he 2021 chip crisis showed how one missing semiconductor can bring automotive production to a standstill.
The next supply challenge may be more complex: selected semiconductors and printed circuit boards (PCBs) are tightening at the same time – driven by different mechanisms, but converging in the same vehicle electronics.
On the semiconductor side, the pressure starts with AI. Rapid growth in data centers is driving demand for High Bandwidth Memory (HBM), advanced DRAM and NAND. These products generate significantly more value per wafer than many mature automotive memories. Analyst estimates put the value generated from a 12-inch wafer at close to $20,000 for mainstream NAND used in AI-related applications, compared with roughly $6,000–8,000 for SLC NAND typically used in automotive and industrial applications.
The consequence is straightforward: Memory manufacturers have an economic incentive to prioritize high-value AI-related products, while investment in older memory technologies becomes less attractive.
Automotive is particularly exposed because many ECUs still rely on DDR4/LPDDR4 for digital cockpits, infotainment, instrument clusters and ADAS/domain controllers, NOR Flash for boot code, firmware and safety-critical software, and SLC/MLC NAND for maps, diagnostics, OTA packages and other persistent data – and these platforms are planned to remain in production for years.
The effects are already visible in both price and availability. Between 2025 and 2026, estimated DDR4 prices increased by around 196%, SLC NAND by around 406%, and NOR Flash by around 280%. Lead times moved in parallel: DDR4 increased from approximately 8 to 29 weeks, SLC NAND from 10 to 22 weeks, and NOR Flash from 11 to 26 weeks.
The underlying mechanism is likely to persist: AI memory consumes more wafer capacity and advanced material and packaging inputs, while commanding significantly higher value. As memory manufacturers prioritize these products, less flexible capacity remains for legacy automotive technologies – pushing automotive lead times out even when overall semiconductor production continues to grow.
This is not yet a broad semiconductor shortage comparable with 2021. It is a selective squeeze concentrated in legacy technologies that automotive cannot replace quickly.
The PCB squeeze follows a different mechanism. Large AI customers are securing the output of entire Asian PCB factories for several years, and manufacturers are prioritizing these high-volume, high-margin orders. AI server boards also require far more layers and raw material than conventional PCBs, increasing demand for laminates, copper foil and glass fabric.
A limited number of OEMs and Tier-1s have already reacted by entering longer-term agreements or securing dedicated supply for critical electronics. However, many traditional automotive sourcing contracts provide long-term commercial relationships without guaranteeing dedicated production capacity. When AI customers commit capacity several years in advance, automotive buyers without equivalent capacity commitments face materially lower delivery reliability.
The effect is already visible: PCB base-material costs rose around 70% between 2025 and 2026, while FR-4 lead times increased from approximately 3 to 8 weeks. For automotive, the risk is therefore shifting from pure price inflation toward capacity access and delivery certainty.
Around 94% of global PCB production is located in Asia, with China alone accounting for close to 60%, while semiconductor fabrication and many upstream electronics materials are similarly concentrated in a small number of Asian economies.
This concentration becomes a geopolitical risk because a disruption does not need to eliminate global capacity to stop automotive production. A Taiwan Strait escalation, export restriction, trade dispute or disruption of major Asian shipping routes could simultaneously affect semiconductor fabrication, PCB production, critical materials and logistics.
With limited alternative capacity available elsewhere, automotive cannot rapidly shift sourcing to another region.
Automotive cannot simply buy an available chip or PCB from another component supplier.
Vehicles operate under demanding thermal, mechanical and electrical conditions, and failures in safety-relevant electronics can directly affect occupants and other road users. Automotive components therefore undergo extensive validation and approval processes to ensure functional reliability and vehicle safety before they can enter serial production.
Changing a memory device may require electrical validation, software adaptation and ECU testing. Changing a PCB fab, laminate or stack-up can require reliability testing, process validation and customer approval.
These requirements are essential for safety – but they also create a structural disadvantage when supply becomes constrained. Other industries may shift to available capacity relatively quickly; automotive often cannot.
Available global capacity is therefore not the same as available automotive capacity.
Our base case is that supply pressure remains elevated through 2027 and into 2028. For chips, the near-term risk remains concentrated in legacy memory and mature technologies. For PCBs, the concern is increasingly capacity access and delivery certainty as large AI customers lock in multi-year volumes while material costs and lead times increase.
Over time, however, a second semiconductor risk will emerge. As software-defined vehicles consolidate more functions into central and domain controllers, automotive will increasingly depend on leading-edge 5 nm, 3 nm and eventually 2 nm compute SoCs. These chips are produced by a very small number of foundries and use the same advanced manufacturing capacity demanded by AI accelerators and data-center processors. Automotive will therefore compete much more directly with AI companies for leading-edge wafer capacity. AI accelerators can generate 20–100 times more revenue per finished chip than automotive compute SoCs, giving hyperscalers and AI chip companies substantially greater economic headroom to secure scarce 5 nm, 3 nm and eventually 2 nm capacity.
Additional semiconductor and PCB capacity should gradually improve overall availability, but this will not remove the structural risk. Legacy technologies may be phased out, new plants need time to ramp, and leading-edge foundry capacity will remain highly concentrated and commercially attractive to AI customers.
The semiconductor risk therefore becomes bipolar: mature nodes and legacy memory remain critical for peripheral electronics, while SDVs create a new dependency on a small number of highly critical leading-edge compute chips.
The bottleneck shifts from general capacity to qualified, architecture-compatible and strategically accessible capacity.
For mature semiconductors, OEMs and Tier-1s should identify the devices that cannot be redesigned within available inventory coverage. Critical DDR4/LPDDR4, NOR and sole-sourced parts may justify longer-term capacity commitments, allocation agreements or last-time buys.
For future SDV platforms, leading-edge compute must become a strategic sourcing topic: OEMs and Tier-1s should secure access to leading-edge foundry capacity at 5 nm, 3 nm and eventually 2 nm capacity before AI and data-center demand absorbs the available supply.
At the same time, high-volume ECUs need funded migration roadmaps away from technologies approaching phase-out.
For PCBs, the priority is different. Companies need visibility beyond the nominal board supplier: which fab produces the board, which laminate and glass fabrics are specified, where critical upstream materials originate and whether production capacity is contractually secured.
For critical designs, alternative fabrication sites and material sets should be prepared in advance rather than first addressed when the primary source becomes constrained.
At the next platform generation, resilience needs to become a design decision. Memory flexibility should be considered at architecture freeze. PCB stack-ups and material specifications should be standardized where technically and commercially sensible.
This means identifying the relatively small number of dependencies capable of stopping production – and protecting them before the market forces the decision.
You need to load content from reCAPTCHA to submit the form. Please note that doing so will share data with third-party providers.
More InformationYou are currently viewing a placeholder content from Turnstile. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More InformationYou are currently viewing a placeholder content from Vimeo. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More InformationYou are currently viewing a placeholder content from YouTube. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More InformationYou need to load content from reCAPTCHA to submit the form. Please note that doing so will share data with third-party providers.
More InformationYou need to load content from hCaptcha to submit the form. Please note that doing so will share data with third-party providers.
More InformationYou need to load content from reCAPTCHA to submit the form. Please note that doing so will share data with third-party providers.
More InformationYou need to load content from Turnstile to submit the form. Please note that doing so will share data with third-party providers.
More InformationYou need to load content from hCaptcha to submit the form. Please note that doing so will share data with third-party providers.
More InformationYou need to load content from reCAPTCHA to submit the form. Please note that doing so will share data with third-party providers.
More InformationYou are currently viewing a placeholder content from Turnstile. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More Information
