DDR5 16G chips averaged $39 in February, up 7.4% month-over-month, while 1Tb TLC flash wafers jumped 25% to $25, marking the steepest single-month gains reported yet. These price movements come despite a seasonal slowdown from the Lunar New Year, signalling that underlying structural pressures remain intact.
The fundamental problem is straightforward: AI infrastructure continues to pull memory capacity toward server DRAM and high-bandwidth memory, leaving conventional DRAM and consumer NAND segments undersupplied. The demand from hyperscalers such as Microsoft, Google, Meta and Amazon has forced the three biggest memory manufacturers to pivot toward higher-margin enterprise-grade components, creating a zero-sum dynamic where every wafer allocated to an HBM stack for an Nvidia GPU denies supply to the DRAM module of a mid-range smartphone or the SSD of a consumer laptop.
For Australian businesses dependent on electronics imports, this matters directly. Australia is almost entirely dependent on foreign-controlled microchip technology, making it increasingly vulnerable to global supply-chain shortages. Manufacturers relying on stable component costs face an uncomfortable reality: prices are expected to rise substantially before any relief arrives. Conventional DRAM contract prices are forecast to increase by 55–60% quarter-on-quarter, whilst NAND Flash prices are expected to rise by 33–38% quarter-over-quarter.
Yet it's worth acknowledging the legitimate economic argument behind this market shift. Memory manufacturers are responding rationally to price signals. Cloud service providers tend to be less sensitive to price increases, which is why memory suppliers are expected to raise average selling prices more aggressively. The AI infrastructure buildout is genuine and economically significant. Hyperscalers are willing to absorb costs because the strategic value of AI capabilities justifies it.
The supply-side story is less straightforward. Even if investment in memory capacity ramps now, lead times of 18–24 months will delay relief, with the last downcycle having burned manufacturers who overbuilt, leaving no one wanting to overbuild again—but underbuilding has consequences. This creates a genuine policy quandary: memory manufacturers cannot expand capacity overnight, and the historical boom-bust cycle makes them cautious about throwing capital at new fabs.
The increasing gap between demand and supply is driving rapid spot price increases and procurement capital pressures which, if continued, could lead to industry cycle collapse. That prospect should concern not just hardware makers but policymakers. A destabilised electronics manufacturing base affects too many downstream sectors—defence, automotive, consumer devices—to dismiss as mere market volatility.
Australia's semiconductor vulnerability during this period reinforces a longer-term strategic problem. Commercial semiconductor chip manufacturing capability is nearly absent in Australia, and opting out of semiconductor manufacturing will severely constrain Australia's growth as a technological nation. This memory crunch, painful as it is, illustrates why friend-shoring supply chains and domestic manufacturing investment matter.
The path forward requires balancing competing priorities. Memory manufacturers need confidence to invest in capacity. Electronics makers need prices to stabilise. Governments need supply resilience. No single actor can deliver all three. The pragmatic approach is to acknowledge that near-term price pain is likely unavoidable given the structural shift toward AI, whilst simultaneously building the policy framework—including targeted investment in regional semiconductor capabilities—that reduces Australia's dependence on global memory cycles.