Tenstorrent, a chip startup focused on inference, has raised $1.5 billion at a $6.2 billion valuation, TechCrunch reported on Monday — a financing that more than doubles what the company was worth in late 2024 and lands five months after two of the biggest names in chips reportedly tried to buy it.
The 10-year-old company is one of a wave of AI hardware startups trying to chip away at Nvidia’s dominance. Its bet is narrower than most: instead of chasing the training market Nvidia owns, Tenstorrent is building chips tailored for serving open-weight AI models — the kind anyone can download and run — and for video generation.
The exit it didn’t take
Five months ago, Tenstorrent reportedly received acquisition interest from Intel and Qualcomm at valuations of as much as $5 billion. The new round values the company above even the top of that range, and its leadership says the decision to stay independent came down to a demonstration.
Jim Keller, the prominent microprocessor engineer who serves as Tenstorrent’s chief executive, told Bloomberg that the company decided to remain independent and raise more capital after it showed that its Galaxy servers could run DeepSeek models. Running a leading open-weight model family well, on its own servers, was the proof point that let Tenstorrent argue it had a business worth funding rather than a team worth acquiring. The claim is the company’s own account of a turning point, not an independent benchmark — but it was enough to anchor a $1.5 billion raise.
Who wrote the checks
The round was led by Trace Capital, a new firm launched by Quiet Capital co-founder Ben Mahdavi, and Innovation X. Other backers included Cadence Ventures, Eclipse, Fidelity, SanDisk and T. Rowe Price, according to TechCrunch.
The $6.2 billion valuation more than doubles the $2.6 billion Tenstorrent fetched in late 2024 — a jump that reflects how aggressively investors are pricing any credible alternative in AI silicon, even one whose products serve a niche. Tenstorrent’s new capital will go toward scaling its chips for open-weight model serving and video generation.
A wave of expensive challengers
Tenstorrent is not alone in commanding a startling price. Etched, another AI chip maker, was fielding investment offers at valuations of $40 billion or more, TechCrunch reported earlier this month. The pattern is consistent: Nvidia’s grip on AI hardware has made even a small, defensible slice of the market — inference rather than training, open-weight models rather than proprietary ones — worth billions to investors who believe the AI build-out will need more than one supplier.
Open-weight serving is a plausible slice to defend. Models released with open weights can be run by anyone, on any hardware that runs them efficiently, which makes the serving layer a contest on cost and performance rather than on access to a proprietary ecosystem. That is the contest Tenstorrent says its Galaxy servers entered when they ran DeepSeek’s models — and the one it now has $1.5 billion to scale into.
What the raise doesn’t answer
For all the money, Monday’s announcement was light on operating detail. Neither TechCrunch’s report nor Keller’s comments to Bloomberg included revenue figures, shipment numbers or independent performance comparisons, and a valuation set in a private round is a price on potential, not proof of traction. The acquisition interest from Intel and Qualcomm cuts both ways: it validates that established chipmakers take Tenstorrent’s technology seriously, and it marks the price at which they valued it only five months ago — below what private investors have now agreed to pay.
What is clear is the choice. Tenstorrent could have become part of Intel or Qualcomm. Instead, with a DeepSeek demonstration as its evidence and a fresh $1.5 billion as its runway, it is betting that the market for running open AI models is big enough — and open enough — for an independent chip company to grow into a $6.2 billion valuation.
