A car circuit board that shipped in two to three weeks now takes twenty. The bottleneck is not the board shop, and it is not China, which produces more than 57 per cent of the world’s PCB output. It is a machine made by one Japanese company.
The numbers came out of an Aumovio ecosystem conference in China, where the company and its customers and suppliers spent a session on the shortage. Aumovio’s China chief operating officer said automotive PCB lead times have stretched from two to three weeks to an average of twenty, and that the delay sits with copper clad laminate and prepreg rather than with board capacity.
An inverted pyramid with one supplier at the point
The chain narrows at every level. At the top sit 2,000 to 3,000 PCB factories worldwide, most of them in China. Below them around 20 copper clad laminate makers, with eight holding the volume. Below that around 15 glass fabric producers, six of them material. At the bottom sit the air jet looms that weave high-end electronic cloth, and that equipment is almost entirely supplied by one Japanese manufacturer, Toyota Industries, which holds more than 90 per cent of the segment.
Its JAT910 series is the accepted tool for stable volume production of ultra-thin cloth at 1080 and below, with close to 95 per cent share in that grade, an 18 to 24 month delivery cycle, an order book running to 2030 and a halt on new high-end loom orders from May 2026.
AI changed the mix, and the mix starved the car
Thick cloth such as 7628 used to be about 70 per cent of the mix, serving automotive, industrial control and consumer boards, while ultra-thin grades at 1037 and 1027 were 15 to 20 per cent. AI server boards at high layer counts use almost nothing else, so capacity moved and automotive thick cloth was squeezed out. Copper foil moved the same way, from standard to RTF and on to the HVLP low-profile grades servers need.
Min Sheng Securities puts global HVLP4 capacity at about 700 tonnes a month against 2025 demand of 850 tonnes, a gap above 40 per cent, with 2026 demand forecast to pass 3,000 tonnes against effective capacity near 1,300. Prismark puts 2025 PCB output at USD 85.8 billion, up 16.7 per cent, with server and data storage boards up 46.9 per cent and automotive up 5.7 per cent. Automotive board growth is forecast at about 3.3 per cent a year to 2030.
Automotive demand cannot compete with that curve, and it cannot flex either. A board is designed for a specific model, safety parts carry a quality red line, and content per vehicle keeps rising from roughly 0.5 to 1 square metre and a few hundred renminbi on a combustion car to 3 to 5 square metres and several thousand on an electric one.
Why inventory does not solve this
Stockpiling fails on three counts. Materials are scarce enough that there is no spare capacity to fill a warehouse. Prepreg has a shelf life, in some grades about three months, and expired material is a reliability issue on a safety part. And boards are model-specific, so a build plan that slips turns stock into scrap.
The response is multi-sourcing with a five-step qualification path, traceability from board back to laminate, and visibility down to fourth and fifth tier suppliers instead of the first and second. Suppliers report AI customers booking 2027 and 2028 capacity in advance, which pushes automotive orders further back. The industry consensus is that 2027 is the hardest year for automotive supply security and that real relief arrives in 2028.
A glass furnace costs at least RMB 2 billion, runs at 1,400 degrees continuously once lit, and locks its product mix for roughly a decade. That is why upstream capacity does not respond to price the way downstream capacity does, and why the shortage will outlast the order book that caused it.
Editor’s note: This is an adapted translation of the original Gasgoo report. It has been trimmed and restructured for readability for an international business audience. The full original (in Chinese) is at https://auto.gasgoo.com/news/202611/11I70477503C108.shtml.