A Monday note from Morgan Stanley concluded that Nvidia and Broadcom have meaningful protection against the deteriorating U.S. data-center power situation, while warning that a slowdown in AI buildout could hit suppliers of memory, optical, and related chip components with order disruptions, according to Reuters.
In a report last month, the bank put the net power shortfall facing U.S. data-center developers at 34% through 2028 — around 32 gigawatts — even when factoring in behind-the-meter generation and fuel cells. Morgan Stanley said it does not see those bottlenecks putting Nvidia or Broadcom’s 2027 forecasts at risk, according to Reuters.
Morgan Stanley credited the two chipmakers’ ability to monitor where their products are being installed, pursue growth across multiple geographies, and work closely with data-center operators and power-sector partners as a structural cushion that smaller, lower-tier suppliers simply do not have, according to Reuters. That flexibility could allow Nvidia to redirect chips toward data-center projects that have secured power if another facility is delayed, while suppliers tied to the delayed project have fewer options, according to Benzinga.
Should deployment timelines slip, buyers may delay or walk away from orders altogether, and Morgan Stanley identified memory, optics, power-management, and analog components as the product categories carrying the heaviest inventory risk in that scenario, according to Reuters.
In a related line of argument, Morgan Stanley has contended that Nvidia’s competitive edge sharpens precisely when power is scarce, because customers under electricity pressure prioritize the most computationally efficient hardware available, according to Benzinga.
Nvidia CEO Jensen Huang has echoed that framing, noting that before any computing gear goes into a facility, developers must first line up real estate, electricity, and the building structure itself — a sequence he said routinely runs two to three years, according to Benzinga.
The power crunch is part of a broader strain on U.S. electricity infrastructure driven by gigawatt-scale AI data centers coming online this year. Hyperscale campuses running Nvidia’s AI training chips can draw 120 to 140 kilowatts per rack, and the transmission infrastructure needed to power new facilities takes five to ten years to plan and energize — far longer than the 18 to 36 months it takes to build a data center.
Goldman Sachs has also flagged mounting constraints on the U.S. data-center buildout but expects limited near-term impact from political opposition. Morgan Stanley’s analysis identifies labor, power, and political challenges as the combined pressure points, according to the outlet.
https://finance.yahoo.com/technology/articles/morgan-stanley-nvidia-broadcom-shielded-120314954.html?shem=aimgspe,
