Poolside is no longer just training models — it’s building an empire of compute
The US-based AI company is partnering with cloud infrastructure provider CoreWeave on Project Horizon , a 500-acre data center campus in West Texas designed to generate its own power from nearby natural gas fields. The plan: transform a remote ranch into a vertically integrated compute and energy hub, with CoreWeave anchoring the first 250 MW of capacity and another 500 MW reserved for expansion. Initial operations are expected in late 2026 , with full buildout by 2027 .
Meanwhile, Nvidia is preparing to invest up to $1 billion in Poolside as part of a $2 billion round valuing the company at around $12 billion, though some reports suggest it could reach up to $14 billion — nearly five times its $3 billion valuation last year. The round, already over half committed, includes $700 million from existing investors and potential participation from MagnetarDealroom has a profile for this one. Try Dealroom →. The capital will fund the purchase of more than 40,000 Nvidia GB300 NVL72 GPUs , deepening a partnership that is both financial and strategic.
Founded in 2023 by Jason WarnerDealroom has a profile for this one. Try Dealroom → (former GitHub CTO) and Eiso KantDealroom has a profile for this one. Try Dealroom → , Poolside began as a coding automation startup. Now it’s pursuing control over the entire AI stack — from energy and silicon to intelligence itself. Project Horizon is its most ambitious step yet: an attempt to make compute as abundant and reliable as electricity in the industrial age.
But delivering a 2-gigawatt data center — roughly the power of Hoover Dam — will test the limits of execution. The facility’s success hinges on managing cooling, grid stability, and environmental impact at an unprecedented scale. By colocating next to existing gas infrastructure and reusing a prebuilt plant, Poolside and CoreWeave aim to reduce one of AI infrastructure’s biggest costs: energy . Their modular, phased buildout approach lets them deploy capacity efficiently while controlling capital risk.
Still, the true challenge lies beyond construction. Poolside must convert this compute capacity into differentiated AI products and sustainable revenue to justify the capital intensity of its infrastructure vision. If it succeeds, it won’t just train the next generation of models — it will own the infrastructure that makes them possible .
Sources:
M.A. / A.M.
