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

Tenstorrent raises $1.5 billion to stay independent

Tenstorrent, a ten-year-old startup building processors for AI inference, has raised $1.5 billion at a $6.2 billion valuation, according to TechCrunch. The round lands five months after the company reportedly attracted acquisition interest from Intel and Qualcomm at valuations of up to $5 billion, interest it has now set aside in favour of continuing on its own.

The new figure is a sharp step up from Tenstorrent's previous mark. The company was valued at $2.6 billion in late 2024, meaning the latest round more than doubles its worth in under two years.

Who is backing the round

The financing was led by Trace Capital and Innovation X, TechCrunch reports. Trace Capital is a newly launched firm started by Ben Mahdavi, who previously co-founded Quiet Capital. They were joined by Cadence Ventures, Eclipse, Fidelity, SanDisk and T. Rowe Price.

The investor mix is notable: alongside financial institutions, the round pulls in names from across the hardware supply chain, including Cadence's investment arm and storage maker SanDisk, suggesting strategic as well as purely financial conviction in Tenstorrent's roadmap.

A benchmark that tipped the decision

Jim Keller, Tenstorrent's CEO and a prominent microprocessor engineer, told Bloomberg that the company decided to remain independent and raise more capital after demonstrating that its Galaxy servers could run DeepSeek models.

The choice of workload is telling. DeepSeek's models are among the most widely used open-weight releases, and showing that Tenstorrent's hardware can serve them gives the company a concrete proof point for prospective customers. The fresh capital will go toward scaling chips built specifically for serving open-weight AI models and video generation.

One of several challengers to Nvidia

Tenstorrent is part of a wider crop of AI hardware startups trying to loosen Nvidia's hold on the accelerator market. TechCrunch notes that the group includes Etched, an AI chip maker that was recently fielding investment offers at valuations of $40 billion or more.

The money flowing into these companies reflects a shared investor thesis: demand for AI compute is large enough, and the supplier base concentrated enough, to make room for serious alternatives, particularly in inference, where cost per query matters more than raw training throughput.

Why it matters

Attention in AI hardware has long centred on training frontier models, a segment Nvidia effectively owns. Inference, running trained models to answer real requests, is widely expected to become the bigger and more price-sensitive half of the market as AI deployment spreads. Tenstorrent is betting that open-weight models, which anyone can download and host themselves, will drive a large share of that demand, and that silicon designed from the ground up for inference can undercut general-purpose GPUs on efficiency.

The raise also signals that investors see an independent path for chip challengers rather than an exit to an established buyer. Tenstorrent passed on the acquisition interest from Intel and Qualcomm, and its backers are instead financing a direct competitor to the incumbent. For organisations wary of depending on a single GPU supplier, a credible, well-funded alternative focused on open-weight serving is exactly the kind of option the market has lacked.

Tenstorrent remains a small challenger set against a vastly larger incumbent, but the combination of demonstrated model support, Keller's engineering reputation and $1.5 billion in new capital makes it one of the more serious attempts yet to build that alternative.