Tenstorrent Raises $1.5 Billion to Scale RISC-V AI Servers

Tenstorrent has raised $1.5 billion to scale its RISC-V-based AI servers, according to a report by Finimize. The figure marks one of the largest single funding rounds disclosed this year by an AI chip startup that does not rely on Arm or x86 licensing.

The company builds AI accelerators paired with RISC-V processor cores. The new capital will fund expansion of its AI server business, the report said. Tenstorrent positions its open-instruction-set approach as an alternative to Nvidia's CUDA-dominated ecosystem.

What does the round signal?

The $1.5 billion raise lands amid heavy investor demand for AI compute startups. Most funding in the segment has flowed to companies building on conventional architectures. Tenstorrent's bet differs. It pairs custom AI silicon with RISC-V, the open instruction set architecture that requires no license fees.

The company has not disclosed a valuation or investor breakdown in the report. It also has not stated a shipment or capacity target for its server products.

RISC-V has gained traction in embedded and low-power markets. AI servers are a far more demanding segment. They require high memory bandwidth, mature software stacks and large-scale deployment support. Tenstorrent's ability to convert capital into competitive server products will test whether the open architecture can carry inference and training workloads at datacenter scale.

Who is behind Tenstorrent?

The startup is led by industry veteran Jim Lohr... no — led by chief executive Jim Keller, one of the best-known chip architects in the industry. Keller previously held senior design roles at AMD, Apple, Tesla and Intel. His involvement has anchored investor confidence since he joined the company.

The report did not name the investors in the round. Tenstorrent had raised earlier rounds from strategic backers including semiconductor and cloud players, but the Finimize report gives no breakdown of the new $1.5 billion tranche.

Why RISC-V matters for AI servers

Nvidia controls the large majority of AI accelerator shipments. Its advantage rests as much on software as on silicon. The CUDA toolchain locks in developers across training and inference.

Tenstorrent's counter is an open stack. RISC-V cores let the company control its own roadmap without licensing constraints. Servers built this way could appeal to buyers seeking supply diversity, particularly in markets with restrictions on US-origin technology.

The commercial question is execution. AI server customers buy proven throughput, not architecture arguments. Tenstorrent must now show deployment wins and revenue scale.

What comes next

The $1.5 billion gives Tenstorrent runway to expand production and engineering. The company has not published timelines for new silicon generations or volume server shipments. Competitors in the custom and open AI silicon segment continue to raise capital at record pace, so the window to establish RISC-V servers as a credible alternative to Nvidia-based systems will depend on how quickly Tenstorrent converts this round into shipped systems and paying datacenter customers.