Good Numbers, Falling Stock
NVIDIA (NVDA) reported earnings that were great, as expected, yet the stock fell about 0.8% during the coverage. For NVDA, good is not always good enough. The stock has slid in six, now seven, out of the last nine quarters after earnings. The bar sits so high that everyone else moves underneath it while NVDA stays on top. When the bar is that high, almost any result gets hammered down.
The numbers matter less than what CEO Jensen Huang says next. Guidance is the key thing to watch. The real story is the numbers plus the guidance, because NVDA must prove the AI infrastructure growth cycle still has a lot of runway left for them, rather than opening runway for competitors.
The Margin Question
NVDA is guiding gross margins in the range of 73.5% to 74.5%, below the 75% it currently reports. The market may be taking issue with that step down. Has NVDA reached peak margins, or is something else driving this?
It seems odd to break down whether a possible 50 basis point move is the reason NVDA might not climb on growth that no one else touches. The company's numbers are one of one; no competitor comes close.
NVDA also announced it is partnering with suppliers to secure critical components. Jensen Huang has been a strong diplomat here, with a good relationship with TSMC and partners across Taiwan and South Korea.
The Flywheel and Mobilizing Capital
NVDA sits in the middle of a flywheel. It benefits from the hundreds of billions invested in AI infrastructure while also working through partnerships to create more demand for it. Jensen Huang talks a lot about AI factories, which need heavy upfront investment.
A few weeks ago NVDA announced a large partnership with key institutional investors, aiming to mobilize $500 billion in third party capital. "Mobilize" is the key word: NVDA is not writing a check itself. It wants to build financial structures that pull in capital and let it be deployed.
This ties to Huang's bigger thesis: NVDA compute should be viewed as "investable assets," and AI factories as "investable infrastructure." How he extends that thesis will show whether the margin conversation is reliable now or whether it will not hold long term.
Three Things to Watch from Jensen Huang
Demand. NVDA benefits from the hundreds of billions invested, but it needs to show the money, that demand is not just big but also broadening, wide enough to carry NVDA through its product transition.
New demand vectors. A Perplexity-NVDA partnership around a portable computer suggests NVDA is not only moving toward big AI factories but also getting closer to the end user.
The product transition. NVDA must show where it stands in the shift from Blackwell to Vera Rubin. The report mentioned Vera Rubin is in full production. Rather than showing Blackwell slowing while Vera Rubin ramps up, NVDA has the chance, with market momentum and existing customers, to stack the two growth cycles instead of ending one to launch another. Stacking them is fuel for growth, and research before the numbers pointed in that direction.
NVDA is also helping existing customers get more value from infrastructure they already bought, for example with Dynamo. The result is a two-way, give-and-take relationship with customers and the partner ecosystem. The open question is how investors and the street react when they want a bit more.
Guidance and the Ports Pike Opportunity
The initial read on guidance was that it came in higher than estimates but lower than buy side consensus, again pointing to very lofty expectations.
Headlines driving NVDA lately include the Ports Pike opportunity in Ohio. Limited pre-call commentary says the infrastructure there could represent about 1.5 million GPUs, which could mean $150 billion to $200 billion in revenue. That site is expected to come online in 2028, so it is a long way out. NVDA also noted leases signed over the next 15 years or so, which it plans to lease to third parties, adding more paths to growth, though some are further down the road.
Sustainability and Customer Concentration
The main question these facts raise is how sustainable the growth is. NVDA sits at the center of the AI boom, but being that nucleus does not make it immune to the economics of the boom.
NVDA still has significant customer concentration. Last quarter, three of its biggest customers made up over half of revenue. That concentration could open the door to competition, since some of NVDA's largest hyperscale customers are also investing in their own custom silicon.
So NVDA must prove why customers should keep buying, and keep spending at the current pace. That brings in the ROI question: eventually those hundreds of billions have to turn into revenue, productivity, and real business value for customers.
