Tenstorrent raises $1.5B at $6.2B valuation to scale open-weight AI chips

Tenstorrent, the AI chip startup led by veteran microprocessor engineer Jim Keller, has raised $1.5 billion at a $6.2 billion valuation. The round, confirmed by the company and first detailed in a Bloomberg interview with Keller, lands just five months after the ten-year-old Nvidia challenger reportedly drew acquisition interest from Intel and Qualcomm at valuations of as much as $5 billion.

The number matters. Tenstorrent last raised capital in late 2024 at a $2.6 billion valuation, according to the company's disclosed figures. The new round therefore more than doubles the startup's worth in roughly a year. It also prices Tenstorrent above the level at which Intel and Qualcomm were reportedly prepared to buy it.

Who put up the $1.5 billion?

Trace Capital led the round alongside Innovation X. Trace Capital is a new firm launched by Ben Mahdavi, co-founder of Quiet Capital, and Tenstorrent's raise represents a high-profile early commitment from the fund.

The syndicate extends well beyond the two leads. According to details of the round, the participant list included:

• Cadence Ventures, the investment arm of chip-design software maker Cadence
• Eclipse, a venture firm with a long record in deep-tech and infrastructure investing
• Fidelity, the asset manager
• SanDisk, the memory and storage company
• T. Rowe Price, the asset manager

The presence of strategic investors from adjacent layers of the hardware stack — EDA (electronic design automation, the software used to design chips) and storage — signals that established industry players see Tenstorrent's roadmap as credible enough to fund at multibillion-dollar scale.

Why did Tenstorrent stay independent?

Jim Keller, Tenstorrent's CEO and one of the best-known microprocessor engineers in the industry, gave Bloomberg a direct account of the decision. Keller told Bloomberg that the company decided to remain independent and raise more capital after it showed that its Galaxy servers could run DeepSeek models.

That is the pivotal technical fact behind the financing. Tenstorrent's Galaxy servers — the company's rack-scale inference systems — demonstrated they could serve DeepSeek's open-weight models (AI models whose trained parameters are publicly available for download and deployment). The demonstration validated the company's core thesis: that its silicon can competitively serve open-weight AI workloads rather than depending on closed, proprietary model ecosystems.

Keller's own framing to Bloomberg, kept verbatim from the interview, is that the company "decided to remain independent and raise more capital" on the strength of that result. The choice carries weight because Keller's track record gives it credibility; he is described in the source as a prominent microprocessor engineer whose career spans major chip architectures.

What will the money fund?

The new capital will go toward scaling chips tailored for serving open-weight AI models and video generation, the company indicated. Both targets are deliberate.

Open-weight models such as DeepSeek's have shifted the economics of AI deployment: enterprises can download the weights and run inference (the process of executing a trained model to generate outputs) on hardware they control. That shift creates demand for inference-optimized chips, which is precisely Tenstorrent's stated focus as a startup. Video generation adds a second, computationally heavy workload — generating video requires sustained inference throughput, a load that rewards purpose-built silicon.

Tenstorrent is not building a general-purpose accelerator. It is building chips for serving models whose weights anyone can obtain. That positioning distinguishes it from incumbents whose margins rest on a closed ecosystem spanning chips, software and models.

How big is the threat to Nvidia?

Nvidia dominates the market for AI accelerators, and Tenstorrent belongs to a wave of startups trying to chip away at that dominance. The source places Tenstorrent explicitly in that cohort.

The cohort is attracting capital at steep valuations. Among Tenstorrent's peers is Etched, an AI chip maker that was fielding investment offers at valuations of $40 billion or more, as TechCrunch reported earlier this month. The Etched figure — more than six times Tenstorrent's new valuation — illustrates the scale of investor appetite for credible Nvidia alternatives, even before those companies have matched the incumbent's installed base or software maturity.

The stakes are straightforward. Nvidia's grip on AI compute shapes what infrastructure costs, who can deploy frontier models, and how quickly open-weight alternatives spread. Every funded challenger changes the bargaining position of cloud providers and enterprises building AI systems.

What does the valuation jump signal?

Three data points frame Tenstorrent's trajectory:

• Late 2024: $2.6 billion valuation in its prior raise
• Roughly five months ago: reported acquisition interest from Intel and Qualcomm at valuations of as much as $5 billion
• Now: $6.2 billion valuation on $1.5 billion in fresh capital

The sequence shows a company that chose the public-market-style path — more capital, more dilution, continued independence — over an exit at a lower number. It also shows investors repricing the company upward by $1.2 billion beyond the top of the reported acquisition range, a premium the company earned, in Keller's telling, by proving its Galaxy servers could run DeepSeek models.

What comes next?

The $1.5 billion gives Tenstorrent the balance sheet to scale production of chips for open-weight model serving and video generation. The test ahead is execution: converting a demonstrated capability — DeepSeek models running on Galaxy servers — into shipped silicon at volume, in a market where Etched is commanding reported valuations above $40 billion and Nvidia's position remains the reference point every challenger is measured against.

By Marcus Bennett