Key Takeaways
- Morgan Stanley sees Nvidia and Broadcom as relatively insulated from current US data-center power constraints.
- Power shortages could delay AI infrastructure deployments and create uncertainty for secondary chip suppliers.
- Data-center expansion is increasingly dependent on reliable electricity, grid capacity and alternative power solutions.
AI’s Power Problem Is Becoming a Chip Supply-Chain Problem
The artificial intelligence boom is running into a very physical limitation: electricity. As US data centers race to add computing capacity, growing power constraints are creating a new risk for the semiconductor industry, although Morgan Stanley sees Nvidia and Broadcom as better positioned than many suppliers.
Why Nvidia and Broadcom Look Better Protected
Morgan Stanley’s research indicates that power availability remains one of the biggest bottlenecks for US data-center expansion. The investment bank has previously warned that limited grid capacity, permitting delays and infrastructure constraints could slow new projects.
Despite those challenges, Nvidia and Broadcom are considered relatively insulated because of their strong visibility into customer deployments, chip placement and broader data-center expansion plans. Their positions across the AI infrastructure ecosystem could give them greater visibility as customers adjust deployment schedules.
That does not mean the wider semiconductor industry is equally protected.
Secondary Chip Suppliers Face Greater Risk
If a data center cannot secure enough electricity to bring new computing capacity online, customers could delay installations or adjust orders. That creates potential inventory and demand risks for suppliers further down the semiconductor chain.
Memory, optical components, power-management products and analog chips could be particularly exposed if AI infrastructure projects are postponed.
Morgan Stanley has also highlighted the growing importance of alternative power solutions. Its research points to increased interest in natural gas generation, nuclear power, batteries, fuel cells and other technologies capable of supporting large data-center loads when traditional grid connections are constrained.
TwikUp’s Perspective
The AI investment story is increasingly becoming an infrastructure story. GPUs may be the visible engine of artificial intelligence, but they cannot operate at scale without electricity, cooling, transmission capacity and physical data-center space.
That could shift investor attention beyond chip designers toward companies supplying power equipment, generation and grid infrastructure. For the semiconductor industry, meanwhile, the key question may increasingly be whether customers can actually power the computing capacity they have ordered.