The price of DRAM memory chips, the working memory in everything from smartphones to servers, has more than doubled since early 2025 and surged a further 40–90 per cent in the first quarter of 2026 alone, according to Taipei-based analysts. The immediate effect is visible on retail shelves: Apple raised Mac and iPad prices by hundreds of dollars, Microsoft imposed its third Xbox price increase in 13 months, and the global PC market is forecast to contract 11.3 per cent this year as component costs bite. Industry executives describe the squeeze as a “hundred-year flood” for memory pricing, with no easing expected before 2028.

The mechanism is a direct consequence of the artificial-intelligence buildout. The three dominant DRAM manufacturers—Samsung, SK Hynix and Micron—have redirected more than 80 per cent of their advanced production capacity toward high-bandwidth memory (HBM) for AI servers, where margins are richer. Data centres now consume roughly half of global DRAM output, up from 32 per cent five years ago. At the same time, the physical infrastructure to power those data centres is straining: more than 2,500 gigawatts of generation and storage projects sit in grid connection queues worldwide, and US regulators have begun scrutinising how large-load customers such as data centres jump the queue. In Texas alone, nearly 90 per cent of 438 GW of large-load interconnection requests come from data centres, prompting the grid operator to create a fast-track process to separate speculative proposals from real projects.

The reallocation of memory supply pits deep-pocketed hyperscalers—Microsoft, Amazon, Google and Meta—against consumer-electronics makers and the automotive industry. Memory makers are enjoying a windfall, but the concentration of advanced fabrication in South Korea and Taiwan leaves global supply chains vulnerable. A class-action lawsuit filed in California in late June accuses the three memory giants of deliberately curtailing standard DRAM output to inflate prices, echoing antitrust cases from the 2000s. Meanwhile, the energy sector is consolidating rapidly: US utility M&A hit a record $203.6 billion in the first five months of 2026, as companies seek scale to finance the generation and transmission needed for data centres. In the labour market, the picture is uneven. Firms that invest heavily in AI are still hiring, including for entry-level roles, but newly minted computer-science graduates from top US universities report sending thousands of applications without securing a full-time position, as AI automates the coding tasks that once served as a training ground.

Capacity expansion is underway but will not bring relief quickly. South Korea announced a $520 billion national semiconductor plan, and Micron has committed $200 billion to US fabrication plants, yet meaningful new wafer output is not expected before mid-2027. On the energy side, the first realistic transmission plan for Texas’s queue is not due until autumn 2027. The next factual milestone to watch is the quarterly earnings guidance from the major memory suppliers, which will signal how much capacity they intend to allocate back to consumer-grade chips for the 2027 contract cycle.