U.S. data centers could face a 32GW power gap by 2028, with supply chains for ASICs, storage, and optical modules under pressure.

Morgan Stanley’s latest report projects U.S. data centers will face a roughly 34% net power gap between 2026 and 2028, equivalent to 32 gigawatts (GW). The power bottleneck is expected to shift from “chip shortages” to “power shortages,” initially hitting downstream segments of the supply chain including Application-Specific Integrated Circuits (ASICs), storage, optical modules, power management ICs (PMICs) and analog chips. The report notes that NVIDIA and Broadcom, due to their high visibility into the final deployment locations of their chips and global footprints that reduce reliance on the U.S. single market, will not face material impacts to their 2027 earnings forecasts for now. In contrast, ASICs are more sensitive to power resources, while products like storage, optics and power management components are more vulnerable to delays in data center projects, facing risks of order postponements, cancellations and inventory adjustments. Morgan Stanley points out that Oracle’s 1.3GW Project Lighthouse in Wisconsin has become a real-world example of this risk: the project’s full-power supply could be delayed as early as October 2028, and in a pessimistic scenario, it could even be pushed to spring 2029. Even if AI servers are already racked, they cannot be converted into actual computing power and revenue without sufficient power. The report forecasts that under the baseline scenario, the U.S. will add about 19GW of gas turbine and engine capacity between 2026 and 2028, Bloom Energy’s fuel cells will contribute around 6GW, and nuclear power and crypto mining retrofits can also provide supplementary power, but the total is still insufficient to fill the power gap. Meanwhile, overseas expansion is constrained by factors such as approvals and geopolitics, making it difficult to shift demand quickly.