OpenAI Triples Compute to 1.9GW, Nvidia Backs Nebius in AI Expansion
OpenAI’s computing capacity more than tripled to 1.9 gigawatts in 2025, with new data centers prioritizing deployment of Nvidia’s latest GPUs across the U.S., Europe, the Middle East and South America. Nvidia also backed infrastructure startup Nebius and is deepening its AI build-out as Meta rolls out custom AI chips.
1. OpenAI Compute Build-Out
OpenAI saw its compute capacity more than triple to roughly 1.9 gigawatts in 2025 under industrial compute lead Sachin Katti, reflecting an aggressive push to support next-generation AI. The company aims to add a gigawatt of infrastructure weekly and is evaluating new facilities across the U.S., Europe, the Middle East and South America.
2. NVIDIA Hardware Role
OpenAI has prioritized Nvidia’s newest GPUs for its global expansion, securing multibillion-dollar supply agreements and choosing sites that allow rapid deployment of high-density AI servers. Nvidia chips remain central to training advanced models despite challenges like power and memory constraints.
3. Nvidia’s Investment in Nebius
Nvidia provided strategic backing to AI infrastructure startup Nebius, reinforcing its stake in the data center equipment market and supporting Nebius’s strong chart momentum. This investment enhances Nvidia’s ecosystem of AI partners.
4. Competitive Dynamics with Meta
Meta Technologies’ introduction of custom AI chips for data centers adds competition to the high-performance compute space. Nvidia continues to expand its AI hardware build-out, positioning itself against emerging in-house solutions from major cloud and tech players.