Morgan Stanley reports that Nvidia and Broadcom remain resilient despite a projected 32 GW power shortfall in US data centers through 2028. While AI leaders maintain stable growth, smaller secondary component suppliers face higher risks of order delays if project timelines stall due to power constraints.
The race to build large-scale artificial intelligence infrastructure in the United States is running into a physical challenge: a lack of sufficient electricity. According to recent findings from Morgan Stanley, data-center developers face a projected net power shortfall of 32 gigawatts through 2028, representing a 34% gap in required supply. This growing electricity bottleneck is changing how the industry looks at growth, shifting the focus toward power availability rather than just chip demand.
Despite this challenging utility environment, industry leaders Nvidia and Broadcom appear well-positioned to maintain their long-term growth. Analysts note that these companies possess deep visibility into global chip deployment and maintain close coordination with both semiconductor fabricators and power infrastructure providers. This strategic advantage helps them navigate potential bottlenecks more effectively than others, leading to stable performance forecasts for 2027.
However, the story is different for smaller, secondary entities in the semiconductor supply chain. The report highlights that manufacturers of memory, optical components, power-management hardware, and analog chips are more exposed to risks. If data-center operators cannot secure enough power to bring facilities online, project schedules will likely slip. For these secondary suppliers, delays in data-center completion can lead to order cancellations or reduced demand for their products, creating a direct link between grid capacity and their bottom lines.
The power shortage is no longer just a utility issue; it has become a central factor for the semiconductor sector. As the industry faces regulatory friction and labor shortages alongside these grid constraints, the pace of capital spending on AI is seeing new friction. Investors monitoring this sector may want to look beyond just chip sales figures. The key monitorables for the coming quarters will be the speed at which data-center projects are commissioned, the effectiveness of on-site power generation solutions, and whether infrastructure growth can keep pace with the demand for advanced AI chips.