The global scramble for memory chips that pushed up smartphone prices late last year has moved squarely into the automotive sector. Car-grade memory — particularly DRAM — is in short supply, sending spot prices tumbling upward and forcing automakers to confront higher component costs just as electric-vehicle (EV) production and software content accelerate.
Specialist trackers and banks report eye‑watering moves. TrendForce expects broad DRAM prices to rise 55–60% and NAND Flash 33–38%; car‑grade DDR5 spot prices have surged as much as 300%. UBS noted that car‑grade DRAM rose around 180% in three months, while monitoring firms say DDR4 prices — the workhorse for automotive storage — have climbed more than 150% since mid‑2025.
The cause is structural. Samsung, SK Hynix and Micron together control more than 80% of DRAM capacity, and major producers are reallocating wafer starts away from low‑margin DDR4 toward higher‑value HBM and DDR5 chips for AI datacentres. HBM (high‑bandwidth memory) is purpose‑built for large AI models and commands prices many times that of standard DRAM, making it a far more attractive product line for foundries and integrated vendors.
That shift leaves automakers exposed. Automotive procurement accounts for under 5% of global memory purchases, so carmakers are often outbid by deep‑pocketed hyperscalers and AI companies. Suppliers and tier‑one chip brokers report DRAM prices in the automotive channel rising two‑ to four‑fold; some OEMs say they cannot compete on price with AI buyers who are investing billions to secure capacity.
The cost impact is no abstraction. A mid‑level smart EV typically uses five to six DRAM packages; a component that once cost around 700 yuan per vehicle has swollen to roughly 2,000 yuan at current quotes. HSBC estimates upstream cost inflation — memory and other materials — could add 1,000–3,000 yuan to EV production costs, squeezing margins or forcing higher retail prices at a time when consumer sensitivity to price remains high.
Some manufacturers have partial protection. Large OEMs such as Volkswagen and BYD and certain suppliers that negotiated long‑term deals or pre‑built inventories have secured preferential allocation, and smaller domestic chipmakers and traders have stepped in to help fill gaps. Beijing Junzheng, for example, says it built up inventories and signed wafer supply agreements to shield incumbent customers, yet it still struggles to meet a rising tide of new demand.
Market responses will be mixed. Automakers can try to absorb costs through vertical integration, renegotiated supplier contracts, product redesigns that reduce memory density or by postponing feature roll‑outs. Policymakers can tilt procurement or offer incentives for domestic car‑grade memory production. But absent a faster rebalancing of industry capacity or a drop in AI demand for HBM/DDR5, tightness is likely to persist through 2026; whether pressures ease in 2027 depends on investment decisions by the big memory makers and shifting demand mixes.
This episode underlines a wider reordering of global supply chains: the AI boom is not only inflating cloud bills and server inventories, it is diverting basic semiconductor resources and altering the economics of legacy chips. For the auto industry — where software and compute are becoming integral to product differentiation — the memory squeeze is a reminder that digital ambition now competes directly with industrial scale for scarce semiconductors, with implications for vehicle pricing, margins and the pace of electrification.