Gasgoo Munich-The memory chip squeeze isn't over — and now, high-end automotive PCBs (printed circuit boards) are running short too.
Orders that once took two or three weeks to deliver now face a 20-week wait. Price hikes are cascading from copper-clad laminates (CCL) and fiberglass cloth down to the most obscure consumables. China accounts for over 57% of global PCB output by value, yet its Achilles' heel may rest in a single weaving machine further upstream.
What's driving this shortage isn't the ebb and flow of car demand — it's the explosive growth in AI compute. With unprecedented force, AI is wresting materials, capacity, and equipment away from the auto industry.
At the recent AUMOVIO Ecosystem Conference, AUMOVIO joined its clients and suppliers for an in-depth discussion on this supply-chain crunch.
From a 3-Week to a 20-Week Wait: The Bottleneck Isn't at the Board Maker, It's Further Upstream
The shift shows up first in lead times.
"Automotive PCB lead times used to run two to three weeks — now they average 20," the AUMOVIO China COO admitted during a roundtable, noting that delivery cycles have stretched severalfold.
He stressed that the real drag on deliveries isn't board-making capacity, but upstream materials like copper-clad laminates (CCL) and prepreg (PP).
Price increases are starting upstream, too. Since the fourth quarter of 2025, hikes have rippled from CCL and fiberglass cloth down to drill bits and cover plates, an insider at a PCB maker revealed. "Every layer of the PCB is under cost pressure."
According to local reports, leading CCL producers have raised processing fees for laminates, prepreg, and copper foil multiple times since 2026.
AI-generated image; Source: Doubao
Why is China's vast PCB industry getting choked by upstream materials?
According to that PCB insider, the supply chain resembles an inverted pyramid, narrowing at every tier.
At the top sits a highly fragmented PCB sector — roughly 2,000 to 3,000 factories worldwide, most concentrated in China. One step down, the CCL market has about 20 global players and eight dominant suppliers. Further down, fiberglass cloth has around 15 global players and six main producers. At the very base, the jet looms essential for high-end electronic cloth depend almost entirely on a single Japanese company.
Public data shows that global capacity for high-end electronic cloth — especially ultra-thin varieties — hinges on premium jet looms supplied by a handful of makers, led by Toyota. The Japanese giant commands over 90% of this equipment segment.
According to Shanghai Securities News and Cailian Press, Toyota's JAT910 series is the industry standard for stable mass production of 1080 and finer ultra-thin cloth, holding nearly 95% of that niche. Lead times stretch 18 to 24 months, the current order backlog extends to 2030, and Toyota stopped accepting new orders for high-end looms in May 2026.
AI, meanwhile, is rewriting the structure of upstream materials.
A top PCB executive explained that thick cloth — such as type 7628 — once dominated, supplying auto, industrial, and consumer boards and accounting for about 70% of the mix. Ultra-thin variants like 1037 and 1027 saw limited use, making up just 15% to 20%.
Over the past two years, the high-layer-count boards demanded by AI servers have shifted almost exclusively to ultra-thin cloth, upending that mix.
As demand surges, upstream material suppliers are pivoting capacity toward ultra-thin cloth — squeezing out the thicker grades needed for automotive boards.
Copper foil is climbing the value ladder in tandem, shifting from standard foil to reverse-treated foil (RTF), and then to the low-profile foil (HVLP) that AI servers require.
Just how severe is this squeeze? A few numbers tell the story.
Minsheng Securities estimates that global monthly capacity for HVLP4 high-frequency, high-speed copper foil — a material AI servers require — stands at just 700 tons. Monthly demand in 2025 has already hit 850 tons, leaving a shortfall exceeding 40%. The brokerage predicted in November 2025 that as 1.6T optical modules scale up, monthly demand in 2026 could top 3,000 tons, against effective capacity of just 1,300 tons.
Forecasts for the 2026 gap vary across agencies, but all agree on the direction: demand will outstrip supply, with shortfalls topping 20%.
Steady automotive demand is no match for the steep, hockey-stick curve of AI.
Data from Prismark backs this up. Global PCB output value hit an estimated $85.8 billion in 2025, up 16.7% year-on-year. Within that, PCBs for servers and data storage surged roughly 46.9%, while automotive boards edged up just 5.7%. Looking ahead to 2030, automotive boards are forecast to grow at a compound rate of just 3.3% (based on March 2026 estimates, with slight variations across quarters).
