SDSignal Desk

Tenstorrent raises $1.5B at $6.2B valuation after passing on a possible sale

· Updated · Signal Desk newsroom

Analysis of reporting by TechCrunch. How we work

Image: TechCrunch

Jim Keller’s Tenstorrent raised $1.5 billion at a $6.2 billion valuation, more than double its late-2024 mark, five months after Intel and Qualcomm reportedly showed acquisition interest at up to $5 billion. The money funds inference chips for open-weight models and video generation.

Key takeaways

  • Tenstorrent raised $1.5 billion at a $6.2 billion valuation, up from $2.6 billion in late 2024.
  • Intel and Qualcomm reportedly showed acquisition interest at valuations of up to $5 billion.
  • Trace Capital and Innovation X led; Cadence Ventures, Eclipse, Fidelity, SanDisk and T. Rowe Price joined.
  • Keller told Bloomberg the company chose independence after showing its Galaxy servers could run DeepSeek models.
  • TechCrunch reports rival Etched has fielded offers at $40 billion or more.

The round

Tenstorrent, the inference-focused chip startup led by veteran processor designer Jim Keller, has raised $1.5 billion at a $6.2 billion valuation, TechCrunch reports. That is more than double the $2.6 billion valuation from late 2024, and it comes five months after Intel and Qualcomm reportedly showed acquisition interest at valuations of up to $5 billion. Trace Capital and Innovation X led the round, with Cadence Ventures, Eclipse, Fidelity, SanDisk and T. Rowe Price participating. Keller told Bloomberg the company chose independence after demonstrating that its Galaxy servers could run DeepSeek models. The money goes toward chips built to serve open-weight AI models and video generation.

Tenstorrent is about ten years old and positions itself as an inference specialist rather than a training-chip competitor. Open-weight models are those whose parameters are published for anyone to run, as opposed to closed models served only through a company’s own API.

An inference bet on open models

The AI hardware market has long depended heavily on one company, and inference is where costs accumulate once models start answering real requests. A chip company that explicitly targets open-weight models is building for the part of the ecosystem anyone can download, modify and run, which benefits developers and buyers who do not want every workload priced by a single vendor. A well-funded alternative supplier also strengthens enterprises’ negotiating leverage on inference costs. Efforts like AWS TorchNeuron (TorchNeuron) show the same push against NVIDIA’s software advantage.

Keller’s DeepSeek demonstration aims at the open-weight world, where Chinese labs have been among the most aggressive releasers. That is commercially smart but ties Tenstorrent to a model supply that is politically contested in Washington; tighter restrictions on Chinese models could make its most compelling demos harder to sell to some customers.

Independence and competition

Passing on a possible acquisition keeps control of the roadmap but means competing alone against NVIDIA’s software lock-in, long the real moat. A large round buys time, not developers, so software and tooling need to keep pace with hardware. Tenstorrent sits in a crowded wave of NVIDIA challengers; TechCrunch reports Etched has fielded offers at $40 billion or more, and valuations are climbing fast. NVIDIA, for its part, is financing parts of the custom-silicon market (NVIDIA and MediaTek). Beating NVIDIA on a benchmark is easier than winning procurement. More in our Nvidia coverage, funding coverage and startups coverage.

Who feels it

Developers
More hardware for serving open-weight models, if software matures.
Enterprises
Stronger leverage on inference costs.
NVIDIA
A capitalized challenger in inference.
Investors
Rapidly rising chip valuations with much riding on execution.

What to watch

  1. Named production inference customers.
  2. Benchmarks and pricing for open-weight and video models.
  3. Software tooling progress versus NVIDIA.
  4. U.S. policy on Chinese open-weight models.

FAQ

How much did Tenstorrent raise?

$1.5 billion at a $6.2 billion valuation.

Who leads Tenstorrent?

Veteran processor designer Jim Keller.

Why didn’t Tenstorrent sell?

Keller told Bloomberg it chose independence after showing its Galaxy servers could run DeepSeek models.

What will the money fund?

Chips for serving open-weight AI models and video generation.

Read the original

Continue at the source.

TechCrunch

Companies: NVIDIA

More on NVIDIA

All NVIDIA coverage →