The scramble for the specialized memory chips that power artificial intelligence has a new timeline. According to Finimize, SK Hynix says shortages of AI memory could peak in 2027 — meaning the supply squeeze that has defined the chip market may still have room to tighten before it eases.
Memory chips are the unglamorous but essential counterpart to the flashier processors made by companies like Nvidia. AI systems need vast amounts of high-speed memory to train and run large models, and demand has outpaced what manufacturers can produce. SK Hynix's assessment suggests that imbalance is not a passing blip but a multi-year condition, with the tightest point still a couple of years out.
Investors are already reacting to the memory boom. Reporting on the day's market moves, MSN noted that rival memory maker Micron gave stocks a roughly $250 billion AI boost — a sign of how much financial weight the market is now placing on the companies that supply AI hardware. When a single chipmaker can swing hundreds of billions in market value, it underscores how central memory has become to the broader AI trade.
Taken together, the two signals point in the same direction: memory is now a key bottleneck in the AI build-out, and the firms that make it are among the biggest beneficiaries. A shortage that peaks in 2027 also implies years of elevated prices and fierce competition for supply, which could ripple through to the cost of everything AI touches.
Why it matters: if the companies actually making these chips expect the crunch to intensify through 2027, the price and availability of AI — from data centers to the products built on them — will hinge on how fast memory supply can catch up.