The global rush to build artificial-intelligence infrastructure is now triggering a serious constriction in supply chains, this time around memory chips. A shortage of memory components, essential for everything from smartphones to high-end AI data centres, has emerged as demand in the AI and consumerelectronics sectors surges. Prices for both conventional DRAM and high-bandwidth memory (HBM), used widely in AI servers, have more than doubled in some segments.
Leading chipmakers, including Samsung Electronics, SK Hynix and Micron Technology, are finding it hard to raise output quickly enough to meet surging demand. Many have allocated production lines to HBM and other high-margin chips for AI, but in doing so have reduced output of the legacy memory used by PCs and smartphones. Inventory levels of DRAM plummeted to just two to four weeks in October 2025 from 13 to 17 weeks at the end of 2024.
The shock is already radiating through the technology and consumer-electronics supply chain: tech giants like Microsoft, Google, and ByteDance have scrambled to secure chip allocations, placing long-term or open-ended orders with chip suppliers. Retailers have started to ration memory products, while producers of smartphones and laptops have warned that steep price rises may be on the cards. In some regions, hard-disc drives and memory modules are subject to purchase limits as sellers try to prevent hoarding.
To companies, the shortage has moved well beyond a supply-chain hiccup; increasingly, it is being viewed as a macroeconomic risk. Executives and analysts warn that delays in AI data centre builds could slow broader technology upgrades and dampen productivity growth. Inflationary pressure is another concern, especially for economies already struggling to control rising costs.
Underpinning the crisis is a structural mismatch between where chip producers are investing and where demand remains high: The shift to making high-end chips for AI was reasonable because the need for AI technology has greatly increased since generative AI tools came out, and HBM chips provide the necessary speed for those demanding tasks. But the willingness to divert resources away from traditional memory has left consumer-electronics makers scrambling.
Efforts are underway to mitigate the shortage. Chipmakers have announced plans to expand capacity, though new factories, especially for conventional memory, are not expected to come online until 2027 or 2028. Some companies are scrambling to secure supply with long-term contracts far in advance. However, many observers caution that in a market characterised by volatile demand and thin inventories, only the largest and most financially resilient firms are likely to withstand the challenges, while smaller players and emerging markets may face difficulties.
For companies – from multinational tech giants to regional smartphone makers – the unfolding supply-chain crisis underscores the perils of supply-side dependency. A tight memory market deepens the case for supply-chain diversification and strategic inventory planning. As AI continues to reshape technology investment, firms must take stock of their upstream dependencies. For nations like Sri Lanka – and for emerging-market firms globally – the current crunch serves as a warning: overreliance on imported components, especially in a fast-shifting global tech landscape, can leave economies vulnerable to external shocks.
The unfolding scramble has also forced investors and policymakers to rethink the resilience of the semiconductor industry, which is increasingly treated as a strategic asset rather than a regular commercial sector. Capital is now pouring aggressively into memory-capacity expansions, but analysts caution that such investments risk creating cyclical whiplash: a period of severe undersupply could well be followed by a glut when new fabs come online. Overall, the shortage of memory chips shows how surging demand for AI capacity can strain essential supply chains. A race to ever more powerful data centres has turned into a broader warning of systemic fragility that promises to stall the very upgrades the boom of AI has promised.
