeps compounding into 2027, a $6.2 billion entry price for one of the few independent RISC-V bets may look cheap. If inference prices keep collapsing the way model prices have, every accelerator startup's math gets harder, Tenstorrent's included.