Morgan Stanley has identified power availability as the next constraint on the AI chip supply chain, and the investment bank names Nvidia and Broadcom as the suppliers best shielded from the bottleneck, according to a report carried by AOL.com.

The assessment shifts attention away from leading-edge fabrication capacity, which dominated supply-chain concerns during earlier phases of the AI buildout, and toward the electricity and power-delivery infrastructure required to run and cool large GPU clusters.

What does the power crunch mean for chip supply?

Data centers built for large-language-model training and inference draw power at a scale that existing grids and on-site power systems were not designed to deliver. When the constraint moves from wafer supply to electricity, the exposure of individual chip suppliers depends on where their products sit in the system.

The report's core distinction, as summarized by AOL.com, separates suppliers whose products are directly tied to power delivery from those whose exposure is indirect. According to Morgan Stanley, Nvidia and Broadcom fall into the second group.

Why are Nvidia and Broadcom shielded?

Morgan Stanley's designation of the two companies as "shielded" rests on their position in the AI hardware stack rather than on immunity to all supply-chain risk.

Both companies design chips but do not manufacture them. Their exposure to a power crunch comes primarily through the availability and operating conditions of the systems their customers deploy, not through a direct dependency on power-component supply for their own bill of materials.

The bank's framing implies a hierarchy of risk:

• Suppliers of power-conversion, delivery and thermal-management components face the most direct pressure.

• System integrators and data-center operators face scheduling and deployment risk as grid capacity lags demand.

• Chip designers such as Nvidia and Broadcom face second-order effects, mainly through the pace at which end customers can energize new installations.

What does this change for the market?

For investors and procurement teams, the report reframes how to evaluate AI-related semiconductor names. The question is no longer only who can secure advanced silicon, but who can operate within the power envelope that data centers can actually draw.

Companies judged "shielded" by the bank retain revenue visibility even if power constraints slow the broader buildout, because their order books depend on committed AI infrastructure spending rather than on per-site electrical headroom.

Morgan Stanley's analysis, as reported by AOL.com, arrives as utilities and data-center operators across major AI markets confront multi-year timelines for grid interconnections and generation capacity. Those timelines now set the effective pace of AI hardware deployment, and by extension the demand profile across the chip supply chain.

The report does not remove risk from Nvidia and Broadcom entirely. Slower energization of new sites can defer revenue recognition and push out upgrade cycles. But relative to suppliers whose products are themselves the bottleneck, the bank's designation places the two designers on the favorable side of the power equation.

For the semiconductor trade, the takeaway is structural: power has joined fab capacity and advanced packaging as a first-order constraint on AI scaling, and supplier positioning against that constraint is now a screening criterion in its own right.