.png){kind=logo}
AI data-centre expansion is pushing global memory supplies into a severe shortage. Industry reports call the squeeze “RAMageddon” or “RAMpocalypse.” Demand for artificial intelligence systems now competes directly with phones, computers and standard business servers.
Nvidia recently told major customers that AI server prices could rise by more than 15%, according to Bloomberg. The increase would apply to some systems shipping in early 2027. Higher memory costs are also lifting prices across hardware, cloud capacity and managed technology services.
AI accelerators use high-bandwidth memory, or HBM, to move large datasets quickly between processors and storage. Producing equivalent HBM capacity can consume three to four times more wafer space than standard DDR5 memory. Therefore, every production shift reduces the supply available for conventional products.
Samsung Electronics, SK Hynix and Micron dominate global DRAM production. They have directed more capacity toward HBM and high-capacity server memory, where margins remain stronger. Hyperscalers including Microsoft, Google, Amazon and Meta also reserve supply through large, long-term orders.
This change has pushed HBM toward roughly one-fifth of DRAM wafer input among leading suppliers. Meanwhile, TrendForce forecast conventional DRAM contract prices would jump 90% to 95% quarter-on-quarter in early 2026. Some large server-memory orders now carry lead times exceeding 40 weeks.
Indian IT providers rely on imported servers, storage systems and networking equipment for client projects. Rising component prices increase the cost of new data centres and hardware upgrades. Short quote periods also make fixed-price contracts harder to manage when suppliers reprice equipment before delivery.
Data-centre operators also need more memory for AI inference, analytics and cloud workloads. Consequently, the same expansion that creates demand also raises each project's equipment bill during deployment.
Hardware makers have responded with higher list prices and revised contracts. Dell, Lenovo, HP and HPE have indicated increases for servers and computers. Consequently, Indian businesses face larger budgets for AI deployments, cloud services, workplace computers and routine infrastructure replacement.
The pressure has also reached consumers. Counterpoint Research reported a 10% annual fall in Indian smartphone shipments during the June quarter. Rising device prices weakened demand, while lower-priced phones faced greater pressure from memory costs. The downturn marked the weakest June-quarter performance in six years.
India is increasing semiconductor and data-centre investment, although new plants cannot quickly repair global supply. Micron opened its assembly and test facility in Sanand, Gujarat, in February. The $2.75 billion project has started commercial production of finished memory and storage products using imported wafers.
The government also seeks up to $200 billion in data-centre investment over several years. That expansion will raise local demand for servers and memory. Meanwhile, the India Semiconductor Mission supports new fabrication, packaging and testing projects designed to reduce import dependence.
China is also expanding domestic production while United States export controls restrict its access to advanced HBM and manufacturing equipment. CXMT is adding DRAM capacity, while YMTC is growing in NAND flash. These additions may improve conventional memory supply, but advanced AI memory still requires specialised tools and packaging.
Building a modern fabrication plant can cost billions of dollars and take years. Therefore, current projects offer little immediate relief. Indian IT firms, data-centre operators and electronics makers must manage higher costs while global suppliers continue prioritising AI-grade memory.
ALSO READ: Apple Could Make iPads and MacBooks in India Under New Manufacturing Plan
