OpenAI is scaling its global compute infrastructure toward 30GW by 2030, underscoring the rapid escalation in power and data center capacity required to support advanced AI systems. A new company graphic shows available compute growing from roughly 0.2GW in 2023 to an estimated 1.9GW in 2025, with a stated January 2025 commitment of 10GW and more than 8GW already identified. The trajectory implies sustained multi-year investment in hyperscale AI infrastructure, with compute capacity expanding at approximately 3x year-over-year.
The roadmap reflects a shift toward “compute as core infrastructure” for AI development, where access to power, cooling, and high-performance interconnects becomes a primary constraint. OpenAI’s long-term planning suggests a move beyond incremental scaling toward coordinated, multi-region deployments capable of supporting globally distributed AI systems and real-time inference at scale.
- Targets 30GW of AI compute capacity by 2030
- ~10GW committed as of January 2025, with 8+GW already identified
- Estimated growth from ~0.2GW (2023) to ~1.9GW (2025)
- Approximately 3x year-over-year expansion in available compute
- Focus on supporting large-scale training, inference, and agentic AI workloads
- Increased reliance on hyperscale data centers and ecosystem partnerships
- Infrastructure constraints shifting toward power, cooling, and network throughput
