Nvidia Faces Another Challenger as Tenstorrent Raises 1.5 Billion Dollars at a 6.2 Billion Valuation
Jim Keller's AI chip company Tenstorrent has closed a $1.5 billion Series E at a $6.2 billion post-money valuation, led by Trace Capital and Innovation X, to scale production of its Galaxy servers. It reportedly held takeover talks with Intel and Qualcomm earlier this year before choosing to raise money instead.
Nvidia's list of would-be challengers just got a very well-funded new entry. AI computing company Tenstorrent announced on October 19 that it has closed a $1.5 billion Series E at a $6.2 billion post-money valuation, led by Trace Capital and Innovation X. Backers include Fidelity, ARK Invest, T. Rowe Price, Morgan Stanley and Tether Investments.
The company is run by chip veteran Jim Keller, who summed up the pitch this way: "We built one architecture and sell it at every size." The money is earmarked for a few concrete things:
- Ramping production of its Galaxy servers for large language model and video generation workloads, per SiliconANGLE
- Advancing its RISC-V technology, expanding its workforce globally and continuing to license CPU and AI core IP to chip designers
- Global commercial deployment, in a market where the company says it can compete on price and take orders that may be too small for bigger rivals
The valuation more than doubles the $2.6 billion Bloomberg reported after its previous round in 2024. The round is also notable for what it avoided: TNW reports that Tenstorrent held takeover talks with Intel and Qualcomm earlier this year, and that Keller told Bloomberg it chose to raise money instead of selling.
Performance claims should be read with care. One analysis relays Tenstorrent's claim that Galaxy runs DeepSeek R1 at 350 tokens per second at about one-fifth the per-token cost of Nvidia's GB300, which is the company's own figure and untested by independent buyers.
Nvidia's lead rests as much on software as on silicon, so the real test for Tenstorrent is whether customers will run production workloads on an open, RISC-V-based alternative at the scale this funding is meant to support.