"AI is explosive and cannibalistic; automotive is steady and predictable," the PCB executive noted. "The steeper curve will always capture the premium resources."
Making matters trickier, automotive boards offer little room for flexible reallocation. They are highly customized — each board maps to a specific model and can't simply be swapped — and safety-critical parts leave zero margin for error. Meanwhile, electrification and smarter cockpits keep pushing up PCB content per vehicle, from roughly 0.5 to 1 square meter worth a few hundred yuan in combustion cars, to 3 to 5 square meters worth thousands in new-energy vehicles.
Demand is climbing, yet capacity is being diverted. That is the essence of a structural shortage.
From Firefighting to Deep Visibility: Seeing Risk Through the Fifth Tier
Viewed this way, the shortage won't be solved by simply expanding capacity in the near term. The supply chain's playbook is shifting, from reactive firefighting to proactively building technical resilience.
Step one: put the eggs in different baskets.
The AUMOVIO China COO said the company has made multi-sourcing the cornerstone of its procurement strategy, requiring as many critical part numbers as possible to have at least two qualified suppliers. It's also continuously broadening the supplier pool at both the board-fabricator and CCL levels.
The PCB insider noted that first-tier suppliers have largely pivoted to AI, and even second-tier players may be 50% to 70% allocated to AI. Buyers can dig further down to the third tier, where some suppliers are already undergoing validation.
"In the future, a single part number may need two swappable materials," the insider observed. Board makers can't pull this off alone — they have to loop in both customers and suppliers.
Yet "adding a backup" is far more than just dropping a new material code into the system.
Building a swappable backup system requires at least five steps. During material selection, only candidates meeting process parameters enter the pool. In validation, parts must clear dual hurdles at both the board maker and the customer. From there, the full chain — from PCB to CCL — must be traceable, and mass-production milestones must be pulled forward. Finally, suppliers, board makers, and customers must collaborate openly, doing the legwork well in advance.
Image source: Jingwang Electronics official website
Could companies buffer the blow by stockpiling inventory, as they might during other shortages? The insider said no.
First, materials are currently too tight to even meet immediate demand, leaving no spare capacity for stockpiling.
Second, materials like prepreg have strict storage conditions and shelf lives. Some prepreg expires in just three months; using it past that date directly compromises vehicle reliability — an untouchable red line.
Third, automotive boards are model-specific. If a vehicle is redesigned or forecasts miss the mark, hoarded boards become dead stock — unusable elsewhere, tying up capital and warehouse space.
This aligns with a broader rethink in global supply-chain management: the pure, cost-optimized just-in-time (JIT) model is giving way to a "JIT plus just-in-case" hybrid. Companies are holding moderate strategic buffers for critical, long-lead-time items, but their true confidence rests on the ability to switch fast.
Step two: push the risk radar deeper into the supply chain.
"In the past, visibility into Tier 1 and Tier 2 suppliers was considered good enough; now you have to see all the way to Tiers 4 and 5," a senior consultant noted. This shortage, he argued, has exposed more than just competition for materials and capacity — it has revealed a blind spot to "outsiders." By staring only at the existing supply base, companies struggle to notice new entrants siphoning resources from the outside.
Step three: rebuild trust and collaboration amid the volatility.
Automakers can't offer perfectly stable forecasts, but they can be transparent. A procurement chief at one automaker suggested bringing OEMs, Tier 1s, and all the way down to Tier 5s into a single communication loop to share real demand. He also advocated using big data and AI tools to sharpen predictions: "When humans judge the future based on limited information, the margin of error is large. AI, drawing on a larger knowledge base, might find more solutions."
Collaboration also means locking in future capacity well ahead of time.
A supplier executive noted that AI customers often lock in orders for 2027 or even 2028, forcing upstream capacity to prioritize them. To counter this, the automotive supply chain must work with strategic customers early to map out next year's real demand, relay it up the chain layer by layer, and ask for customer support in fast-tracking alternative suppliers.
