NVIDIA Corporation (NVDA.US) reports tonight: Morgan Stanley says the business remains strong, but what the market is really nervous about is the interim CEO's debut and the guidance outlook.

date
09:53 17/11/2026
avatar
GMT Eight
NVIDIA is about to report earnings; Morgan Stanley maintains an "Overweight" rating with a price target of US$267.
NVIDIA Corporation (NVDA.US) will report results for the third quarter of fiscal 2027 on Wednesday morning Beijing time. Morgan Stanley's take in its earnings preview is direct: the focus of this report is not the current quarter's numbersthe results themselves are already fully priced inbut the first management appearance after Jensen Huang's departure, and guidance for next year that the company "may not even give." Morgan Stanley maintains its "Overweight" rating and a US$267 price target. The quarter's numbers: the part without suspense Morgan Stanley expects F3Q27 revenue of US$108.1 billion, slightly below the US$108.9 billion consensus; EPS of US$2.47, in line with consensus. Analysts wrote in the report that channel checks show "the business remains strong," and even as some Southeast Asian projects slow, "they cannot meet demand." What is really worth hearing on the call are three things from interim CEO Colette Kress: her stance on support for and investment in the AI ecosystem, whether NVIDIA Corporation's innovation engine can continue, and her assessment of the risk of further tightening of export controls. Jensen Huang typically did not give opening remarks but was extremely active in the Q&AMorgan Stanley notes this part, "broad enthusiasm about AI's transformative impact," may diminish at the margin. The report also points readers to its November 13 special report "NVIDIA: Which Way Forward," arguing the management transition itself deserves more discussion than any single quarter's profit and loss. The divergence beyond this quarter: Morgan Stanley is about US$70 billion below consensus Over a longer horizon, the divergence becomes immediately visible. Morgan Stanley expects fiscal 2028 revenue of US$623 billion and EPS of US$13.37, versus consensus of US$694.8 billion and US$15.75a gap of about US$100 billion that is the result of "last week's cut to numbers": Morgan Stanley's estimate for next year's revenue growth is 52%, versus company guidance of about 70% and customer demand growth of 100%. Still, analysts believe this number "has been fully de-risked," and channel checks still point to continued optimism; the real suspense is that, with an interim CEO in place and questions mounting, they "are not sure whether the company will give guidance next year." Focus one: demandthe gap between 100% and 70% NVIDIA Corporation previously guided fiscal 2028 revenue growth of about 70%, based on customer demand growth of nearly 100%. But after the midterm elections, concerns about land, power and data center (LPS) constraints have risen markedly: SoftBank, a partner on NVIDIA Corporation's PORTS-Pike project, has written down US$3.7 billion on related power equipment. Morgan Stanley's judgment is that the midterm elections pose a greater threat to 2028 and 2029 than to 2027data centers under construction will not be affected by new bans, and the ecosystem has time to route around bottlenecks; the hyperscalers it spoke with acknowledge challenges but are "confident these challenges can be overcome." Financing is the other side of the same issue. Jensen Huang historically supported regional neoclouds and used NVIDIA Corporation's balance sheet to help customers lock up land and power resources; whether these commitments continue after the management change will be a focus. Morgan Stanley cites an example: CoreWeave, Nebius and IREN together guide to about US$100 billion of capex in 2027, but their combined market value has evaporated by US$38 billion over the past monthif the phrase "demand growth of 100%" partly depends on financing, then whether financing still comes, especially for long-tail small customers in the ACIE segment, needs continued monitoring. Morgan Stanley's fallback is that demand from hyperscalers alone is enough to support its forecast, with next year's shipments equivalent to roughly "high teens GW," and SpaceX alone planning to deploy 10GW. The report also deliberately flags a more fundamental variable: investors' willingness to put money into AI is "a bigger risk to the buildout trajectory than anything NVIDIA Corporation can control"in other words, what really matters tonight is whether changes in the financing environment make it into management's language. Focus two: supply and guidancereiteration is a positive Morgan Stanley does not expect the full-year growth guidance to change, but out of caution it has cut its own revenue forecast by about 10% on China export control risk. On the supply side, especially memory, conditions "remain very tight," with no sign of easing"if anything, what we are seeing is tightening." One telling detail: some investors question why NVIDIA Corporation has been slow to reiterate guidance after the CEO change. Morgan Stanley reads this instead as a positive signalby convention earnings often come with a raise, and pre-announcing may be "better left for the earnings day itself." Focus three: Rubin rampfirst quarter of high-volume shipments This is Rubin's first quarter of high-volume shipments; a quarter ago the expectation was that Rubin would account for about 20% of data center revenue (about US$20 billion). The ramp has not been smooth: key test partner King Yuan Electronics has cut its 2026 guidance for a second consecutive quarter, partly attributing it to a slower-than-expected Rubin ramp. But Morgan Stanley does not think this will hit NVIDIA Corporation's revenue recognition: first, shorter test times can explain part of the gap; second, its estimate of flat quarter-on-quarter growth in the January quarter "should still be conservative," with the bigger Rubin ramp-up coming in the January and April quarters. The latest testimony on the demand side In an appendix, Morgan Stanley compiles what cloud companies said about AI capex on their Q3 earnings calls, which can serve as background for tonight's call: Amazon.com, Inc. said it "installs an AI accelerator roughly every 14 seconds," with payback periods still shorteningits 2026 capex is about US$220 billion, 2027 will be significantly more, and it has "clear visibility into future returns"; Alphabet Inc. Class C said AI investment "drove the highest query growth in five years" and believes every dollar of capex "can support more revenue over time"; Oracle Corporation said "demand for AI is real and enormous; building capacity and financing are just taskstasks may be hard, but they can all be solved"; Meta raised the low end of its capex range from US$35 billion to US$37 billion; Microsoft Corporation said "I would be surprised if we have less than 13GW of compute coming online before the end of next year"; SpaceX said there is a "non-zero possibility" of bringing more than 20GW of in-house compute online next year, and plans to deploy 440,000 Vera Rubin units by the end of the second quarter to build the strongest training cluster; on the neocloud side, Nebius said "customers are committing to us for longer terms, at higher prices, with more prepayments," while CoreWeave said existing capacity "could be sold out even to buyers in line." Valuation: three scenarios and option-implied probabilities Morgan Stanley's three scenarios: bull case US$336 (about 21x 2027 EPS of US$16), triggered by supply easing driving data center revenue above expectations, a revenue-sharing agreement bringing billions of dollars of pure-profit revenue, and GPU-based AI PCs opening the consumer market; base case US$267 (about 20x US$13.37, a discount to the overall semiconductor sector and a slight premium to the broad marketMorgan Stanley explains that high market share and high gross margins actually leave the valuation multiple "lacking near-term expansion leverage"); bear case US$130 (about 10x), corresponding to two big debates both turning negative: supply catching up to demand faster than expected, or competitors entering and customers' in-house chips taking share. Based on option-implied volatility, the probabilities over one year of breaking above the bull price, rising above the base price and falling below the bear price are about 0.9%, 11.2% and 24.2% respectivelythe options market is clearly pricing downside scenarios more heavily than upside. Sell-side consensus distribution: 77% Overweight, 22% Neutral, 2% Underweight. In Morgan Stanley's risk-reward theme scores, the new data era, pricing power and long-term growth are all positive.