• Morgan Stanley (MS) has reinstated Nvidia (NVDA) as its top semiconductor pick, arguing that the AI infrastructure buildout remains structurally strong but is increasingly constrained by data-center power, construction, and financing rather than GPU supply.
  • Nvidia trades at roughly 15x Morgan Stanley’s FY2028 earnings estimate, a valuation that hinges on extraordinary multiyear growth, while the stock’s trailing P/E sits near 29x.
  • The bank highlights Nvidia’s broad global customer base and geographic flexibility as key advantages as AI projects are routed toward regions with available power and deployable data-center capacity.

A Structural Shift in AI’s Limiting Factors

Nvidia has reclaimed its spot as Morgan Stanley’s highest-conviction semiconductor pick, with the bank telling clients that the AI buildout’s bottleneck is moving from chip supply toward data-center power, construction, and financing. According to people familiar with the matter, the reinstatement reflects a view that Nvidia is unusually well positioned to navigate this next phase, as its diverse customer base—spanning hyperscalers, sovereign initiatives, enterprises, and neocloud providers—gives it flexibility to direct systems toward regions where electricity and facilities are available first.

The call comes as Nvidia’s latest reported results underscore robust demand. Fiscal Q2 2027 revenue reached $96.2 billion, up 106% year over year, with $89.0 billion from Data Center—roughly 93% of total company revenue. Management guided fiscal Q3 revenue to about $108 billion, implying another large sequential step-up. “The operational story is a product-and-platform transition,” noted one analyst, pointing to Blackwell Ultra driving current deployments while the Vera Rubin platform moves into full production at customers including CoreWeave (CRWV), Google (GOOG) Cloud, Microsoft (MSFT) Azure, Oracle (ORCL) Cloud Infrastructure, and Nebius (NBIS).

Valuation Debate: Forward vs. Trailing Multiples

Morgan Stanley’s cited valuation of about 15x its FY2028 earnings estimate should be viewed as an analyst-specific forward multiple, not the market’s conventional trailing P/E. At the latest available snapshot, NVDA traded at $230.86 in pre-market activity on October 2, up 1.09% in the prior session, with a $5.59 trillion market capitalization. Its quoted trailing P/E was 29.19.

The gap between the two figures illustrates just how much projected earnings growth is baked into the bull case. Consensus estimates call for FY2028 revenue of about $691.9 billion and diluted EPS of $15.75. At the latest price, that equates to an approximate 14.7x FY2028 EPS multiple—close to the 15x characterization. But those estimates are inherently vulnerable to execution, demand, supply-chain, and policy changes.

Meanwhile, the broader analyst backdrop remains overwhelmingly bullish: 23 of 23 consensus ratings are bullish, with an average target of $345.35 and a median target of $330. The target range spans $282 to $515. Those figures describe sentiment, not a guarantee of returns.

Power, Construction, and the Financing Hurdle

The economic insight behind Morgan Stanley’s call is that AI infrastructure is becoming a physical-infrastructure problem. Training and inference clusters consume enormous electricity, making access to generation, transmission, substations, transformers, and grid interconnection critical to deployment timing. Land, permits, cooling, networking, and construction capacity can also constrain projects even when customers can procure GPUs.

Financing is another pressure point. Reuters reported growing Wall Street debate over Nvidia’s financing approach and whether advanced chips can reliably serve as long-term loan collateral, with lenders seeking stronger guarantees in parts of its reported $500 billion financing effort. “A pullback in lender appetite or deterioration in AI-project economics could defer orders,” one market strategist warned.

Nvidia’s breadth potentially lets it direct systems where electricity and facilities arrive first, rather than relying on a single country or client. That flexibility is a key pillar of the bull case. The trade-off, however, is that a supply bottleneck moving downstream does not eliminate risk—it changes the risk. Nvidia may be less limited by chip fabrication, but its revenue conversion may become more exposed to customer construction delays, power interconnection queues, credit conditions, and returns on enormous AI capital expenditures.

Export Controls and Geopolitical Crosscurrents

Nvidia remains deeply exposed to U.S.–China technology policy. U.S. export controls restrict direct sales of its most advanced AI processors to China. Reporting in August indicated that Chinese companies were accessing Nvidia computing capacity through overseas data centers in Southeast Asia, a route not clearly covered by rules focused on physical chip exports. Proposed U.S. legislation would extend restrictions to remote access to controlled compute.

This week’s enforcement news heightens the issue: U.S. prosecutors charged a Southern California man over an alleged scheme to route about $300 million of export-controlled AI hardware through Singapore and Malaysia to China. Prosecutors said the equipment included high-end servers containing Nvidia GPUs; Nvidia itself was not alleged to have violated export controls. Bloomberg similarly reported an expanding international crackdown into alleged diversion networks.

For Nvidia, the implications are mixed. Stronger controls can reduce access to a major end market and add compliance costs, while also potentially accelerating Chinese development of domestic alternatives. At the same time, global demand outside China and geographically diversified data-center construction can partially offset constrained direct China sales. Policy uncertainty can complicate long-range forecasts, particularly if controls tighten around cloud access, foreign subsidiaries, or third-country deployment.

The Bottom Line

Morgan Stanley’s call is a vote that Nvidia’s competitive position will let it benefit from the next phase of AI investment, even as the bottleneck shifts beyond chips. The company’s reported financial momentum—$24.1 billion in operating cash flow in fiscal Q2 after $50.3 billion in Q1—and broad analyst support substantiate the bull case. But the key risks are no longer confined to semiconductor supply. They extend to power, construction, credit, export controls, and the ultimate profitability of global AI infrastructure. As one analyst put it, the debate is shifting from “Can enough GPUs be made?” to “Who pays for, powers, and benefits from AI-scale infrastructure?”

Correction: An earlier version of this article misstated the sequential change in Nvidia’s fiscal Q2 net income. It rose 2%, not 18%.