US data-center developers face a worsening electricity shortage over the next several years, risking supply chain delays for key semiconductor segments amid growing AI demand, Morgan Stanley said. However, chip giants Nvidia and Broadcom are largely shielded from these power constraints due to their strategic positioning and supply chain coordination.
- US data centers face a 34% power shortfall by 2028, equal to 32 GW
- Nvidia and Broadcom insulated via strong chip placement and coordination
- Memory, optics, and power-management chip makers bear biggest risks
What happened
Morgan Stanley has highlighted a significant power shortage looming for US data centers due to the soaring electricity requirements of generative AI workloads. Their latest estimate points to a 34% net power deficit through 2028, even after accounting for supplementary power sources such as behind-the-meter generation and fuel cells. This shortage translates to approximately 32 gigawatts less power availability than needed.
This energy crunch is beginning to impact the semiconductor supply chain by introducing potential delays in chip deployment. While investment in data center infrastructure continues aggressively to meet AI demand, challenges including labor shortages, political hurdles, and power availability threaten to slow the pace of new capacity coming online.
Why it matters
The power shortfall is particularly critical for the chip supply ecosystem supporting AI infrastructure. Morgan Stanley notes that Nvidia and Broadcom remain well insulated from these disruptions, thanks to their clear visibility over chip placement strategies, geographic expansion plans, and strong coordination across data center operators, semiconductor suppliers, and power providers. This protection helps preserve their 2027 revenue and shipment forecasts.
In contrast, other chip categories—specifically memory, optical components, power management, and analog devices—are more vulnerable to inventory disruptions or deferred orders if power constraints cause customers to delay data center expansions. Such interruptions could ripple through the supply chain, affecting overall AI hardware availability and timelines.
What to watch next
Monitoring how US data center operators respond to this power bottleneck will be crucial. Potential solutions might include accelerated investment in alternative power generation, improved energy efficiency, or political developments that either facilitate or restrict new infrastructure. The pace and scale of these responses will directly influence semiconductor demand and supply chain stability.
Investors and industry observers should also focus on secondary chip manufacturers vulnerable to power-related delays. Any significant order cancellations or supply chain bottlenecks in memory, optics, or power management could signal broader challenges for the AI hardware ecosystem, potentially affecting pricing, availability, and innovation timelines in the coming years.