AR Glasses at 100,000 Units: The PCBA Consistency Threshold

On 9 September 2026, AR glasses maker INMO announced the completion of its C3 funding round, bringing cumulative Series C financing to nearly one billion yuan. In March, its self-built smart glasses factory in Chenghua District, Chengdu entered production with an annual capacity of 100,000 units. Sales in the first half of 2026 already exceeded the whole of 2025, cumulative shipments approached 200,000 units by early June, and overseas revenue accounted for roughly 32 percent of the total.

The capacity figure is the number that changes the manufacturing conversation. AR glasses began as a development-stage product: hardware designs changed frequently, orders arrived as samples and small batches, and the most valuable thing a supplier could offer was fast engineering response. At 100,000 units a year, the requirement shifts. The product must not merely work for this batch; it must perform, look, and last the same way across all of them.

What Changes at 100,000 Units

In low-volume production, variation can be absorbed by inspection and screening. Boards are measured, marginal units are set aside, and the yield loss is a manageable line item. At scale, that approach stops working for two reasons.

First, inspection capacity does not scale linearly with volume, and full screening of every parameter becomes impractical. Second, the economics change. A defect rate that looks trivial in percentage terms becomes a concrete quantity of rework and delivery loss once tens of thousands of units are involved. A fraction of a percent on 100,000 units is hundreds of glasses, each requiring diagnosis, disassembly, and replacement of a component that may not be individually replaceable.

The parameters that matter at this stage are unglamorous: microvia quality, plating uniformity, line width consistency, placement offset, and solder joint quality. These are exactly the variables that small-batch production can filter out after the fact and that scale production must control in the process. For a supplier, the deliverable changes from demonstrated capability to repeatable capability, which is a different thing and requires process control rather than inspection alone.

The Paradox: Lighter Glasses, More Complex Interconnect

AR glasses differ fundamentally from phones. There is no single main board with everything mounted on it. The electronic system is distributed around the frame, the temples, and the optical module, and it all has to fit within a device whose weight is a primary user-facing specification.

The main processor, memory, wireless communication, power management, cameras, and display-related components must be connected within a very constrained volume. That architecture naturally favours HDI and FPC, and often rigid-flex solutions. HDI uses laser microvias to increase local routing density and avoid the board area consumed by conventional through holes. Flex circuits route along the temples and frame, connecting separate structural components that may move relative to one another.

The difficulty is not simply making the board smaller. As board area shrinks, the components still need power, data, and control connections, so traces and pads become more densely packed. Small variations in laser drilling, via filling plating, trace etching, and layer registration then translate directly into assembly yield problems. A via that is marginally off-target or a pad that is slightly under-etched becomes a defect that is invisible in a sample and systematic in production.

Weight and thickness constraints compound the problem. Thinner boards and thinner stackups are more sensitive to process variation because the same absolute tolerance represents a larger proportion of the feature size. A two-micron variation in dielectric thickness means more in a thin wearable stackup than in a thick server board, and its effect on impedance in high-speed sections is proportionally larger.

Where Manufacturing Difficulty Actually Sits

The hardest part of AR glasses production is not the board alone. It is the interaction between board fabrication and assembly.

Component density inside the glasses is high relative to available volume, and the product is weight-critical, so the PCBA stage is tightly constrained. As component spacing narrows, solder paste printing, placement accuracy, and hidden joint quality under ball grid array packages all influence final product yield. Paste volume variation that would be tolerable on a large board with generous apertures becomes critical when apertures are small, because the area ratio of a small aperture reduces paste transfer efficiency.

This is why the inspection chain matters as much as the fabrication specification. Paste inspection catches volume errors before they become joints. Optical inspection verifies placement and visible joints. X-ray resolves joints beneath packages where no camera can see. Electrical test confirms that the assembled device matches the design. Each method covers failure modes the others cannot detect, and none is sufficient alone. A complete test flow is what prevents a marginal joint from becoming an intermittent fault in a customer’s hands, and it belongs in a documented quality system rather than being improvised per project.

