- PC makers including HP and Dell are struggling to secure PCBs for 2027 as AI servers absorb board capacity
- Canon says AI-driven PCB shortages have doubled its expected medical sales hit to about ¥90 billion
- One supplier cites some orders taking up to 30 months even as Unimicron’s chairman used the same figure for equipment, not boards
The AI infrastructure boom has already chewed through the world’s supply of high-end memory, and it is now working its way down to the green slab everything else sits on. PC makers planning their 2027 lineups,
HP and Dell, among others, are struggling to secure enough printed circuit boards, according to a Nikkei Asia reportsummarized by Guru3D.
One board maker told the publication that lead times on a significant number of its products have doubled, with some orders stretching to 30 months, while another said an order placed today would not be filled until late 2028.
The board that nobody worried about could be a significant problem
The soaring demand for PCBs is not happening in a vacuum; Samsung has already warned that the memory shortage could last until 2028, and the same logic essentially applies to boards: the companies building AI data centers need more of the product, need better versions of it, and will pay more for it.
The crux of the matter is that AI server boards are a different beast from the motherboard in a mid-range laptop. Würth Elektronik estimates that PCBs for AI servers carry three to five times as many layers as conventional designs, built on low-loss laminates and specialty glass fabrics, so each one essentially consumes far more material and factory time.
Board makers have responded rationally, aiming to maximize profits: Market researcher in4ma reported in July that some PCB manufacturers have told existing customers they will stop producing standard boards in favor of higher-margin AI server work. Hyperscalers are also buying their way into the supply chain directly, with the world’s biggest hyperscaler, Amazon, taking a stake of roughly NT$1.59 billion ($48.2 million in USD) in Taiwan’s Gold Circuit Electronics in September, according to TechSpot.
That leaves PC makers in a particularly odd and unenviable position, with an unnamed executive at an American PC maker telling Nikkei that even with shipments set to fall again next year, “we are still struggling to secure enough PCB supplies to meet our needs.”
IDC says global PC shipments fell 20.1% year on year in the third quarter to 62.7 million units, with HP down 30.9% and Dell down 25%, although IDC attributes part of that to distributors working through stock bought early in the year ahead of price hikes. Either way, this is a buyer who wants fewer boards and still cannot get them, which points to an allocation problem rather than demand. It is the same dynamic that has data centers on course to grab 70% of high-end memory chips this year.
The 30-month figure is doing the heavy lifting in headlines built around Nikkei’s coverage, and it centers around a single unnamed supplier describing some orders, not an industry average.
The trend around it is striking enough on its own – in March 2026, German board maker ILFA said lead times for standard FR-4 laminate had gone from a few days to around four weeks. By April, a Broadcom director was telling Nikkei that PCB lead times had gone from six weeks to six months. This effectively means that an order placed in late October that lands in late 2028 implies a wait of roughly two years, while a 30-month lead time from today would run to around April 2029, so the two supplier claims describe different ends of the range.
A week before Nikkei’s report, Unimicron chairman S.C. Chien told the TPCA Show in Taipei that lead times for some key production equipment had reached 30 months, with more than 80% of equipment seeing longer waits and over a fifth seeing waits more than double their usual length. Chien also warned that low-CTE glass cloth, copper-clad laminate (CCL) and ABF could fall about 20% short of demand in 2027.
Nikkei also reports that one leading laminate supplier has orders filling through 2030, despite plans to triple capacity by 2028. An order book, however, is not a delivery schedule, and suppliers benefit from scarcity narratives; Acer’s Jason Chen has already accused memory makers of hyping 2030 shortage talk to protect their margins. The upstream bottlenecks are real, though. Nittobo controls roughly 90% of the T-glass used in advanced substrates, and its capacity expansion will take years to reach the market.
One could argue that there is no case here for panic-buying a motherboard just yet. A board is a small slice of a laptop’s bill of materials; no PC maker has put a number on the retail impact, and memory remains the dominant cost driver
But the PCB squeeze adds to an already grim picture, in which major vendors had warned of 15-20% price rises starting inthe second half of 2026. The large OEMs with long-term contracts should weather it best thanks to already negotiated and inked supply commitments.
Smaller motherboard and graphics card brands, buying without that leverage, are where a doubled lead time or a tripled laminate bill is most likely to surface in 2027, either as a higher sticker price or a thinner spec sheet if not both at the same time.
