The complaint, delivered in the careful language of a CEO who knows every word will be parsed, was aimed at an unnamed customer, and no one in the industry had any doubt who he meant. Over the past few years, Micron Technology’s chief executive said, some customers pressured memory manufacturers to cut prices, and the consequences were damaging — prices fell to a third of their earlier level in 2023, leaving memory makers without the cash to invest in new capacity. The remarks, widely read as a swipe at Apple, Micron’s largest customer, capture how the balance of power between memory suppliers and consumer electronics giants has shifted since AI began consuming every available chip.
The math behind the complaint is not in dispute. Memory prices collapsed in 2023 as smartphone and PC demand stalled, and the three companies that dominate DRAM and NAND production — Micron, Samsung, and SK Hynix — cut output to stop the bleeding. Micron’s prices fell to roughly a third of their peak, and the company, like its rivals, shelved expansion plans. The buyers who had pushed hardest for low prices, according to industry executives, were the handset makers and PC makers who treat memory as a commodity to be squeezed in quarterly negotiations.
Then AI changed the calculation. Training and running AI models consumes memory in quantities the industry had never seen: a single advanced accelerator needs memory measured in the hundreds of gigabytes, and a data center housing thousands of them needs memory by the ton. High-bandwidth memory, the specialized type used in AI accelerators, has been in chronic shortage for two years, and its prices have climbed even as the broader market stabilized. Memory makers, finally, are in a seller’s market.
The CEO’s point, made in an interview published this week, is that the buyers who squeezed prices in 2023 are now paying the price. Memory capacity that was never built, he argued, is part of the reason AI customers face shortages today. Companies that demanded rock-bottom prices from their suppliers are now competing for the same limited supply of high-bandwidth memory as everyone else, and they are paying premiums to secure it. The implied lesson for Apple, the largest consumer of conventional memory in the world, is that the era of cheap memory is over.
Apple has not responded publicly to the remarks, and Micron’s chief did not name the company, a fact that itself is part of the story. The two companies have a long and complicated relationship: Apple is among Micron’s biggest customers, but it also buys from Samsung and SK Hynix, and it has built a reputation as a tough negotiator that plays suppliers against each other. The public friction, industry executives say, reflects a deeper change: memory is no longer a commodity that phone makers can dictate terms on, because the AI boom has made it strategic.
The strategic shift is visible in the numbers. Memory makers are now allocating production toward high-bandwidth memory and other AI-specific products, and the capacity that used to serve smartphones and PCs is being redirected. Analysts estimate that a meaningful share of the industry’s wafer output is now committed to AI memory, leaving less for traditional devices. The result is a market where phone makers must compete with data centers for supply — and data centers, with their enormous budgets, are winning.
The balance of power has also changed structurally. The memory industry consolidated sharply after the 2023 downturn, and the surviving players are more disciplined about capacity, signaling to customers that prices will not return to the old lows. Micron, Samsung, and SK Hynix have all said they will add capacity only where demand justifies it, and they have shown a new willingness to walk away from low-margin business. For buyers accustomed to annual price declines, the message is unwelcome.
For Apple, the stakes go beyond memory prices. The company sells more than 200 million iPhones a year, each containing memory that must be bought in vast quantities, and its margins depend on component costs. A persistent memory shortage, or a prolonged period of high prices, would eat into those margins or force higher device prices. Apple has responded by diversifying suppliers and, according to people familiar with its plans, by investing in memory supply relationships that lock in volumes, but it cannot escape the underlying reality: the market has turned.
The broader meaning is a shift in how the tech economy’s biggest companies deal with each other. For two decades, consumer electronics makers dictated terms to their component suppliers, using their scale to force prices down. AI has flipped that dynamic for memory, and suppliers are savoring the reversal. The Micron chief’s remarks were pointed precisely because they were rare: suppliers rarely criticize their largest customers in public, and doing so signals confidence that the power balance has genuinely changed.
Whether the confidence is justified will be tested in the coming quarters. Memory prices are notoriously cyclical, and the current boom could fade as new capacity comes online. But the structural factors — AI demand, industry consolidation, disciplined capacity — suggest the seller’s market will last longer than past cycles. For Micron’s CEO, the interview was an opportunity to state a simple case: memory is no longer cheap, the buyers who demanded it be cheap share the blame for the shortage, and the suppliers who were squeezed are now in control.
