The US is on track to add 100+ GW of new data center capacity by 2030, driven by the OpenAI Stargate JV ($500B), hyperscaler AI arms race, and nuclear revival. The critical bottleneck is power grid interconnection — PJM alone has a 5-year queue with 3,000+ GW of pending requests. New builds are increasingly moving to the South and Midwest to escape Northern Virginia's power constraints. The US maintains a decisive lead in AI chip performance (NVIDIA Blackwell architecture) and private capital depth.
China is executing the most ambitious state-coordinated infrastructure buildout in history. The EDWC initiative is creating a national computing power network spanning 8 hub clusters. The key constraint is AI chip performance — US export controls mean China cannot access NVIDIA's latest GPUs, forcing reliance on Huawei Ascend (estimated 60–70% of H100 performance). However, China's build speed, government coordination, and domestic market scale mean it could reach 60–70% of US capacity by 2030, even with the chip gap. The DeepSeek R1 breakthrough showed China can achieve frontier AI results with fewer, less powerful chips.