A severe shortage of DRAM — the volatile memory chips that power everything from smartphones to servers — is quietly building into one of the most consequential supply chain disruptions the technology industry has faced in years. While much of the world’s attention has been focused on tariffs and trade wars, the companies that design and build the devices consumers rely on daily are confronting a stark reality: there simply isn’t enough memory to go around, and the consequences could be devastating for product launches, corporate bottom lines, and entire categories of consumer electronics.
The warning signals have been growing louder for months, but a recent interview with Phison Electronics CEO Khein Seng Pua brought the crisis into sharp relief. Speaking with The Verge, Pua described a situation in which the DRAM shortage has become so acute that it could “kill products” and potentially take down companies that are unable to secure adequate supply. Phison, a major controller chip maker for solid-state drives and other storage products, sits at a critical junction in the memory supply chain, giving Pua a front-row seat to the unfolding crisis.
An Industry Running on Empty: Why DRAM Supply Can’t Keep Up
The roots of the current shortage are multifaceted. On the demand side, the explosive growth of artificial intelligence workloads has created an insatiable appetite for high-bandwidth memory (HBM) chips, which are essentially stacked DRAM dies engineered for the kind of parallel processing that AI training and inference require. Companies like Nvidia, AMD, and a host of AI startups are consuming vast quantities of HBM for their accelerators, and the three major DRAM manufacturers — Samsung, SK Hynix, and Micron — have been aggressively reallocating production capacity toward these higher-margin products.
That reallocation comes at a direct cost to conventional DRAM supply. The same fabrication lines that once produced DDR5 modules for PCs, LPDDR5X for smartphones, and server-grade memory are now being retooled or prioritized for HBM production. As Pua explained to The Verge, the math is unforgiving: when a fab shifts capacity to HBM, the output of standard DRAM drops correspondingly, and there is no quick fix. Building new semiconductor fabrication facilities takes years and billions of dollars, and even expanding existing ones requires lead times measured in quarters, not weeks.
The AI Hunger Machine and Its Collateral Damage
The irony of the situation is hard to miss. The very technology that is supposed to transform industries — artificial intelligence — is now threatening to starve the broader electronics market of a fundamental component. Every major cloud provider, from Microsoft and Google to Amazon and Meta, has been racing to build out AI data center capacity, and their orders for HBM-equipped GPUs and accelerators have been enormous. SK Hynix, which has emerged as the dominant supplier of HBM chips to Nvidia, has reportedly sold out its HBM production through 2025 and is already taking orders for 2026 allocations.
This dynamic has created a two-tier market. Companies with deep pockets and strategic relationships with memory manufacturers can secure supply, albeit at rising prices. Smaller firms, startups, and companies in less glamorous segments of the electronics industry — think IoT devices, budget smartphones, automotive electronics, and consumer peripherals — are finding themselves squeezed out. Pua’s warning to The Verge was blunt: some products will simply never make it to market because their makers cannot procure enough DRAM at prices that make economic sense.
Prices Climbing, Margins Shrinking
DRAM contract prices have been on a steady upward trajectory. According to industry tracker TrendForce, DRAM prices rose significantly in the first half of 2025, with DDR5 modules seeing some of the steepest increases. For PC makers already operating on thin margins, these price hikes are painful. For smartphone manufacturers in the mid-range and budget segments, where bill-of-materials costs are tightly controlled, they can be fatal to a product’s viability.
The pricing pressure extends beyond the chips themselves. As DRAM becomes scarcer, the components that depend on it — SSDs, memory modules, embedded systems — also see their costs rise. Phison, as a controller chip company, depends on the availability of NAND flash and DRAM to sell its products. If memory manufacturers are diverting DRAM production to HBM, the downstream effects ripple through the entire supply chain. SSD makers need DRAM for cache buffers in their drives. Smartphone makers need it for system memory. Server builders need it for everything. When supply tightens across the board, the competition for remaining inventory becomes fierce.
Echoes of Past Shortages — But This Time Is Different
The tech industry has weathered DRAM shortages before. The 2017-2018 memory shortage, driven by strong smartphone demand and a transition to new manufacturing processes, sent prices soaring and prompted antitrust investigations in multiple countries. But industry veterans say the current situation has a different character. The 2017 shortage was largely cyclical — a temporary mismatch between supply and demand that corrected as manufacturers ramped up production. The current shortage, by contrast, appears structural.
The structural nature of the problem stems from the fact that HBM is not a temporary fad. AI workloads are expected to grow exponentially for years to come, and the demand for HBM will grow with them. Each new generation of AI accelerators requires more memory bandwidth and more HBM capacity. Nvidia’s next-generation Rubin platform, for example, is expected to use even more HBM per chip than the current Blackwell architecture. This means that the share of global DRAM capacity devoted to HBM is likely to increase, not decrease, putting permanent pressure on conventional DRAM supply unless significant new capacity comes online.
Who Gets Hurt First — and Who Survives
The companies most vulnerable to the shortage are those without the scale or strategic importance to command priority from Samsung, SK Hynix, or Micron. Large OEMs like Apple, Dell, HP, and Lenovo have long-term supply agreements and the purchasing volume to ensure they receive allocations even in tight markets. But smaller PC makers, niche electronics companies, and firms in emerging markets may find themselves rationed or priced out entirely.
Pua told The Verge that the situation could lead to market consolidation, as companies unable to secure memory supply are forced to delay, cancel, or abandon products. In the worst case, some firms could face existential threats if they cannot deliver products to market on schedule. This is particularly concerning for the SSD market, where Phison operates, because SSDs are a commodity product with razor-thin margins — any significant increase in component costs can render a product uncompetitive overnight.
The Geopolitical Overlay: Tariffs and Trade Restrictions Add Fuel
Compounding the supply-demand imbalance are the geopolitical tensions that continue to roil the semiconductor industry. U.S. export controls on advanced chip technology to China have forced Chinese companies to stockpile memory and other components, further tightening global supply. Meanwhile, tariffs imposed on semiconductor imports have added costs at multiple points in the supply chain, making it more expensive to move chips across borders.
The situation in China is particularly complex. Chinese memory manufacturers like CXMT (ChangXin Memory Technologies) have been ramping up domestic DRAM production, but their output remains a fraction of what Samsung, SK Hynix, and Micron produce. And with U.S. restrictions limiting China’s access to the most advanced chipmaking equipment, Chinese firms are largely confined to producing older-generation DRAM, which does not fully address the market’s needs for DDR5 and LPDDR5X.
What Comes Next: No Quick Fix in Sight
Industry analysts do not expect the shortage to ease significantly before late 2026 at the earliest. Samsung, SK Hynix, and Micron have all announced plans to expand capacity, but new fabs take two to three years to build and equip. In the meantime, the memory manufacturers are in an enviable position: high demand, limited supply, and rising prices translate directly into fatter profit margins. For everyone else in the value chain — from device makers to consumers — the outlook is less rosy.
Consumers should expect to see higher prices for PCs, smartphones, and storage devices in the coming quarters. Some product categories may see reduced configurations — less RAM in base models, smaller SSD cache buffers — as manufacturers try to stretch limited supply. And some products, as Pua warned, may simply never appear. The DRAM shortage is not just a supply chain inconvenience; it is a force that will shape which products exist, which companies survive, and how quickly the broader tech industry can grow in an era increasingly defined by artificial intelligence’s voracious appetite for memory.