Micron Technology's latest quarterly earnings call surfaced a forecast that should alarm both automakers and consumers eyeing future car purchases. The company predicts that as automakers introduce more vehicles capable of driverless operations in geofenced areas (L4 autonomy), bills of materials will escalate from 16GB of memory to over 300GB. For context, that is roughly equivalent to the total storage capacity of a high-end laptop, crammed into a vehicle's compute systems.
The forecast is not mere speculation. Micron reported Q2 2026 revenue of $23.86 billion, almost triple the $8.053 billion it posted for Q2 2025. That surge stems largely from AI infrastructure demand. But looking ahead, the company sees autonomous vehicles as another frontier market with insatiable appetite for fast, expensive memory. The challenge is that global memory supply is already stretched thin, and the competition for production capacity is fierce.
Here is the uncomfortable truth: the automotive industry is already losing the battle for semiconductor supply. A shortage of DRAM chips, which power ADAS systems, digital cockpits and infotainment units, is tightening supply and driving up costs across the automotive industry because data centres building AI infrastructure are consuming enormous quantities of memory chips, and manufacturers are chasing that higher-margin business instead of prioritising automotive supply. This is not a manufacturing capacity problem; it is a prioritisation problem. Chip makers have chosen where to send their wafers, and automotive is not winning.
DRAM prices could rise between 70% and 100% in 2026 compared to 2025 prices, a significant increase for premium cars with advanced cockpit and autonomy features which already had north of $150 of DRAM in 2025. General Motors signalled the cost impact bluntly. The company expects DRAM, unfavourable foreign exchange changes and higher prices on tariffed materials like copper and aluminium will increase costs by $1 billion to $1.5 billion this year.
The economics are stark. Memory manufacturers can allocate a unit of production capacity to either high-bandwidth memory for AI servers (where profit margins are extremely high) or automotive DRAM (lower margins). If automotive clients are ready to pay more to match the wafer value DRAM makers would get from other industries, then they will get the volume they need. In other words, automakers can secure supply, but only at premium prices that will flow directly to consumers.
The timing compounds the problem. New production lines from major memory manufacturers are not expected to come online until 2027 or even 2028, meaning no substantial capacity increase is imminent. During that window, as DRAM makers focus on newer generations of chips, the supply of older DRAM used in current cockpit and ADAS designs will dry up regardless of price, and for OEMs and tier 1 suppliers, most cars planned for production in 2028 have cockpit and ADAS designs that use current or older-generation DRAM and will all need to be redesigned to be compatible with the next generation of DRAM chips.
Premium vehicles and electric cars face the sharpest impact. Luxury and high-tech vehicles are most exposed, as premium models with advanced cockpits, rich displays and autonomy features require more DRAM. Conversely, traditional automakers building conventional vehicles with fewer electronic features enjoy some insulation from the crunch. Automakers using automotive system-on-a-chip technology that combines computing functions like ADAS, infotainment and vehicle control onto a single chip cannot build cars without DRAMs, though being a traditional automaker might have a slight advantage if needed they could remove certain content or features.
Micron's 300GB forecast assumes an optimistic timeline for Level 4 autonomy adoption. By 2030, almost three million cars are expected to be fully autonomous, and more than 15 million will support L2+/L3 autonomy. That scale could absorb the entire global memory supply increase for years. But Micron has a vested interest in that forecast. Higher automotive memory demand translates directly to higher revenue and margin recovery across its business. The company has already benefited enormously from AI infrastructure buildout; another growth vector softens the eventual plateau.
The counterargument deserves airing: perhaps the shortage will prove temporary and manageable. DRAM supply is more elastic if automakers are willing to pay, and DRAM capacity will be tight but still attainable if automotive clients are willing to pay more to match the wafer value DRAM makers would get from other industries. For buyers of premium vehicles, paying higher prices may be simply the cost of adding advanced features. For those shopping entry-level cars, disabled features or feature mix variability could become routine.
The structural issue is clearer. Memory manufacturers, facing a once-in-a-generation AI boom, have no commercial incentive to prioritise a slower-growth, lower-margin automotive sector. Supply constraints will likely persist through the decade. Automakers now face a choice: invest heavily in alternative architectures and local supply partnerships, accept rising costs and margin compression, or selectively scale back the electronic features that once promised to differentiate their vehicles. None of these paths is easy or cheap.