DRAM shortage lasts 10 years Chairman dismisses AI bubble talk

The Memory Paradox: Why the AI Hype Might Be Overheating the Supply Chain

Amid the roaring excitement surrounding artificial intelligence and the subsequent massive investment in data centers, a voice from the heart of the semiconductor supply chain is offering a dose of sober reality. Simon Chen, chairman of Adata, suggested that while demand is undeniably explosive, the market narrative regarding memory chips needs recalibration. He forecasts that the global DRAM shortage will persist for another decade and dismisses notions of an imminent AI bubble.

Chen’s comments came at a pivotal moment, following recent stock volatility triggered by record quarterly results from TSMC. This environment naturally fuels speculation about whether the current investment cycle has become unsustainable. However, Chen argues that judging the market solely by short-term capital spending or a single company’s utilization rate misses the long-term picture of global demand.

The real story, he contends, is far broader than just data centers. While AI fuels massive GPU consumption, the demand for memory is set to expand across every sector imaginable. Chen points to an ecosystem of future edge devices—including autonomous vehicles, smart homes, unmanned factories, driverless cars, robots, and low-Earth-orbit satellites—a pool he estimates could reach tens of billions of units.

This vast array of applications shifts the focus from a single sector bubble to holistic industrial growth. He believes that when looking at the overall cycle, investors should wait until after 2030 to truly assess potential bubbles, perhaps revisiting the question of an AI bubble around 2040 or 2050.

Despite this long-term view, the immediate market is experiencing significant price pressure. Chen predicts that DRAM contract prices are set to climb by another 20% to 30% in the third quarter, while NAND flash is expected to rise by 35% to 40%. This escalation reflects the relentless upstream demand for memory.

The current dynamic shows memory makers actively responding to this environment. Companies like SK hynix are aggressively funding expansion, having recently secured a record $26.5 billion in its U.S. IPO to bolster High-Bandwidth Memory (HBM) manufacturing. Furthermore, competitors are also pushing boundaries; Nanya plans to quadruple capital spending to $6.2 billion by 2027, and China’s CXMT is reportedly nearing Micron-scale DRAM output this year.

Chen asserts that these market movements do not signal a widespread loosening of the market. Instead, he notes that the major suppliers have learned from past downturns and are opting for cautious, measured expansion rather than repeating the disorderly capacity races that previously caused price craters.

The data speaks to this resilience. Adata, one of the Taiwanese module makers, had strategically stockpiled more than NT$30 billion in chip inventory by late February. This proactive positioning allowed them to capitalize on rising contract prices, which have surged as much as 171% year over year, driven primarily by AI demand.

Ultimately, the memory industry is navigating a complex intersection of technological ambition and physical supply constraints. The outlook remains bullish for module-makers and price growth in the short term, underpinned by massive future applications, even as broader macroeconomic forces require careful consideration of long-term investment cycles.