Tenstorrent has turned reported takeover interest into a far larger wager on going it alone.

Earlier this year, Intel and Qualcomm were said to have explored buying the Santa Clara inference-chip specialist at valuations of up to $5 billion. Tenstorrent declined to address what it called “market speculation,” but the message of its latest financing is clear: the company would rather build than sell.¹

On Monday, the 10-year-old startup announced a $1.5 billion round at a $6.2 billion post-money valuation, more than doubling the $2.6 billion valuation it secured in late 2024. Trace Capital and Innovation X led the investment, with a broad investor roster including Cadence Ventures, Fidelity, SanDisk, T. Rowe Price, Morgan Stanley and Tether Investments.²

The decision followed a technical proof point that helped shape Tenstorrent’s strategy. Chief executive Jim Keller told Bloomberg that the company chose to remain independent and raise capital after demonstrating that its Galaxy servers could run DeepSeek models. The new money is earmarked for chips built to serve open-weight AI models and video-generation workloads.¹

That puts Tenstorrent in a difficult but increasingly crowded race. Its original pitch was to win a slice of Nvidia’s market through inference chips and licensing designs to companies seeking custom silicon. It has since expanded into servers, and plans to use the financing to scale manufacturing.²

The two accounts align on the central calculation: Tenstorrent sees more value in staying independent than accepting a deal. But independence now comes with a sharper test — proving that a well-funded challenger can translate open-model momentum into hardware sales in Nvidia’s backyard.