Where Mid-Volume PCB Orders Go When High-End Capacity Is Full

On 7 August 2026, financial media reported that Yibo Technology had recorded a year-on-year increase of more than seventy percent in signed sales orders, with the growth rate rising month by month. The fastest-growing categories were automated test equipment, optical modules, and robotics. Within PCB fabrication, certain bottleneck processes had reached full capacity and order backlogs were forming. Optical module PCBA had entered volume production with monthly shipments in the hundreds of thousands of units, and semiconductor-related business was described as one of the fastest-growing directions, with its revenue share potentially rising above twenty percent.

The most useful signal in that report is not the growth rate. It is the phrase about bottleneck processes, because it describes what happens to a market when demand concentrates in a small number of technically demanding steps rather than spreading evenly.

What an Order Backlog Actually Signals

A backlog in a specific process step rather than across an entire factory is a distinctive condition. It means the factory has capable equipment in general, but one or two operations have become the rate limiter for everything passing through them.

In high-end board manufacturing, those steps are predictable. Laser drilling capacity limits high-density interconnect work. Lamination presses limit high-layer-count boards, because each additional layer pair consumes a press cycle. Back drilling and high-aspect-ratio drilling limit thick boards with high-speed channels. Impedance-controlled etching and the inspection capacity to verify it limit any design with tight tolerances.

Each of these steps is difficult to expand quickly. Equipment has long lead times, and a new machine is not capacity on the day it is installed: it must be commissioned, ramped to yield, and approved by customers first. During that period, the backlog persists regardless of how much floor space the factory has.

That is why a backlog at the bottleneck step is a more informative indicator than a headline capacity figure. It tells you where the market’s technical demand has concentrated, and it tells you which programmes will be affected first: those whose designs require the constrained step, and those whose volumes are too small to compete with large accounts for the limited slots.

The Demand Path From AI Computing to Test Equipment

The composition of the reported growth is worth reading closely, because it shows how AI infrastructure demand propagates beyond servers.

Accelerator capacity expansion directly increases demand for GPU server boards. But it also increases demand for the equipment that tests those devices. Automated test equipment requires high-pin-count interface boards carrying dense, impedance-controlled signal paths, often at very fine pitch, and those boards are difficult to produce. When chip production rises, test capacity must rise with it.

Optical modules follow the same logic. As accelerators multiply, data must move between them at higher rates, which pulls optical transceivers toward 800G and 1.6T and increases both the number and the complexity of the boards inside each module. High-speed serializer-deserializer channels, radio frequency paths, and optical conversion circuitry impose strict requirements on material loss, layer structure, via stub, and differential impedance, in a module whose internal space is extremely limited. The board therefore shifts from a conventional multilayer design toward a high-density, high-speed structure.

AI demand therefore does not stay inside the data centre. It moves into the equipment that builds and tests computing hardware, the interconnect that carries data between systems, and the automation that manufactures it. Each needs a different board family, and each consumes the same constrained process steps.

Why Effective Capacity Is the Real Question

The industry has spent the past two years recalibrating what capacity means. The metric that matters now is not how many square metres a factory can nominally process, but how much high-end output it can genuinely deliver.

Several factors separate nominal from effective capacity. Material availability is one: a factory cannot produce M8 or M9 class boards without a qualified supply of that laminate, regardless of its press count. Process maturity is another: a line that has never run a particular material grade requires development before it can produce it to specification. Qualification is a third: customers approve specific factories for specific products, and that approval cannot be transferred to a new line without testing.

The practical result is a market that can look simultaneously over-supplied and under-supplied. There is ample capacity for conventional boards, and persistent tightness for high-speed, high-layer-count and high-density products. Programmes whose designs sit in the constrained region experience the market as a shortage; programmes whose designs sit outside it may see competitive quotations and short lead times.

This is also why technical demand has become layered. Optical modules, automated test equipment, and some high-speed control products primarily pull demand for boards in the eight to sixteen layer range and above. AI servers, switches, and high-end test systems pull the sixteen to seventy-eight layer range. Within those ranges, high-density interconnect, any-layer structures, and fine-line processes producing traces at 0.075 mm and below become progressively more important, and differential impedance control tightens toward plus or minus five percent on demanding designs.

