AI servers are crowding out PCB capacity, with delivery times for some orders extended to 30 months

According to a report by Nikkei Asia, PC manufacturers such as HP and Dell have already begun to worry about whether they will be able to secure enough circuit board supply when formulating their 2027 product plans. A manager at a U.S. PC manufacturer said that even though global PC shipments are expected to possibly continue declining next year, the company is still finding it difficult to obtain enough PCBs to meet production needs.

This situation is quite abnormal. According to traditional market rules, falling PC demand usually means that pressure on component supply will ease accordingly, but the current situation is exactly the opposite. AI server construction is absorbing large amounts of global PCB manufacturing resources, making it impossible for consumer electronics manufacturers to ensure sufficient production components even in the face of weak market demand.

The core of the problem lies in the major differences between the circuit boards used in AI servers and ordinary PC motherboards.

Ordinary desktop motherboards usually use relatively conventional multilayer PCB structures, while high-end AI servers need to carry large numbers of GPUs, CPUs, memory, and high-speed interconnect devices, placing higher demands on signal integrity, heat dissipation capability, and electrical performance of circuit boards. These circuit boards often use structures with dozens of layers and use low-loss copper-clad laminates and special glass fiber materials to ensure signal stability during high-speed data transmission.

Because manufacturing is more difficult and unit prices are more substantial, PCB manufacturers are naturally more willing to allocate production lines to AI server customers. For manufacturers, the equipment time and material resources occupied by producing one high-end AI server circuit board may be equivalent to multiple ordinary consumer-grade circuit boards, and AI infrastructure customers usually also have stronger purchasing power and willingness to pay.

As a result, some of the production capacity originally used to produce desktop motherboards, laptop motherboards, and graphics card PCBs is gradually being shifted to the AI server sector.

A more serious problem comes from the most basic raw material of PCBs—copper clad laminate (CCL).

Copper clad laminate is the core foundational material for manufacturing printed circuit boards, usually made by pressing together glass fiber cloth, resin, and copper foil. Manufacturers then process circuit traces on its surface and form a complete PCB structure through multilayer stacking.

A Taiwanese PC manufacturer told Nikkei Asia that CCL prices have now reached at least 12 times the July level. This increase reflects the rapidly worsening tightness of upstream material supply. However, this is market feedback provided by a single company and does not mean that CCL of all specifications has seen increases of the same magnitude.

At the same time, orders at one of the world's major CCL suppliers are already booked through 2030. Even if the company plans to expand capacity to three times the current level by 2028, the existing order backlog remains very large.

This means the problem is no longer limited to whether PCB manufacturers are willing to expand production, but whether upstream material suppliers can provide enough raw materials in time.

PCB manufacturers are also continuing to extend delivery lead times. One supplier said that delivery times for a considerable portion of its products have doubled; another company disclosed that if a new order is submitted now, delivery is expected only in the second half of 2028.

As a result, lead times for some orders have reached 30 months. For PC manufacturers that need to plan product development and production cycles in advance, this situation will seriously affect product supply arrangements for 2027 and even 2028.

Expanding capacity is also not an easy task. In addition to needing to purchase equipment, build factories, and ensure raw material supply, the PCB industry also faces a serious shortage of professional technical personnel.

As manufacturers expand investment one after another, experienced PCB engineers and production technicians have become targets of competition. According to reports, salaries for technical positions related to the PCB industry have risen by 20% to 30%. This further increases the cost of corporate expansion and also means that even if factories are built, they may not be able to immediately reach expected capacity.

The current PCB supply tightness has already begun to affect other industries.

Canon of Japan said this week that, affected by the shortage of circuit boards and related components driven by the AI industry, some of its production activities have been constrained since August this year. Because medical equipment production has been affected, Canon expects to lose about 90 billion yen in medical equipment business sales this year as a result.

This shows that the impact of AI infrastructure construction on traditional electronics industries is no longer limited to PCs and graphics cards; medical equipment, industrial electronics, and other products that rely on complex circuit boards may likewise be affected.

At the same time, the PCB shortage is combining with the earlier memory crisis.

Since 2026, DRAM and NAND flash memory prices have continued to rise, and PC manufacturers have already had to reassess product configurations and pricing strategies. Market research firm Gartner even previously predicted that by 2028, PCs priced below $500 may gradually disappear, because rising costs of memory and other key components will make low-priced computers increasingly difficult to sustain in production.

The graphics card market has not been spared either. Taking Nvidia's GeForce RTX 5080 as an example, retail prices in some U.S. channels have already approached $2,000, far above its initial official suggested retail price.

Now, PCBs have also joined the ranks of price increases and supply tightness, meaning that the cost pressure facing the PC industry is further expanding.

It is worth noting that although PCBs do not account for a prominent share of the overall material cost of a computer, they are irreplaceable foundational components. Whether motherboards, graphics cards, or laptops, all must rely on circuit boards to connect chips, memory, power modules, and other components. Once PCB supply is insufficient, manufacturers cannot complete whole-machine production even if they have already obtained CPUs, GPUs, and memory.

Nvidia has already begun to lock in advanced packaging and chip substrate capacity in advance by establishing cooperative relationships with professional substrate manufacturers. Japan's Ibiden, South Korea's Samsung Electro-Mechanics, and Taiwan's Unimicron have all entered the relevant cooperation system. Among them, Unimicron plans to expand related capacity to 2.5 times the current level within the next three years.

However, although chip packaging substrates and ordinary PCBs are technically related, they cannot fully substitute for each other. The growing demand for advanced substrates from AI chips is also further intensifying resource competition across the entire circuit board industry chain.

For the 2027 PC market, the current situation is especially worth watching. Even if consumer demand remains weak, PC manufacturers may not necessarily obtain lower procurement costs as a result. On the contrary, strong demand from AI server customers for high-end PCBs may continue to squeeze production resources for traditional consumer electronics products.

For some time to come, motherboards, graphics cards, and laptops may all be affected by rising PCB costs and extended delivery lead times. Although no PC manufacturer has yet announced specific retail price adjustment ranges, with memory, SSDs, and GPUs already having broadly increased in price, the PCB supply crisis will undoubtedly further weaken the room for price cuts in consumer electronics products.

From an industry perspective, this crisis reveals another side of AI infrastructure expansion: artificial intelligence is not only competing for the most advanced GPUs, HBM, and advanced-process wafers, but is also consuming large amounts of basic electronic manufacturing resources that the market had not previously paid special attention to.

As global capital and manufacturing capacity continue to tilt toward AI data centers, the traditional PC industry, even with mature product design and sales channels, may be unable to complete production as planned because of a single ordinary circuit board.