Inside the Sudden RAM Price Surge
The memory industry just changed overnight.
Yesterday, Micron—a global leader in the semiconductor, memory, and storage industries—announced that it was shutting down its consumer electronics division, Crucial. On Wednesday, they publicly stated that they “made the difficult decision to exit the Crucial consumer business to improve supply and support for [their] larger, strategic customers in faster-growing segments.” The company was referring primarily to AI customers.
Crucial’s shutdown was a surprise. AI has maintained significant control over the memory market for quite a while, but the complete shutdown of one of the three main companies that sell consumer RAM—in favor of AI—was a gamble that caught many off guard.
Working with large language models takes a significant amount of memory, or RAM. AI training depends heavily on high-bandwidth memory systems, making RAM one of the industry’s most valuable resources. As a result, AI companies have struck deals with global leaders in the industry. In October, OpenAI entered into a deal with SK Hynix and Samsung, who agreed to produce up to 900,000 DRAM modules per month for OpenAI’s Stargate Project, an initiative dedicated to developing OpenAI’s infrastructure.
Crucial’s shutdown is a serious threat to PC builders and tech enthusiasts, who are already suffering from skyrocketing GPU and storage prices, thanks to high demand from AI companies. The shutdown has negatively impacted consumer RAM pricing. Over the span of a few weeks, consumer memory, like the RAM used in desktop PCs, has experienced a threefold increase in price. RAM leaders have no incentive to sell cheap consumer RAM—enterprise AI customers are dramatically more profitable than consumers.
There’s a lot behind Micron’s major decision. By shutting down all of its consumer operations, Micron is taking a risk. They’re betting on the significance of the AI boom, hoping that hype will stay around. But as the AI bubble grows, so does the risk of staying in it.
“Are we in an AI bubble? Of course!” Pat Gelsinger, former Intel CEO, says. “Of course we are. I mean, we’re hyped, we’re accelerating, we’re putting enormous leverage into the system.” Gelsinger says that although the AI-driven market is already in a bubble, it would be “several years” before he sees it ending.
Companies need to be mindful of the dangers that lie in the AI bubble. Today, nearly everything in tech feels linked to AI. Almost every time there’s significant fluctuation in a big tech stock’s price, it can be attributed to an AI deal. Billions of dollars circulate between different AI companies every day.
NVIDIA, a world leader in AI and the first public company to reach a market cap of $4 trillion, has tried to reach out of the bubble. Almost all of NVIDIA’s revenue comes from AI customers and infrastructure, but the company, which commands 90% of the AI chip market, maintains a major spot in other industries—like gaming graphics—to some extent, as a security measure. NVIDIA, among other major names in the AI space, is concerned about the risk of its current clientele concentration—especially when things could go wrong so quickly. DeepSeek R1’s release, which threatened demand for NVIDIA’s infrastructure, dropped NVIDIA shares by 17% in one day.
Micron’s exit from the consumer market is a message. Unlike NVIDIA, they’re going all in, betting that the AI bubble’s inevitable pop will leave them in a good position (reminiscent of some of the dot-com bubble’s biggest players).
The tech market is in an interesting position. Industries like the memory, storage, GPU, and semiconductor markets have been reshaped by AI, and numerous companies have shifted resources towards AI development. Micron’s exit from the RAM market shows how easy it is to dive headfirst into the AI bubble—but we have yet to see how it will all end.