At the same time, higher speed and higher power are arriving together. High-speed interfaces need tightly controlled impedance, while power modules may require heavy copper for current carrying and heat dissipation. Robotics, automotive, and low-altitude applications add flexible and rigid-flex interconnects. The result is that advanced products are no longer described by a single attribute such as high layer count. They are combinations: high layer count plus high-density interconnect plus high-speed material plus flexible interconnect plus power design. Producing that combination requires capability across several process families rather than excellence in one. Confirming the combination is available, rather than assuming that a supplier who does one thing well can do all of them, is precisely what a capability review establishes.

Where Mid-Volume Orders Go

When leading manufacturers prioritise large customers and large volume programmes, mid-volume, high-complexity, and fast-iteration work finds different routes.

This is not a simple transfer of volume from one factory to another. It is a re-sorting of suppliers against three criteria that matter for iterative programmes: technical fit, delivery speed, and quality stability.

Technical fit. A programme that iterates frequently needs a supplier who can build the design’s difficult features at prototype quantity without treating the order as an interruption. That means the microvia size, line width, layer count, and impedance tolerance the design requires must be routine for the supplier’s process, not experimental. Where a prototype cannot be built to the intended geometry, the measurements taken from it describe the limitation of the fabrication rather than the behaviour of the design, and the iteration produces no useful information.

Delivery speed. An iterative programme lives or dies by cycle time. Each revision requires fabrication, assembly, and measurement before the next decision can be made, so engineering time lost between stages compounds. Keeping board fabrication and assembly with one supplier removes the information transfer across a company boundary that otherwise consumes days per revision.

Quality stability. The point of prototyping is to produce a design that can be manufactured, not merely one that works. A prototype built on a process that cannot be repeated at volume teaches the wrong lesson. Suppliers that report capability as measured distributions rather than single results, and that maintain process control on plating, registration, and impedance, provide more useful information at the prototype stage.

Selecting a Supplier for Iterative Programmes

For engineering teams, several practical considerations follow from this environment.

Match the supplier’s strength to the design’s constraint. A programme whose difficulty is density needs high-density interconnect and fine-line capability. A programme whose difficulty is current needs heavy copper and thermal design. A programme whose difficulty is signal integrity needs low-loss material and tight impedance control. Engaging a supplier who is strong in the specific constraint avoids the most common source of prototype disappointment.

Confirm material and stackup before layout is frozen. The material grade a design assumes determines both performance and availability. Verifying that the grade is qualified and in supply, and that the stackup is producible to the required tolerance, converts an assumption into a design input. Where the loss budget permits, hybrid stackups that use advanced laminate selectively reduce cost and material risk at the same time.

Plan the transition to volume early. The most expensive mistake in an iterative programme is discovering at ramp that the prototype process is not the volume process. A supplier that can carry a design from rapid prototyping through low-volume assembly and into larger volume under one quality system preserves the process knowledge accumulated during development instead of discarding it.

Design the test flow alongside the board. Dense, high-speed designs place fine-pitch packages in small areas, and the joints that matter most are frequently hidden. Paste inspection, optical inspection, X-ray, and electrical test cover different failure modes, and traceability links the results to individual units. A structured test strategy is what allows a prototype result to be trusted and a production population to be characterised.

The broader conclusion is that the PCB market has segmented along technical lines. Conventional capacity is plentiful; effective high-end capacity is not. The strategic response is not to chase the largest supplier, but to match the design’s hardest requirement to a supplier who treats it as routine.

Frequently Asked Questions

Why do backlogs form in specific processes rather than whole factories? Because a few operations, such as laser drilling, lamination pressing, and back drilling, limit throughput for high-end work. Adding capacity at those steps requires long equipment lead times plus process development and customer qualification.

How does AI demand reach test equipment and optical modules? More accelerators require more test hardware, and higher data rates require faster optical interconnect. Both are PCB-intensive products with demanding electrical requirements, so AI growth pulls several board families at once.

What is effective PCB capacity? The output a factory can genuinely deliver in high-end products, accounting for qualified material supply, process maturity, and customer approval, rather than nominal square metres of capacity.

Why do mid-volume orders get re-routed when capacity is tight? Large customers get priority on constrained processes, so mid-volume and fast-iteration programmes move to suppliers with the right technical fit, delivery speed and quality stability.

What should an engineering team verify in a supplier? That the design’s hardest requirement, whether density, current, or signal integrity, is routine for the supplier’s process, that material and stackup are confirmed before layout, and that fabrication, assembly and test can be coordinated to keep iteration cycles short.