OctoAI

01 / WHAT IT BUILT

What the company did

OctoAI was a Seattle-based AI inference platform, founded in 2019 as a University of Washington spin-out of the Apache TVM project and led by CEO Luis Ceze. It served open-source foundation models (Llama, Mixtral, SDXL, Stable Diffusion, Whisper) behind OpenAI-compatible REST APIs, plus OctoStack for private VPC and on-premises deployment, helping companies run generative AI models more efficiently. It raised about $130 million in total, including an $85 million Series C in 2021 at a roughly $900 million valuation, from Madrona, Amplify Partners, Tiger Global, and Addition. In September 2024, it was acquired by Nvidia for a reported $165 million (potentially over $250 million with retention incentives) — a steep down-round exit for investors; OctoAI wound down commercial availability of its services on October 31, 2024, and Ceze joined Nvidia as Vice President of AI systems software.

03 / EVENT HISTORY

Events in the structured record

1. 2019
   Founded

   OctoAI (formerly OctoML) was founded as a University of Washington spin-out of the Apache TVM project, focused on machine-learning model optimization.

2. 2021
   Raised $85M Series C

   OctoAI raised an $85 million Series C at a roughly $900 million valuation from Madrona, Amplify, Tiger Global, and Addition.

3. 2024
   Acquired by Nvidia

   In September 2024, Nvidia acquired OctoAI for a reported $165 million; commercial services wound down on October 31, 2024, and CEO Luis Ceze joined Nvidia.