The problem
Every serious AI lab rents the same vendor’s GPUs at monopoly prices, writes to a proprietary software stack it cannot fork, and has no realistic path to designing its own silicon — x86 and Arm licences plus closed toolchains keep custom AI chips a hyperscaler-only game.
How it works
Founded in Toronto in 2016 by Ljubisa Bajic, Milos Trajkovic, and Ivan Hamer — with legendary chip architect Jim Keller joining as CTO in 2020 and becoming CEO in 2023 — Tenstorrent builds AI computers around its Tensix cores across the Grayskull, Wormhole, and Blackhole product lines. The differentiator is openness: open-source software stacks that let developers own their tooling, and licensable AI plus RISC-V IP for customers that want custom silicon, manufactured through partners such as Samsung and TSMC. Series D money is earmarked for the open-source stacks, developer hiring, and new global design centres.
Pain points
GPU cost and allocation shortages, CUDA-only codebases, and instruction-set licensing fees that price all but the largest firms out of custom silicon.
Business model
Selling AI hardware (processors, systems, and AI clouds) and licensing silicon IP — with early demand from Samsung for manufacturing and Hyundai for automotive AI systems.
Challenges
Nvidia ships at a scale and software maturity Tenstorrent cannot yet match; the $150M contract base must convert into repeat volume; and each new architecture generation (through Blackhole) burns cash long before it earns it.
Funding
- Raised: $693M+ Series D (Dec 2024) at a $2B pre-money valuation, led by Samsung Securities and AFW Partners with Bezos Expeditions, LG, Hyundai, Fidelity, Baillie Gifford, Export Development Canada, and the Healthcare of Ontario Pension Plan; $1.18B+ total across 10 rounds.
- Valuation: $2B pre-money at the December 2024 round.
Latest — November 2025
Tenstorrent is in advanced talks to raise $800M in new funding at roughly a $3B–$3.2B valuation led by Fidelity, as investor appetite for a credible open-architecture alternative to Nvidia keeps intensifying.