We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
OpenAI Blog·Sep 23·5 min read
OpenAI, Oracle, and SoftBank expand Stargate with five new AI datacenter sites
TL;DR
Large-scale AI model training requires massive computational infrastructure, but U.S. capacity was fragmented and insufficient. OpenAI, Oracle, and SoftBank announced five new Stargate AI datacenter sites as part of a $500 billion, 10-gigawatt infrastructure expansion across the United States.
✦ Why It Matters
Engineers can plan larger-scale AI experiments knowing U.S. compute capacity will expand significantly, reducing infrastructure bottlenecks.
Key Takeaways
- Five new Stargate datacenters will increase available GPU capacity for model training nationwide.
- 10-gigawatt power infrastructure expansion removes energy constraints on large-scale AI workloads.
- Coordinated multi-site deployment reduces single-point-of-failure risk compared to centralized facilities.
- Job creation in infrastructure signals sustained long-term investment in AI compute ecosystem.
Full Summary
Training state-of-the-art AI models demands enormous computational resources—GPUs (graphics processing units) and specialized hardware running continuously. OpenAI, Oracle, and SoftBank jointly announced expansion of the Stargate project, a $500 billion U.S. infrastructure initiative targeting 10 gigawatts of total power capacity across multiple new datacenter sites.
The five new locations complement existing facilities and represent a coordinated approach to distributed datacenter deployment rather than single-site concentration. This expansion directly addresses the bottleneck of GPU availability and power infrastructure that has constrained AI model development.
The initiative is expected to generate tens of thousands of jobs in construction, operations, and engineering. For AI researchers and engineers, this means increased access to training infrastructure and reduced competition for compute resources, enabling faster iteration on large-scale models.