Key developments: China’s AI compute capacity is now estimated at 24 GW and is set to add another 50 GW across more than a thousand new data centers, narrowing the gap with the United States’ 56 GW total. The surge is driven by state utilities treating AI load as a growth priority, which flips the traditional permitting‑and‑power constraints that are slowing U.S. deployments. At the same time, financing needs are exploding—Morgan Stanley projects $1.5 trillion of external capital will be required for AI infrastructure by 2028, and companies like SpaceX are seeking $40 billion to fund Nvidia AI chips through a mix of bond and loan issuances. The broader industry recognizes that high‑density GPU clusters and liquid‑cooling solutions make power capacity the primary bottleneck and a critical financial asset.
Why it matters: The convergence of massive compute expansion in China and the looming global financing gap underscores a race not just for AI models but for the underlying electricity and cooling infrastructure that powers them. Investors and policymakers must address power‑grid upgrades and cooling technologies to avoid throttling AI growth, while firms that secure financing early will shape the next generation of AI services.
What to watch next: • How U.S. regulators respond to the “power‑first” infrastructure strain and whether new permitting reforms can unblock domestic buildouts. • The rollout of large‑scale financing deals, such as SpaceX’s $40 billion bond/loan package, and their impact on GPU supply chains. • Emerging cooling innovations and grid‑integration projects that aim to mitigate the power bottleneck for high‑density AI data centers. [[1, 2, 3]]