"What you need on this chain are long-termists who share your values," he said. Automotive boards may be less lucrative than AI products in the short run, but the auto industry is a stable, decades-long business. Whether a company can bind partners who remain committed to the automotive strategy and share the burden of volatility will determine whether it survives the grueling supply test of 2027.
From Surviving to Thriving: Commercial Resilience Is a Longer Test
Technical resilience solves for survival. But to thrive over the long haul, a harder, deeper layer of support is needed.
First, face reality: upstream expansion runs on its own clock. The fabled "China speed" doesn't fully apply at the very top of the chain.
The PCB executive noted that building a commercial-scale fiberglass furnace requires at least 2 billion yuan. Once fired up, the furnace must run continuously at 1,400 degrees Celsius — and once the product mix is set, it can barely be changed for a decade.
This is the industry's "earn for a year, nurse for three" logic, which makes upstream firms naturally cautious about expanding. Add the fact that advanced copper-foil post-processing lines and core looms for ultra-thin cloth remain concentrated in Japanese hands, and true relief on the materials front is still some way off.
The supplier executive predicted that 2027 will be the most brutal year for automotive supply security, with substantive relief from capacity bottlenecks likely holding off until 2028.
A dose of caution: the shortfall won't last forever. For one thing, AI data-center demand is cyclical too; if that pullback comes, siphoned capacity could well flow back. For another, an expansion wave in domestic fiberglass, copper foil, and electronic cloth is already underway, flashing warnings of structural overheating. For the auto industry, today's battle is over availability; tomorrow's may be over price — and whether to keep locking in.
Image source: Fulltech official website
Yet within the gap, domestic substitution is gaining pace.
Public reports show that domestic firms have made successive breakthroughs in premium CCL, HVLP copper foil, and high-end electronic cloth. In looms, Chinese equipment has carved out some substitution for standard electronic cloth, but machines for high-end specialty cloth remain unavailable or stuck in validation. With top production lines still dominated by imported equipment, domestic machines still lag noticeably in yield, efficiency, and reliability testing.
Industry insiders expect domestic materials to gradually take the baton in the second half of 2027 as new furnaces and copper-foil capacity come online. Whether they hit that mark, however, hinges on the pace of expansion and automotive-grade certification. For safety-critical auto parts, certification and validation cannot be rushed — that is the baseline technical resilience must hold.
Beyond technical substitution lies a more fundamental concept: commercial resilience.
The consultant explained that product certification, backup alternatives, and rapid switching constitute technical resilience — they solve the short-term question of whether parts can be supplied at all. Securing supply over the medium to long term requires answering three commercial questions.
First, if upstream players invest in heavy-asset expansion, can they secure a clear and stable return? If downstream customers suddenly walk away once the capacity is built, who bears the risk?
Second, resilience comes at a cost. The expenses of multiple certifications, strategic inventories, and redundant capacity — how should they be shared across the supply chain?
Third, how should risk be shared along the chain? From the OEM at the bottom to suppliers at every tier, can the industry build a mechanism for collective burden-sharing?
"Technical resilience will keep us alive in 2026 and 2027; commercial resilience will let us go the distance."
Intriguingly, the very shock triggered by AI may ultimately find some of its solutions in AI as well.
The consultant divides enterprise AI adoption into three stages: efficiency, enhanced utility, and scale. The progression runs from boosting efficiency in small scenarios, to embedding AI into production scheduling, engineering, and financial outcomes, and finally to becoming an AI-native organization.
According to public industry use cases, large language models can already simulate factors like typhoon weather, upstream supply disruptions, and shipping delays — issuing early warnings and recommending production adjustments. Digital twins and supply-chain control towers are making it possible to close the loop between risk identification and strategy generation.
Take a longer view, and the relationship between cars and AI is about more than just competing for resources. The vehicle itself is a highly intelligent platform — and likely a crucial prototype for embodied AI. Its demand for chips and AI will only keep climbing.
The story of the inverted pyramid is, at its core, a story about certainty.
As global industries race to chase AI's high returns, the auto sector must lean on long-termism, transparent collaboration, and commercial consensus to lock in a stable upstream outlook.
There is no single winner in this resilience rebuild. As AUMOVIO stressed during its ecosystem conference roundtable, only by "walking together as a group" can the steady automotive curve hold its ground against the steep AI tide.
