For most of the AI boom, Nvidia Corporation (NASDAQ:NVDA) margins moved in just one direction. That is about to change.
Soaring memory costs will push the company’s gross margins from 75% in the second quarter to 71%–72% in the fourth quarter. Nvidia expects margins to recover only after higher prices take effect in fiscal 2028.
But what caused the decline? An AI buildout so expensive that it inflates the cost of the components Nvidia needs to build its own systems.
That creates a new problem for the Santa Clara, California-based company: Demand is still overwhelming supply, but some of that scarcity is now working against it.
Memory Prices are Becoming Nvidia’s Problem
Nvidia delivered $96 billion of revenue in the second quarter, more than double the year-earlier figure. It also expects fiscal 2028 revenue to grow approximately 70%.
Yet investors are being asked to accept lower profitability while that growth accelerates.
Colette Kress, Nvidia’s chief financial officer, said the company is facing “extreme pricing conditions in memory.”
“The magnitude of the price increase has exceeded our prior expectations and are headed even higher into next year,” she added.
That is the clearest signal yet that memory has become a meaningful constraint inside the AI infrastructure boom.
Every accelerator Nvidia ships carries high-bandwidth memory it buys from someone else, mainly Micron Technology Inc. (NASDAQ:MU), SK Hynix Inc. (NASDAQ:SKHY) and Samsung Electronics, and that memory has become the fastest-inflating line in the bill of materials for an AI server.
Nvidia said memory scarcity is “being driven in large part by the AI build-out itself.” In other words, the same demand surge creating Nvidia’s revenue growth is also making some of its inputs more expensive.
That makes memory inflation different from an ordinary cost shock.
The pressure is visible in Nvidia’s guidance. Gross margins are expected to be 74% in the third quarter, then fall to 71%–72% in the fourth quarter.
That would mark the first major sequential deterioration in margin of the current AI cycle.
Nvidia Can Raise Prices. But Timing Matters
Nvidia can raise prices to offset higher memory costs, but there is a delay. Memory prices are still rising now. Nvidia’s price increases will start showing up in fiscal 2028.
That means margins will take a hit first. They should recover later as higher prices are passed on to customers.
The key question for investors is how long that gap lasts.
Nvidia still has plenty of demand. The company says demand is growing about 100% next year, but supply limits revenue growth to roughly 70%.
As CEO Jensen Huang said, “The unconstrained would be a lot higher.”
The AI Supply Chain Is Changing
Nvidia is still growing at an extraordinary pace. But rising memory costs mean the company may not keep all the benefits of that growth.
Some of the money is flowing to the companies supplying the components Nvidia needs.
Memory is the clearest example.
Nvidia expects revenue to grow about 70% in fiscal 2028. But if memory prices remain high, Nvidia could earn slightly less profit per dollar of sales.
That makes gross margins almost as important as revenue growth in the next phase of the AI boom.
Image: Shutterstock
