What Tenstorrent does
Tenstorrent designs AI compute hardware and software intended to provide an open alternative across scales, from individual accelerator cards to enterprise scale-out servers. The company emphasizes open architectures, vendor-neutral IP, and open-source software to avoid lock-in for AI developers and organizations.
Key capabilities
Its hardware lineup includes Blackhole accelerator cards with passive, active, and liquid cooling options; TT-QuietBox liquid-cooled workstations capable of running models in the range of around 120 billion parameters; and Galaxy scale-out servers for enterprise AI deployment. On the software side, TT-Forge is an open-source MLIR-based compiler compatible with PyTorch, JAX, and ONNX. Tenstorrent also licenses its AI and RISC-V-oriented IP for specific workloads.
Who it's for
The AI infrastructure offering targets researchers, developers, and enterprises seeking transparent, ownable silicon and open tooling rather than proprietary, closed AI compute stacks. The differentiator is openness across hardware, software, and licensable IP.
