The AI Stack
Sign in

Tenstorrent

Tenstorrent develops RISC-V-based AI processors and scalable systems challenging NVIDIA dominance.

Updated April 2026

Overview

Founded
2016
Headquarters
Toronto, Canada
Segment
GPU & AI Accelerators

Product overview

Tenstorrent designs AI accelerators like Grayskull, Wormhole, and Blackhole chips, along with RISC-V CPUs (Ascalon), chiplets, boards, workstations (QuietBox), and servers (Galaxy) for cloud, edge, automotive, and sovereign AI.. Customers include LG, Hyundai, Samsung, and partners like GlobalFoundries, TSMC for production; they serve AI developers, hyperscalers, automotive, and defense sectors.. Distinct from NVIDIA via open-source RISC-V architecture, modular chiplets, no HBM dependency, flexible IP licensing, and emphasis on performance-per-watt efficiency with scalable NoC interconnects.

Revenue model

Licenses AI/RISC-V IP, sells chiplets/chips, AI boards/servers/systems; $40M+ 2024E, ~$200M 2025E revenue with $150M customer contracts.

Moat

Tenstorrent's key competitive moat is its proprietary chiplet-based AI processor architectures like Grayskull, Wormhole, and TT-Ascalon, featuring custom Tensix cores with integrated network-on-chip for efficient tensor computations, scalability from edge to data centers, and cost advantages from using GDDR6 instead of expensive HBM. This is bolstered by an open-source software ecosystem, high customizability via RISC-V IP with innovation licenses allowing customer modifications, and superior flexibility for dynamic AI workloads compared to NVIDIA GPUs.