Anjney Midha founded AMP, which coordinates compute for independent AI teams, and is a venture partner at Andreessen Horowitz. Before investing in AI infrastructure, he built computer-vision and developer-platform businesses at Ubiquity6 and Discord.[1][2][3]
Start with The Need for an Independent AI Grid →Education
Links
AMP a16zFounder · AMP
Venture Partner · Andreessen Horowitz
Visiting Scientist, Applied Physics · Stanford University
Career
Midha co-founded and led Ubiquity6, which worked on computer vision and multiplayer technology. Discord acquired the company, and he became vice president of platform ecosystem. Andreessen Horowitz’s biography credits him with establishing Discord’s dedicated developer-platform organization and launching its Midjourney partnership. Earlier, he was a partner at Kleiner Perkins, where he helped form a dedicated seed fund. He joined a16z in 2023 and its current biography identifies him as a venture partner.[1]
Contributions
In his conversation with OpenRouter co-founder Alex Atallah, Midha discusses what happens after a model has been trained. Developers need a way to discover it, compare available providers and integrate it into applications. The interview considers model diversity, marketplaces for inference and the security problems around valuable token flows. His contribution is investment and infrastructure building around those markets, alongside the operators developing the model-serving products.[2]
Ideas
AMP’s thesis starts with a mismatch between research teams and the computers they need. A lab may run a large training job, then have quieter periods or a different mix of inference work. Provisioning for each peak can leave expensive capacity unused. AMP proposes pooling demand across teams whose busy periods differ, while sharing scheduling and infrastructure work. Its current site identifies National Compute as the operating grid. The aim is to give focused labs access to scale while letting them remain independent.[3]
Tools, writing & talks
Writing, research & talks
Start here
The Need for an Independent AI Gridamppublic.comAMP’s thesis for pooling infrastructure across independent research teams.
Read AMP’s grid thesis to understand the resource-sharing problem before the investment discussion. Basic familiarity with AI training helps, but the demand-pooling explanation needs no programming.
Founder’s organization; company thesis.[3]Conversation on model distribution, inference marketplaces and the economics of token use.
Interview participant.[2]Original sourceOpenRouter's CEO and an a16z partner discuss building a multi-model routing layer, why labs struggle with distribution, usage rankings as a market signal, model fusion experiments, and agent-driven token fraud.
Catalog workOriginal source- Anjney Midha’s a16z writinga16z.com
Biography and an archive of investment writing and conversations.
Investor and author.[1]