The AI boom has made circuit boards and laminate scarce and expensive. Inside the PCB supply chain squeeze, why it reaches every buyer, and how to stay ahead......
AI has turned the printed circuit board, and the laminate beneath it, into one of the most contested products in electronics. Here is what is driving it, why it reaches every buyer, and how to stay ahead.
For most of the last decade the printed circuit board was the part of an electronics project you worried about least. It was available, it was predictable, and its price moved gently. That world has gone. In 2026 the humble PCB, and the copper clad laminate it is built on, has become one of the most fought over products in the entire electronics supply chain. If you buy boards today you are effectively standing in the same queue as Amazon, Google and Microsoft, and they have very deep pockets.
This is not a temporary disruption that will pass with the next quarter. It is a structural reset driven by the largest infrastructure build-out in the history of computing. Understanding why it is happening matters, because the businesses that grasp it, and adjust how they buy, will protect their supply while others are left waiting.
How we got here: the AI demand shock
The single force behind all of this is artificial intelligence, or more precisely the physical infrastructure required to run it. The hyperscalers are constructing data centres at a scale the electronics industry has never seen. Combined capital expenditure across the largest operators is on track for somewhere in the region of 700 billion dollars in 2026, roughly six times what it was only a few years earlier, and new build-outs are planned to double again for 2027.
That money flows directly into hardware: GPUs and AI accelerators, advanced networking, power and cooling, and the high specification boards and substrates that tie it all together. Demand for AI and server PCBs is running more than 30 per cent higher year on year, while most other sectors sit in the low single digits. Crucially, AI hardware does not want ordinary boards. It needs high layer count multilayers, HDI, low loss and low dielectric materials, and high speed designs. Those are precisely the technologies that are hardest to make and slowest to scale, and demand for them has pushed factory utilisation to record levels.
It starts in the raw materials
To understand the squeeze you have to look beneath the board, at the materials it is built from. This is where the constraint really lives, and it is not one shortage but several arriving at once.
Copper clad laminate, the foundational substrate of almost every PCB, is the heart of the problem. Material makers have switched their priority to high speed AI grades and away from standard FR4, leaving the mainstream material starved. Import prices tell the story plainly. By early 2026 copper clad laminate prices had risen by around three quarters year on year and passed twenty thousand dollars a tonne for the first time since records began at the turn of the century. High end grades have climbed further still.
Copper foil, which is bonded to laminate to form the board, accounts for a large share of laminate cost. Foil makers are prioritising the hyper very low profile foil that high frequency AI boards require, which has created shortages in both the premium grades and the standard ones. Underneath that, the price of copper metal itself sits near all time highs.
Glass fibre cloth is the quiet bottleneck few outside the industry notice. Producers are converting standard glass capacity to the specialty low dielectric, low loss yarns that AI and high performance computing need, which squeezes both the advanced material and the standard material at the same time. The thinnest pre-pregs, the ones used for impedance control in dense, high layer AI stackups, are the tightest of all, because they rely on fine weave glass that is already constrained and on precision weaving that is hard to scale quickly.
Resin delivered the most dramatic shock of the year. A single petrochemical complex outage removed an estimated 70 per cent of the world's supply of the high purity resin used in many high frequency and RF laminates, almost overnight. That is not a price problem, it is an availability problem, and it shows how concentrated and fragile parts of this chain have become. Add in a tightening market for sulfuric acid, essential to board manufacturing, along with constraints on industrial gases and specialty metals, and the picture is of an entire material base under simultaneous strain.
It is worth correcting one popular misconception here. You may have heard that data centres are buying up gold and driving its record price. They are not, in any meaningful sense. Gold's surge is a story about central banks and safe haven demand, not circuit boards, and the amount used in electronics is tiny by comparison. Gold still matters to you, because it sits on board finishes and the high price raises finishing costs, but it is a parallel event rather than a cause. The metal data centres really are consuming at scale is copper. A single gigawatt data centre uses tens of thousands of tonnes of it, and a single AI server rack can contain close to two miles of copper cabling.
Why the PCB became the critical issue
The reason this reaches everyone, and not just buyers of cutting edge AI hardware, is the way capacity now cascades through the supply base.
The largest and most advanced factories are laser focused on the high margin AI work, and as they fill up they price out and push down the simpler products, which are then picked up by smaller factories that are themselves already running near capacity. This spillover, or waterfall, effect means that even a business ordering perfectly ordinary boards feels a shortage that was, in a sense, caused by AI.
Factory behaviour has shifted with it. We have reached a point where factories can turn away business without fear of losing revenue, and they are choosing which customers and technologies to support. The constraint has even spread to the equipment itself, with the top factories buying up the available laser drilling machines, making it difficult for others to add capacity at all.
For buyers, three changes matter most. First, lead times have stretched dramatically, with general PCB lead times pushing past twenty weeks and some laminate grades now quoted at up to six months, often on a quota basis. Second, there is what the trade has started calling the rug pull, where laminate prices move so frequently that a quotation can be overtaken before an order completes, and factories pay the current rate even on material ordered earlier. Third, and most fundamentally, the market has shifted from being about price to being about availability. Material is now allocated to whoever commits, not whoever negotiates hardest.
