GPU Supply Crunch Shows No Sign of Easing as Memory Shortage Bites

The GPU supply problem has no clear end in sight, with AI data centers consuming memory as fast as it can be made. As chipmakers redirect resources toward the AI boom, the resulting memory shortage is hitting consumer tech hard: component prices are climbing, supplies are shrinking, and ordinary buyers are bearing the cost.

Known as "chipflation," GPU price hikes have far outpaced broader inflation. A JP Morgan report projects that by the end of 2026, DRAM prices will have risen 400 percent since 2024. Tom's Hardware's chips tracker found Nvidia's GeForce RTX 5090, which launched in January 2025 at a $1,999 suggested price, now retails $4,920 above that figure. In Japan, some gaming GPUs run roughly 41 percent above their American counterparts.

The effects ripple across consumer electronics. JP Morgan's report puts the Consumer Price Index for software and accessories up nearly a quarter over two years, while a 2026 Economist report found similar rises for consoles and laptops, alongside an unusual drop in PC shipments. Resellers have capitalized, with GameStop marketing Sony's PlayStation 5 Pro at roughly $500 above MSRP. Some expect the crisis to stretch into the 2030s, and many chip producers have already sold much of their 2027 supply, focusing sales on 2028 and beyond.

At its core, the shortage is a memory problem. The industry was already transitioning from DDR4 to DDR5, stalling supplies, and the rapid data center buildout has worsened things by demanding vast stores of high-bandwidth memory (HBM), a faster form of DRAM that AI developers need to train and run models. HBM competes for the same wafer capacity as DRAM, and suppliers have shifted production toward it, selling HBM for roughly 30 percent more than traditional DRAM. Transitions to HBM4 and HBM4E are further limiting supplies. A Samsung executive told Reuters HBM could account for nearly 30 percent of total DRAM production in 2027, up from 20 percent the year before, with some expecting almost 40 percent by 2029. AI data centers are projected to consume roughly 70 percent of global DRAM supplies this year.

The market is highly concentrated: Micron, Samsung and SK Hynix account for roughly 90 percent of global DRAM production, and all three expect long-term shortages despite capacity expansions. Micron ceased direct-to-consumer sales last year to focus on AI production. Micron CEO Sanjay Mehrotra said supply will be "much tighter in calendar 2027 and 2028" than in 2026, while SK Hynix CEO Kwak Noh-jung told Reuters he forecasts next year will be "the worst year in the industry's history from the supply perspective." SK Hynix chairman Chey Tae-won said building adequate wafer supply will take "at least four to five years," with production lagging 20 percent behind demand until at least 2030. Micron has reportedly sold 75 percent of its 2027 HBM output at prices "much higher than 2026 prices," and Samsung has sold almost 80 percent of its 2027 memory output, per KB Securities.

Not everyone is convinced. Acer CEO Jason Chen told DigiTimes prices could plateau next year, citing a potential influx of cheap Chinese alternatives. A South Korean publication quoted experts predicting a "flood of new supply" in 2029 would turn the memory market "into a red ocean," causing memory stocks to tumble in October 2026. Nvidia's Jensen Huang dismissed the possibility of a surplus, telling CNBC, "One of the things that's really great about memory is that you can always use more of it."

Source: Engadget