Nvidia (NVDA) CFO Colette Kress told investors last week that the company is seeing “extreme pricing conditions in memory,” that the increases have “exceeded our prior expectations,” and that they are “headed even higher into next year.” Colette Kress said this in prepared remarks on Nvidia’s fiscal second-quarter earnings call on Aug. 26, and then delivered the line companies almost never say plainly: “As a result, we are resetting expectations today.”
Memory is not an exotic part. It is the DRAM in a phone, the RAM in a laptop, the storage in a game console, and one of the most expensive ingredients in a graphics card… and three companies make nearly all of it. Kress was describing the market for AI server memory, not the shelf price of a laptop, and Nvidia said nothing at all about consumer devices. But those products draw on the same suppliers and the same fabs, which is why a chip company’s margin guidance is worth reading even if you never buy a chip.
More News from Barchart
The guidance is where the admission gets specific. Gross margin was 75% in the quarter that was just reported. Kress guided to 74% in the current quarter and told analysts that Nvidia expects “margins to bottom in Q4 in the 71% to 72% range before settling at 72% to 73% in [fiscal 2028] as executed price increases take effect in Q1.” Executed, not contemplated. Nvidia is absorbing part of the memory bill and passing part of it forward, and it has already decided how much.
Sponsored Content: Invest in the Platform Giving Everyday Investors Access to Private Markets
The company also volunteered the cause, which is the unusual part. “Memory scarcity today is being driven in large part by the AI buildout itself,” Kress said, and “unlike a component that simply raises our cost with no offset benefit, tighter memory supply is a symptom of the same demand surge that’s driving our own growth.” Nvidia is not presenting itself as a bystander to the shortage.
It is also the largest single buyer inside it. Nvidia’s supplier purchase commitments stood at roughly $279 billion in its most recent quarterly filing — capacity reserved years ahead of when the parts are needed. Tom’s Hardware reports that about $160 billion of that relates to memory specifically. That split is not obviously a disclosed line item in the filing, and it should be read as that outlet’s reporting rather than as a company figure.
Market structure is why any of this bites. “We have longstanding deep relationships with all three major memory suppliers, and we’re working closely with them to further increase the capacity our roadmap requires,” Kress said. She did not name the three companies, but they are Samsung, SK Hynix (SKHY) and Micron (MU), and there is no fourth option. Adding capacity means building fabs, which takes years, not quarters.
What has already reached buyers is harder to pin down than the corporate numbers. Tom’s Hardware and Network World both report, sourced to supply-chain channels rather than to Nvidia, that the company has warned large customers of server price increases above 15%, on top of an earlier rise in July. Kress’ “executed price increases” is the on-the-record version of the same thing. What Nvidia has not said is anything about what a graphics card or a laptop will cost, and that link — real as the shared supply base is — remains an inference rather than a company statement.
The other side of the trade is simple arithmetic. A memory shortage is a cost to whoever buys memory and revenue to whoever sells it, and the same cycle compressing Nvidia’s margin runs in the suppliers’ favor. Barchart flagged rising memory prices as a risk to Nvidia’s numbers before the company attached a figure to it.
There is a contrast inside Nvidia’s own commentary worth sitting with. CEO Jensen Huang said there is no AI bubble because AI is “incredibly profitable”, and the CFO has now guided margins down over four quarters because of an input cost the company says its own buildout is largely creating. Both statements can hold at once. They are the two halves of the same trade.
The things to watch from here are narrow and checkable — whether gross margin actually bottoms in the 71% to 72% band Nvidia guided to, and whether memory contract prices roll over before fiscal 2028, which is when Kress said supply stops being the constraint. Until one of those things resolves, the company’s own guidance is the clearest available read on where the price of memory goes next.
On the date of publication, Caleb Naysmith did not have (either directly or indirectly) positions in any of the securities mentioned in this article. All information and data in this article is solely for informational purposes. This article was originally published on Barchart.com