Andromeda Cluster

Andromeda Cluster, founded in 2022, operates a global GPU compute marketplace that connects AI teams with high-performance training infrastructure across dozens of providers worldwide. The platform handles the sourcing, benchmarking, pricing, certification and deployment of GPU capacity, so that researchers and engineers can obtain frontier-scale compute without negotiating hardware contracts. The company was founded by Nat Friedman and Daniel Gross, initially to serve their venture portfolio and the wider startup community.

The company's technical work spans GPU compute infrastructure, ML training infrastructure, hardware benchmarking, compute pricing and certification, and compute deployment automation. Its platform began with a single multi-thousand GPU cluster and has since scaled to a network of providers; over a billion GPU-hours have flowed through it. Engineers work on problems of market infrastructure as much as of raw compute: reliable benchmarking of heterogeneous hardware, pricing and certification of capacity, and automation of deployment across providers.

Andromeda serves customers in artificial intelligence, AI research, machine-learning startups and AI model training, with a global geographic footprint. The stated ambition is to become a liquidity layer for global AI compute - market infrastructure comparable to the electrical grid - with an emphasis on letting engineers and researchers concentrate on building models rather than on procuring hardware.