Morgan Stanley Puts Nvidia Back at the Top of Its Chip List, Calling 15 Times Earnings 'Very Undemanding'

Morgan Stanley has reinstated Nvidia as its top pick in semiconductors, telling clients in a Friday note that the chipmaker is well positioned "with a very undemanding valuation." Analyst Joseph Moore wrote the note after investor meetings with CEO Jensen Huang, CFO Colette Kress and investor relations head Toshiya Hari in New York and Boston. Nvidia shares climbed about 1.2% in pre-market trading, moving back toward their 52-week high of $236.54.

The valuation math is the anchor. The stock trades at roughly 15 times Morgan Stanley's fiscal 2028 earnings estimate, and Moore argued the multiple does not need to expand for the shares to work - estimates simply need to keep rising. He did see scope for the multiple to expand if enthusiasm returns to AI, but framed that as upside rather than a requirement, a notably patient framing for a megacap that has spent two years trading on sentiment cycles.

The more interesting argument is about where the bottleneck has moved. Morgan Stanley says the constraint on AI growth is shifting away from chip production toward how fast new data-center space can be built and financed - power, land, capital - and that this shift plays directly to Nvidia's strengths. The company lists 80 cloud partners, 55 of them outside the United States, giving it flexibility to deploy capacity wherever power is actually available rather than where demand sits.

Moore also pushed back on the idea that Nvidia's business is only the half investors talk about. "While 90% of investor conversations are around the biggest hyperscalers and frontier model companies, half of the business - probably the faster growing half - is a broader range of AI model companies, Neoclouds, sovereign entities, ODMs and OEMs, and enterprise customers," he wrote. A wider distribution of AI spending, in other words, feeds a wider distribution of Nvidia's customer base.

The note also flags Nvidia's push to raise revenue per gigawatt of data-center capacity as a structural lever: the Feynman architecture expected in 2028 is projected to lift that figure from around $40 billion to well above $50 billion. On AI agents - the current favorite demand narrative - Moore expects them to increase CPU demand, but drive far more spending on GPUs, since agentic workloads run long and consume compute continuously.

Street positioning around the stock remains close to uniformly bullish: TipRanks counts a Strong Buy consensus across 31 analysts, all rating the stock a Buy, with an average price target of about $324 and Morgan Stanley's own target at $300. Analyst calls are opinions rather than guarantees, and this is the same bank that has cycled the top-pick designation between Nvidia and memory makers before. But the substance of the note - a bottleneck argument about power and financing rather than chips, and a revenue base broader than the headline hyperscaler deals - is a more specific thesis than a simple momentum endorsement.