Tenstorrent raises $1.5 billion for open-model chips
Tenstorrent reached a $6.2 billion valuation after raising $1.5 billion to scale inference hardware for open-weight models and video generation.
The financing
AI chip startup Tenstorrent raised $1.5 billion at a $6.2 billion valuation, more than doubling the company’s reported value from late 2024. The round was led by Trace Capital and Innovation X, with participation from Cadence Ventures, Eclipse, Fidelity, SanDisk and T. Rowe Price.
Tenstorrent is positioning itself as an alternative to Nvidia, with a focus on chips and servers designed for inference. CEO Jim Keller said the company chose to remain independent after demonstrating that its Galaxy server systems could run DeepSeek models. The new capital will support production scale and hardware tailored to serving open-weight models and video-generation workloads.
The company’s timing reflects a shift in the accelerator market. The AI industry is no longer defined only by training the largest closed models; open-weight systems are increasingly being deployed by cloud providers, enterprises and developers who want control over weights, data location and operating costs. That creates demand for hardware optimized around inference economics rather than only peak training performance.
The constraint
Tenstorrent’s valuation is substantial, but funding does not establish that its products can compete broadly with Nvidia’s software, supply chain or installed base. Its reported DeepSeek demonstration is evidence of capability, not proof of production-scale adoption. The company must still turn architecture and server claims into reliable developer tools, predictable availability and competitive total cost of ownership.
The round nevertheless matters because it gives a non-Nvidia accelerator company enough capital to pursue an integrated hardware and software strategy while open models are expanding the addressable market. If Tenstorrent succeeds, the beneficiaries would include model developers seeking alternatives and customers that need to run large models without committing to one dominant vendor.
Uncle Cat take: The $6.2 billion valuation is a bet on inference demand, but Tenstorrent’s real test is whether DeepSeek demos become boring, repeatable production deployments.