Ford CFO Sherry House told analysts at the Wolfe Research Auto, Auto Tech and Semiconductor Conference in February that the company expects about $1 billion in additional commodity costs this year from DRAM pricing and inflation, per the Detroit News. She framed it calmly. “This is something that we’ve been actively managing,” House said. “We do believe at this point in time that we have access to sufficient supply, but we are seeing pressure on pricing.”

That’s the polished version. The raw version is that Samsung, SK Hynix, and Micron control roughly 88% of the automotive DRAM market, and all three are sprinting toward AI money. The margins on high-bandwidth memory for data centers dwarf what automakers pay for the DDR4 and LPDDR4 chips stuffed into infotainment systems and ADAS modules. When the most profitable customer in the room wants your entire production run, the customer buying low single-digit percentages of your output doesn’t get to set terms.

Why This Isn’t 2021

The semiconductor shortage that gutted production lines from 2021 through 2023 was a broad supply crisis. Mature-node analog chips used in everything from power management to body electronics disappeared because capacity didn’t exist. Over 10 million vehicles went unbuilt. Plants shut down. Dealer lots went bare.

This is different. Global DRAM fabrication capacity exists. The wafers are there. The problem is allocation. AI data centers are forecast to consume up to 70% of all DRAM produced globally in 2026, according to industry estimates tracked by multiple analysts. One gigabyte of high-bandwidth memory used in AI servers requires four times the wafer area of standard DRAM, and it generates far higher margins. Chipmakers aren’t going to leave that money on the table to keep building DDR4 for car dashboards.

S&P Global Mobility’s semiconductor research team projects that automotive DRAM prices will spike 70 to 100% this year compared to 2025. A UBS analysis estimates automakers will absorb roughly 80% of those increased costs shared with suppliers, many of whom are already under financial pressure from tariffs and inflation. Suppliers specializing in electronics and ADAS are the most exposed.

DRAM inventory levels tell the story in sharper terms. Average industry inventory fell from about 17 weeks in late 2024 to just two to four weeks by October 2025. That’s crisis territory. Micron has confirmed it sold out AI memory contracts for the entire year and estimates it can only meet about two-thirds of medium-term demand for some customers. SK Hynix has similarly reported its 2026 production capacity is fully allocated.

Where the Pain Lands

The shortage hits selectively, and that selectivity is what makes it strategically interesting. DRAM lives in the compute-heavy systems: digital cockpits, advanced driver assistance, infotainment platforms, and the centralized computing architectures that automakers are racing to adopt. Premium vehicles with the most advanced feature sets carry the highest DRAM content per unit. Basic transportation with minimal electronics faces minimal exposure.

That means the brands most aggressively pursuing software-defined vehicle architectures are the most vulnerable. Tesla and Rivian run memory-intensive centralized computing platforms with limited ability to strip features if supply tightens. Traditional automakers retain some flexibility by adjusting optional content, but Ford and GM are both migrating to centralized architectures that increase their DRAM dependency with every new model year.

GM has indicated that DRAM costs, combined with foreign exchange pressure and tariff-driven increases in copper and aluminum, could add between $1 billion and $1.5 billion to annual expenses. Ford’s $1 billion estimate is specifically tied to DRAM pricing and inflation, with the company banking on material and warranty cost reductions to offset the hit.

Visteon, a major cockpit electronics supplier, flagged a 50% year-over-year increase in memory demand late last year, roughly five times normal levels. That kind of surge suggests panic buying has already started, which could create artificial shortages on top of the structural ones.

The Two-Year Clock

S&P Global Mobility’s analysis separates the crisis into two phases, and the distinction matters.

Through 2027, automotive DRAM capacity will be constrained but elastic. If automakers are willing to pay premiums that match the wafer value chipmakers could get from other industries, they’ll get the volume they need. Margins erode. Vehicle prices absorb some of the cost. But production lines keep running.

The second phase is worse. Beginning in 2028, the supply of older-generation DDR4 and LPDDR4 for automotive applications will start drying up regardless of what automakers are willing to pay. S&P Global notes that the top 10 cockpit designs and 8 of the top 10 ADAS designs planned for 2028 production still use DDR4 and LPDDR4. Those chips won’t be available at any price. The industry has roughly two years to redesign those systems and migrate to newer-generation LPDDR5, which will remain in active production.

That migration timeline is aggressive for an industry that typically qualifies new semiconductor components over 18 to 24 months. Some automakers will make it. Others will face the choice between delaying launches or shipping vehicles with reduced feature sets.

The Pattern GCBC Readers Should Recognize

This follows the same structural logic that drove Honda’s production cuts during the Nexperia crisis last fall. The auto industry represents a small fraction of total demand for a critical component. The suppliers who make that component have more profitable customers. And the automakers who invested most heavily in lean inventory and just-in-time efficiency are the first ones exposed when allocation tightens.

The difference this time is that the competing customer isn’t consumer electronics or cryptocurrency miners. It’s the largest capital deployment cycle in the history of the technology industry. Hyperscale data center spending showed no signs of slowing through Q4 2025, and every major cloud provider has announced expanded buildout plans for 2026 and beyond. The auto industry isn’t competing for scraps from a temporary demand spike. It’s permanently lower on the priority list for three companies that control nearly all the supply.

Ford’s F-Series lineup, America’s best-selling vehicle line, increasingly depends on sophisticated infotainment and ADAS systems that require substantial DRAM content. So does every other premium or technology-forward vehicle in every major OEM’s portfolio. The days when a pickup truck was a mechanical device with an optional radio are long gone.

S&P Global doesn’t expect the DRAM price increase to reduce overall light vehicle production in 2026. The costs are absorbable if painful. But the structural shift in chipmaker priorities isn’t temporary, the migration deadline is real, and any automaker that treats this as a short-term commodity blip rather than a fundamental supply chain realignment is making the same mistake the industry made in 2020 when it canceled chip orders and assumed the supply would be there when demand returned.

It wasn’t then. It might not be in 2028, either.