The AI hardware boom just produced another giant number. Tenstorrent, the chip startup run by legendary processor architect Jim Keller, has raised $1.5 billion in a late-stage funding round that values the company at $6.2 billion, TechCrunch and Reuters both reported on Monday — more than double what the company was worth less than two years ago, and a fresh statement that the race to build alternatives to Nvidia is nowhere near slowing down.
The Series E round was led by Trace Capital — a new firm launched by Quiet Capital co-founder Ben Mahdavi — together with Innovation X, according to TechCrunch. The rest of the investor list, as reported by Reuters, reads like a roll call of money that wants a piece of the AI build-out: ARK Invest, the firm run by Cathie Wood; SanDisk; Morgan Stanley; T. Rowe Price; Tether Investments; Fidelity; Eclipse; and Cadence Ventures. When both Wall Street institutions and a stablecoin giant’s investment arm show up in the same round, the round is not really about one product. It is about positioning for a decade of compute demand.
From $2.6 Billion to $6.2 Billion in Under Two Years
The jump is steep by any measure. TechCrunch notes the $6.2 billion post-money valuation more than doubles Tenstorrent’s December 2024 valuation of $2.6 billion. That earlier round, in December 2024, raised more than $693 million at a $2 billion pre-money valuation and was led by Samsung Securities and AFW Partners. In other words, investors who got in at the end of 2024 have watched the paper value of the company roughly triple in about 22 months — the kind of markup that either signals a genuine breakout or a market pricing in hope at full volume. Probably, as usual, a bit of both.
What changed in between? Start with the road not taken. Keller told Bloomberg that the company chose to stay independent and raise fresh capital after its Galaxy servers demonstrated they could run DeepSeek’s models — a proof point that apparently mattered a great deal to buyers and investors alike. And independence was a real choice, not a default: TechCrunch reports that five months ago, Intel and Qualcomm both showed acquisition interest in Tenstorrent at valuations of up to $5 billion. Keller and his board looked at exits worth up to $5 billion and decided the better bet was raising at $6.2 billion and staying on the field. That is confidence, or stubbornness, or both — and Keller’s career suggests he does not scare easily.
The Keller Factor
It is impossible to tell Tenstorrent’s story without its CEO. Jim Keller is one of the most decorated chip architects alive: he led Tesla’s self-driving chip work and held senior design roles at both Apple and AMD, where his fingerprints are on processor generations that reshaped those companies’ fortunes. In an industry where the scarcest resource is people who have actually shipped great silicon, Keller is a proven quantity, and investors are paying for that track record as much as for any single product.
The money, per the company’s plans reported Monday, goes to scaling its Galaxy servers — systems that combine dozens of Tenstorrent chips into machines built to power chatbots, coding assistants and video generators — and to developing chips tailored specifically to open-weight AI models. That second phrase is the strategic heart of the whole company.
Open Chips Against a Closed Empire
Tenstorrent builds on RISC-V, the open-source processor architecture that carries no license fees. That is a deliberate shot at the two dominant models in the industry: Nvidia’s closed, tightly controlled stack, and Arm’s proprietary licensing ecosystem. Tenstorrent’s pitch is that as open-weight models proliferate — models anyone can download, inspect and run — the hardware underneath them should be open too, cheaper to build on and free of toll booths.
Whether that pitch wins is the multi-billion-dollar question. Nvidia’s grip on AI training and inference is not just about chips; it is about software, networking and an ecosystem developers already know. But the market Monday’s round reveals is one where very serious money believes the empire can be chipped at — pun fully intended — especially in inference and in serving open models, where cost per token rules everything. It is the same restless energy running through today’s other tech headlines, from the OpenAI bugging case in San Francisco to Anthropic’s $48 billion public benefit pledge: the AI industry is simultaneously building fortunes and arguing about what it owes the world.
There is also a buyer-side story here. The labs and cloud providers spending tens of billions on AI infrastructure have every incentive to cultivate a credible second source — someone to negotiate against, and someone to call when the dominant supplier's roadmap does not match their own. Open-weight models strengthen that hand: when the model can run anywhere, the hardware decision becomes a genuine contest again, judged on cost and performance rather than lock-in. Tenstorrent's entire design philosophy — dozens of chips ganged into Galaxy systems, silicon tuned for exactly those open models — is built to win that kind of contest.
For now, Tenstorrent has $1.5 billion of fresh ammunition, a valuation that says the market believes, and a CEO who has beaten longer odds than these. The Galaxy servers will have to do the rest of the talking. Follow the money and the machines with our news coverage, and check the day’s other big stories in games and sports.
