Morgan Stanley has flagged a power crunch tied to AI infrastructure buildouts as the newest pressure point in the semiconductor supply chain — with Nvidia and Broadcom positioned as the most shielded from the fallout.

The investment bank's analysts identify the constraint as a capacity problem rather than a demand problem. Data center operators racing to deploy AI systems are consuming power-related components faster than suppliers can scale output, and that bottleneck now propagates through the broader chip supply chain.

Nvidia and Broadcom stand apart, according to Morgan Stanley, because their positions in AI compute and networking place them upstream of the most exposed links. Their exposure to the power bottleneck is limited relative to suppliers whose products depend directly on constrained capacity.

What does the power crunch actually hit?

The shortage targets the power delivery layer of AI data centers: the circuitry and components that convert and manage electricity feeding high-density GPU clusters. As AI accelerator deployments accelerate, demand for this layer grows in step, and suppliers in that segment face the tightest conditions.

Morgan Stanley's assessment implies a split market:

• Shielded: Nvidia and Broadcom, whose AI compute and networking franchises sit on the demand side of the constraint.
• Exposed: component and power-chain suppliers whose output ceilings now cap how quickly AI infrastructure can scale.

Why does it matter for the market?

The finding reframes the AI supply chain debate. Investor attention has concentrated on GPU supply and advanced packaging capacity. Morgan Stanley shifts the focus toward power infrastructure as the gating factor — a constraint that could throttle deployment timelines even where chips themselves are available.

For Nvidia, the analysis adds a second layer of protection to an already dominant position: the company benefits from AI demand while sitting outside the power bottleneck. Broadcom, whose custom AI silicon and networking business serves hyperscale data centers, carries a similar profile.

Suppliers deeper in the power chain face the inverse situation. Their order books may be full, but capacity limits cap revenue growth and raise the risk that AI projects slip schedules while waiting on power delivery hardware.

What signals should readers watch?

Morgan Stanley's framing suggests three indicators will determine whether the crunch eases or tightens:

• Expansion timelines from power-component suppliers serving data center customers.
• Lead times quoted for AI data center power infrastructure relative to GPU delivery schedules.
• Any evidence of AI deployment delays attributable to power availability rather than compute supply.

The analysis positions the power crunch as the next structural test for the AI hardware cycle. Nvidia and Broadcom, per Morgan Stanley, enter that test with the strongest cover.