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 22·5 min read
OpenAI and NVIDIA announce strategic partnership to deploy 10 gigawatts of NVIDIA systems
TL;DR
Large language model training requires massive computational infrastructure, but deployment capacity has lagged demand. OpenAI and NVIDIA announced a strategic partnership to build and deploy 10 gigawatts of AI datacenters—facilities that process and store data using NVIDIA's GPU systems—with initial deployment beginning in 2026.
✦ Why It Matters
Engineers can plan AI infrastructure investments knowing major players are securing multi-year GPU capacity commitments starting 2026.
Key Takeaways
- Plan GPU procurement strategies around 2026 datacenter availability windows.
- Expect sustained demand for NVIDIA hardware in enterprise AI deployments.
- Consider multi-year infrastructure commitments for production LLM workloads.
- Monitor partnership announcements as signals of compute capacity constraints.
Full Summary
Training and running large language models (LLMs)—AI systems with billions of parameters that generate human-like text—demands enormous computational power measured in gigawatts (units of electrical power equal to one billion watts). OpenAI and NVIDIA formalized a strategic partnership to construct and operate 10 gigawatts of AI datacenters equipped with NVIDIA's GPU systems, which are specialized processors optimized for parallel computation.
The first phase of deployment is scheduled for 2026. This partnership addresses a critical infrastructure gap: the shortage of high-performance computing capacity needed to train frontier AI models and serve them at scale.
By securing dedicated NVIDIA hardware infrastructure, OpenAI gains guaranteed access to the computational resources required for continued model development and commercial deployment. The 10-gigawatt commitment represents a substantial capital investment and signals confidence in sustained demand for AI compute resources over multiple years.