Nvidia FQ3 26 – The Surgeon

Nvidia does enough to keep the party going

• Nvidia produced results tailored with surgical precision to do just enough to keep the AI party rolling, but not so much that expectations become overblown and everyone gets the mother of all hangovers.
• Nvidia produced another set of excellent results with revenues / EPS of $57.0bn / $1.30, comfortably ahead of expectations of $54.9bn / $1.26 and guided nicely for the coming quarter.
• In FQ4 26, revenues are expected to be $63.7bn – $66.3bn ($65.0bn), again nicely ahead of the consensus forecast of $62.3bn.
• The results were once again driven by the Data Center segment, where Compute revenues grew by 56% YoY and Networking by 162% YoY.
• Networking is a reflection of how important scale is becoming, as thousands and soon millions of chips will all be working together as one to train and run inference on massive models.
• Nvidia continues to have more demand than it knows what to do with, which, when it comes to the management of expectations, makes life fairly easy.
• As a result, Nvidia’s revenues are mostly driven by how much capacity it has reserved at TSMC, as pretty much anything that it can’t sell to one customer can be turned around and sold to someone else.
• This is how Nvidia has managed to shrug off the impact of China going to zero with barely a ripple to its financial performance.
• I continue to think that there will be no data centre shipments into China until there is a major change of heart by President Xi, regardless of what the White House does.
• However, there is no reason why gaming and automotive should not continue to ship, but these are a rounding error by comparison and so won’t draw much attention.
• The most impressive item in these numbers is gross margin, which fell by 120bp to 73.4% in FQ3 26 from 74.6% in FQ3 25.
• This tiny decline has been achieved despite the shift from selling the Hopper HGX system to the full cabinet systems, which have a lot more non-Nvidia components in them.
• Usually, one would expect to see a big decline in gross margins as the revenue mix has more resold components in it, but such is the demand for Nvidia GPUs, that Nvidia is clearly able to make very healthy gross margins by reselling the products of its partners.
• Nvidia also continued to do a best-in-industry job of delivering profits for shareholders with a net margin after tax of over 50% and very healthy cash flow.
• This is what keeps the valuation from soaring into lunatic territory as FY 2026 (Jan 26) PER is at 40.7x despite becoming the world’s most valuable company.
• I suspect that this surgical management of expectations is going to continue for a few more quarters as Nvidia remains sold out of data centre GPUs, meaning that if one asks Jensen for some chips, he might be able to oblige in about 12 months.
• This gives the company excellent visibility, as demand wobbles can simply be handed off to the next customer in the queue, meaning that supply from TSMC is the single biggest variable in the numbers.
• TSMC’s reliability is top-notch, meaning that when Jensen and Colette give their forecasts, they have a high level of visibility.
• This means that the share price and, to some degree, the sentiment in the AI industry hangs on their outlook, which I think will continue to be successfully managed with surgical precision for a few more quarters yet.
• This is particularly the case as Nvidia has stated that it thinks it can sell $500bn worth of its products between now and the end of fiscal 2027 (ends Jan 27) and even went so far as to say there could be upside to that estimate.
• Consequently, the outlook is for more-of-same for at least a few more quarters, which will provide great relief to nervous investors.
• However, the quote of the day goes to my colleague and competitor, Daniel Newman from Futurum Research, when speaking to the FT.
• Here he said, “While it’s hard to believe the demand is this stable and significant, at some point the doubters are going to have to start to believe” (see here).
• This comment is highly relevant because history has demonstrated that these doubters will be the last buyers, meaning that when they capitulate, this will be the time to get out.
• This is what I am looking for very carefully, even as Nvidia carries the industry upwards and to the right.