Why Morgan Stanley Restored Nvidia as Its Top Semiconductor Pick

Morgan Stanley restored Nvidia as its top semiconductor pick on October 2, 2026, while keeping its Overweight rating and $300 price target. Analyst Joseph Moore’s case combined a valuation he considered undemanding with expectations for continued AI infrastructure demand—even as constraints shift from chip supply toward the power, land, construction and financing needed to bring data centers online.91011

Why Morgan Stanley changed its ranking

Moore’s view followed investor meetings with Nvidia CEO Jensen Huang and CFO Colette Kress. The discussions reinforced the bank’s assessment that Nvidia’s customer base extends beyond the biggest cloud providers and frontier AI labs, and that the company is still early in a new product cycle.1114

Valuation was another part of the argument. Reporting on Moore’s note put Nvidia at about 15 times Morgan Stanley’s fiscal 2028 earnings estimate. That multiple reflects the bank’s forward earnings forecast; it is not a guarantee that those earnings will materialize or that the shares will reach the $300 target.9

Data-center power constraints make computing per gigawatt matter

Morgan Stanley’s thesis is that AI deployment is increasingly limited not just by chip availability but by access to electricity, land, data-center capacity and financing. When customers cannot quickly add power or buildings, the amount of computing delivered from each available gigawatt becomes more important.1014

The reporting describes Nvidia’s strategy as increasing the number of AI tokens its systems can produce per gigawatt. That makes compute efficiency relevant to customers facing deployment limits. The sources reviewed describe this as a broader strategy; they do not provide a quantified Feynman-specific revenue forecast or show that such a forecast is part of the $300 target calculation.14

Demand may reach beyond hyperscalers—and include CPUs as well as GPUs

Morgan Stanley highlighted demand from a broader range of customers alongside traditional hyperscalers and frontier AI labs.211 It also sees the growth of AI agents as a potential driver of demand for both CPUs and GPUs, rather than evidence that GPU demand is fading.211

That distinction matters when interpreting market interest in CPU-focused companies: demand for CPUs can rise alongside demand for GPUs as AI systems and server configurations evolve. It does not, by itself, establish how much revenue Nvidia will earn.11

How the buyback and Barclays forecast fit in

Nvidia’s additional $150 billion share-repurchase authorization is a separate shareholder-return development that formed part of the market backdrop. An authorization gives the company the ability to repurchase shares; it is not the same as having already spent the full amount.14

Barclays offered a separate check on the demand outlook. Using Nvidia’s disclosures about spending by the five largest cloud providers, Barclays estimated at least $30 billion of potential revenue upside versus its own forecasts in each of 2026 and 2027.18 That estimate is an analyst projection—not reported revenue—and it is not the calculation behind Morgan Stanley’s $300 target.

What the $300 target does—and does not—say

Taken together, Morgan Stanley’s argument is that Nvidia’s valuation looked attractive relative to its earnings expectations, while a wider customer base and the need to get more computing from constrained data-center power could sustain demand. The meetings with Nvidia executives informed the bank’s view, but do not remove the execution risks involved in building and financing infrastructure.9101114

The $300 target remains an analyst forecast, not a promised share price. It depends on customer spending continuing and new capacity being deployed; the per-gigawatt strategy and Barclays’ revenue estimate should likewise be treated as expectations rather than realized results.918