Equally important is traceability. When a wearable is returned, the manufacturer must be able to retrieve the laminate lot, the fabrication parameters, the assembly conditions, and the inspection results for that unit. In a high-mix production environment with several product lines running simultaneously, that record is what allows a quality question to be answered definitively instead of by recall. For teams building wearable-class PCB and PCBA products, the practical capability set to confirm includes 1 to 5 stage HDI, FPC and rigid-flex construction, minimum line width and spacing in the 0.076 mm range, and a coordinated fabrication and assembly flow.

Multi-Model Production and Version Churn

Consumer smart glasses remain in a phase of rapid iteration. When the main processor, camera, battery, or optical structure changes, the PCB layout and bill of materials usually change with it. A manufacturer with several product lines in the market at once, spanning domestic and export channels, has to handle multiple board types, multiple versions, and multiple batches simultaneously.

That creates a manufacturing problem that is different from pure scale. Volume production with a single stable design rewards optimisation: the process is tuned once and repeated. Volume production with continuous version churn rewards flexibility: the process must be re-established quickly for each revision without losing the consistency achieved on the previous one, and engineering effort has to be shared efficiently across a portfolio rather than concentrated on one flagship.

The practical implication is that the interface between fabrication and assembly becomes the critical path. Every time a board revision moves from fabrication to placement, the assembly process has to be adjusted: stencil, paste parameters, placement programme, reflow profile, and inspection settings. When fabrication and assembly sit with different suppliers, each revision requires information transfer across a company boundary, and that is where engineering time is lost and where defects are introduced during transition. Keeping both stages under one supplier, as with coordinated volume assembly, reduces the transition cost per revision rather than per project.

Overseas Compliance Is Not a PCB Certificate

With roughly a third of revenue coming from overseas markets spanning more than twenty countries, regulatory requirements become part of the production plan rather than a separate activity.

It is worth being precise here, because the terminology is often confused. Frameworks such as CE marking and FCC certification apply to the complete product or to the radio module within it. They are not certificates that a PCB manufacturer attaches to a bare board. What the PCB supply chain contributes to those approvals is evidence: material compliance declarations for RoHS and REACH, halogen-free test data where required, and consistent manufacturing records that demonstrate the product shipped matches the product tested.

That distinction matters commercially. A product approved on a prototype built from one material lot, then produced over years with occasional undocumented substitutions, carries a compliance risk that only becomes visible during a market audit. A supplier that records material lots and routes changes through a formal engineering change process helps the customer defend the approval rather than merely obtain it. That is what the test and verification stage ultimately supports: not a single pass result, but a reproducible product.

The broader point about AR glasses is that the hard transition is not miniaturisation. Miniaturisation is an engineering problem that design and process development can solve, and the industry has already demonstrated it. The hard transition is turning a design that works into a product that behaves identically across hundreds of thousands of units, across multiple versions, and across markets with different regulatory expectations. That is a manufacturing system question, and it is what separates a supplier who can build a prototype from one who can support a product line.

Frequently Asked Questions

Why does 100,000 units change the manufacturing requirement? Because screening and inspection can absorb variation at low volume, but at scale the defect rate converts into significant rework cost. The process itself has to be stable rather than the output being filtered.

Why do AR glasses need HDI and FPC rather than a single rigid board? The electronics are distributed around the frame, temples and optical module to save weight and volume, so high-density local routing and flexible interconnections between structural parts are required.

What makes small boards harder to produce? Absolute tolerances become proportionally larger relative to feature size, and small stencil apertures reduce paste transfer efficiency, so process variation affects yield more than it does on large boards.

Which inspection methods are needed? Paste inspection, optical inspection, X-ray for hidden joints, and electrical test. Each covers different failure modes, and traceability links the results to individual units.

Does CE or FCC certification apply to the PCB? No. Those frameworks apply to the finished product or radio module. PCB manufacturers contribute material compliance evidence and consistent manufacturing records that support the customer’s approval.