Here's a new wrinkle in Nvidia Corp. (NVDA)'s already complicated relationship with China: Even if the U.S. government approves export licenses for its AI chips, there might not be enough memory to build them.

According to Rep. John Moolenaar, the top Republican on the House China committee, tight supplies of advanced memory chips are likely to cap how many export licenses Nvidia can actually use to sell its H200 artificial intelligence processors to Chinese customers. And with demand for AI hardware still running white-hot, this creates a fresh near-term risk to Nvidia's sales growth.

In a letter to Commerce Secretary Howard Lutnick, Moolenaar flagged that shortages of DRAM, particularly the high-bandwidth memory used in AI accelerators, present an "immediate challenge" under new licensing terms. The issue? A recently implemented rule requires exporters to certify that approved shipments to China won't trigger shortages in the U.S. market, according to Bloomberg.

Nvidia Says It Can Handle the Pressure

Nvidia, for its part, doesn't seem too worried. The company said it regularly manages its supply chain and can fulfill all approved H200 orders without compromising supply for other products or customers. Translation: We've been juggling scarce components for years now, and we know what we're doing.

But the broader picture is trickier. High-bandwidth memory, which is built from stacked DRAM and critical for AI workloads, comes mainly from three suppliers: Samsung Electronics Co. Ltd (SSNLF), SK Hynix, and Micron Technology Inc. (MU). All three have recently warned that supply remains limited as AI data center demand continues to surge.

Samsung Sees an Opening and Takes It

When there's a shortage, somebody always makes money. In this case, it's Samsung, which is using the global memory crunch to push through sharp price increases across its product line.

The South Korean chipmaker has raised prices on key memory products by as much as 60% since September, according to Reuters. The shortage, driven largely by AI servers that consume enormous amounts of DRAM, NAND, and high-bandwidth memory, is forcing companies expanding their AI infrastructure to absorb significantly higher costs.

And here's the kicker: Those pressures could eventually trickle down to consumer products like smartphones and PCs, which rely on the same components. When AI data centers are panic-ordering memory chips, everyone else gets squeezed.

Someone Has to Eat These Costs

Which brings us to an uncomfortable question for device makers: Who absorbs the pain?

Investors are now weighing whether companies like Apple Inc. (AAPL) and HP Inc. (HPQ) can protect their profit margins or will have to pass costs along to consumers and risk dampening demand, Bloomberg reported.

"They're in a tough position," Rob Thummel, senior portfolio manager at Tortoise Capital, told Bloomberg. "They basically have two options: They can take a hit to margins, which the market won't like. Or they can raise prices to offset the higher memory costs, running the risk of hurting demand."

Neither option is particularly appealing, which helps explain why chipmakers tied to smartphones have been catching downgrades recently. Qualcomm Inc. (QCOM) and Arm Holdings Plc (ARM) are among those cited as facing risk from rising memory costs.

Memory Suppliers Are Having a Moment

While device makers and smartphone chipmakers sweat it out, memory and storage suppliers are enjoying the ride. SanDisk Corp. (SNDK) has led the S&P 500 to start 2026, up about 75% year-to-date. Western Digital Corp. (WDC) and Micron have also ranked among the index's top performers, building on strong gains from 2025.

The market logic is straightforward: When supply is tight and prices are rising, the companies controlling that supply tend to do very well. For now, at least, Samsung, SK Hynix, and Micron are in the driver's seat.

But for Nvidia, the memory shortage adds yet another constraint to an already challenging situation in China. Even with export licenses in hand, building and shipping H200 processors requires components that are increasingly hard to come by. It's a reminder that in the semiconductor industry, no single bottleneck exists in isolation. When memory runs short, everyone downstream feels it, from AI chipmakers trying to serve Chinese customers to smartphone makers trying to keep costs under control.

And as AI infrastructure continues its breakneck expansion, those supply chain pressures aren't going away anytime soon.