The data centres buying up the boards and the materials
The phrase you may have heard, that the giants are buying up supply, is true, and it is worth being precise about how.
The clearest template is the memory market, where the allocation order has become stark. The hyperscalers secure the large majority of the volume they request, major manufacturers get a smaller share, and everyone else competes for what is left. The same logic is now spreading into boards and laminate. The giants use their balance sheets to lock in supply years ahead through long term agreements, reserving production capacity at premium prices before smaller buyers can even quote. Some categories of advanced memory are already sold out for the year, and similar forward commitment is taking hold across the materials that feed AI hardware.
Because these buyers are effectively insensitive to price, the cost and the scarcity flow downhill. The effect has already reached consumers, with memory prices climbing steeply and device makers warning of higher prices and tighter availability. For smaller manufacturers and OEMs, the message is simple and slightly uncomfortable: in a market allocated by purchasing power, the businesses that plan and commit early are the ones that get served.
The numbers, at a glance
If you take only a handful of figures from all of this, take these. Copper clad laminate prices up by around three quarters year on year. PCB pricing up roughly 15 to 25 per cent against late 2025, with the steepest rises on high layer and high frequency work. Lead times stretching to six months on some laminate grades. And a single plant outage that removed about 70 per cent of the world's supply of a critical laminate material. None of these is a normal market movement. Together they describe a different market entirely.
A global build-out on a fragile base
Step back, and the wider picture explains why none of this is likely to ease soon. There are now well over ten thousand data centres operating worldwide, of which more than a thousand are the giant hyperscale facilities, and that hyperscale count has roughly doubled in five years. Global data centre capacity has doubled in around four years and is set to double again, with the market expected to grow from a little over 500 billion dollars in 2025 toward 740 billion by the end of the decade.
That demand sits on a base that is more concentrated and more exposed than most people realise. Key materials trace back to a small number of regions and, in some cases, a single plant. Logistics has its own pressures, with disruption in the Middle East tightening air freight capacity between Asia and Europe and pushing fuel and surcharge costs sharply higher. When demand at this scale lands on a base this fragile, volatility is the natural result.
There is a strategic irony worth noting too. Governments, the UK included, are actively courting data centre investment as a matter of national capability and sovereignty. Yet building the data centre is the part a country can control. The chips, the memory and the advanced materials inside them remain firmly offshore, made by a handful of producers in Asia and the United States. The compute may be increasingly local, but the supply chain that fills it is anything but, and that global dependence is exactly what makes events like a single resin outage felt everywhere at once.
What happens next
It is tempting to assume that any boom must be followed by a bust that brings relief. On the current evidence that is not the safe assumption to make. The conditions that would genuinely slow this down, a pullback in hyperscaler spending, a plateau in AI model scaling, a shift away from GPU heavy architectures, or a visible over-build, do not yet appear in mainstream forecasts. The honest debate in the industry is not whether demand will hold, but whether the build-out is running ahead of the returns that justify it. That debate is real and worth watching, but it does not change the near term reality on the ground.
For the next couple of years at least, the sensible planning assumption is continued tight supply, dynamic pricing, longer lead times, and a market in which allocation, not negotiation, decides who gets material. The annual pricing frameworks and stable lead time assumptions that the industry relied on for years are, for now, no longer viable. This is the tightening phase, not the recovery.
What this means for you, and how Texcel helps
None of this is comfortable, but it is navigable, and the businesses that adapt their buying behaviour will come through it in far better shape than those that wait for normal to return.
On our side, we are working the market hard on our customers' behalf. We stay in close, daily contact with our supply chain partners in the UK and offshore, tracking capacity and material trends as they move. We use our spend and our long standing relationships to secure material allocation and protect production slots, and we qualify approved alternative laminates so that a single constrained grade cannot hold up a build. Where it makes sense, we secure stock ahead of need, and we commit to flagging any change as early as we possibly can.
There are practical steps that make a real difference from your side too. The single most valuable one is to commit early. Sharing forecasts and placing firm, scheduled orders lets us lock in allocation ahead of the queue, while buyers who wait find themselves at the back of it. Where you can, consolidating into larger batches helps secure priority. Approving qualified alternative materials in advance keeps projects moving when a specific grade tightens. And responding quickly on quotations allows us to hold pricing before it moves again. In a market defined by volatility, communication and strong relationships are no longer a nicety. They are the strategy.
Talk to us
The PCB market has changed, and it is not changing back any time soon. But with early planning, honest communication and a partner that is close to the supply base, your supply does not have to suffer for it. If you would like to review your pipeline, secure your allocation, or simply talk through what this means for your next project, [get in touch with the Texcel team](#). We would far rather have that conversation now than after a shortage has already hit.
*The figures in this article are drawn from industry and market sources current as of mid 2026. The market is moving quickly and reported numbers vary by source, product and grade, so they are best read as well sourced indicators of direction and scale rather than fixed values.*
