Tenstorrent

Tenstorrent is an AI infrastructure hardware company that designs open-source RISC-V based chips, workstations, and servers for running AI workloads at scale, with pricing starting at $999 for the Blackhole card. It targets AI researchers, ML engineers, enterprise infrastructure teams, and sovereign AI deployments seeking alternatives to proprietary GPU vendors without vendor lock-in.

Founded in 2016 and led by legendary chip architect Jim Keller, Tenstorrent offers a tiered product lineup designed to scale from individual developers to hyperscale deployments. The Blackhole™ card starts at $999 and can be added or configured into custom rigs with passive, active, or liquid cooling options. The TT-QuietBox™ workstation starts at $11,999, offering whisper-quiet liquid-cooled performance capable of running models up to 80 billion parameters from a desk. For production deployments, the Tenstorrent Galaxy™ servers provide sovereign, scale-out infrastructure. Based on our analysis of 870+ AI tools, Tenstorrent stands out as one of the few hardware vendors committed to fully open architectures — including open-source silicon IP that can be licensed without lock-in.

The software stack is anchored by TT-Forge™, an MLIR-based open-source compiler currently in public beta that works with PyTorch, JAX, ONNX, and other major ML frameworks. Compared to proprietary AI accelerator platforms in our directory like NVIDIA CUDA or Google TPUs, Tenstorrent differentiates through complete transparency — all repositories are open source on GitHub, the company runs an active bounty program rewarding community contributions (recent bounties include optimizing atan2, log1p, and sin/cos/tan operations), and IP can be licensed directly. Partnerships with Infinia Technologies for sovereign AI infrastructure and participation in the CHASSIS Program reinforce the company's position as an alternative for organizations prioritizing supply-chain independence and architectural openness.