The printed circuit board is the laminated, etched-copper substrate that mechanically holds and electrically interconnects almost every electronic component ever shipped, and its current interest lies less in the substrate itself than in three live pressures: how far its cheap lithography can be pushed (100 GHz surfaces, trapped-electron qubits), how badly its design tooling lags integrated-circuit EDA, and how concentrated its manufacturing has become outside Europe.
Summary
A printed circuit board is a rigid insulating laminate carrying patterned copper layers. Components are soldered to pads; signals travel along etched traces and through plated vias between layers. The manufacturing process is subtractive photolithography at a coarse resolution compared with semiconductor fabrication, which is exactly why it is cheap and why it is the default substrate for anything that is not a chip. The parameters that decide what a PCB can do are layer count, minimum trace and gap width, dielectric loss at the operating frequency, and the tolerance stack of drilling and lamination.
The sources here use PCBs in two distinct ways. First, as a low-cost fabrication platform for physics that previously needed custom processes: a peer-reviewed continuous-variable thermodynamic computer built from RLC unit cells on a board and run as a parallelised Langevin sampler ref; a radio-frequency higher-order topological metasurface of split-ring resonators demonstrated at 1.5 to 1.7 GHz explicitly to avoid “expensive processing techniques”; a monolithic single-board Paul trap for electrons requiring no assembly, with observed trapped lifetimes of 2.13 ms and secular frequencies up to 90 MHz; and a reconfigurable intelligent surface operating around 100 GHz, where the authors list fabrication limits, switch size relative to unit cell, switch parasitics and control complexity as the binding constraints.
Second, as a design and manufacturing problem in its own right. Signal integrity analysis on board interconnects is now hard enough that machine-learning surrogates are being built to predict eye height and eye width across buffer technologies over interconnects with 44 design parameters. Generative AI has been applied heavily to integrated-circuit design automation but thinly to boards, with domain-specific data scarcity and poor integration with existing PCB tools identified as the blockers. The first text-based PCB question-answer benchmark, 480 pairs derived from 8 open-source hardware projects, appeared in mid-2026 pcb qa evaluating llms over the first printed circuit board.
The third pressure is industrial. Europe’s share of global bare-board PCB production is reported to have fallen from 16% in 2000 to roughly 2.3% in 2022, with environmental and regulatory barriers cited as constraints on rebuilding domestic capacity, making boards a chokepoint for defence electronics ref. Boards are also now a documented attack surface: illuminating a device with a low-power radio signal allows an eavesdropper to read backscatter from internal cables and PCB traces and infer user activity, a class of vulnerability the NSF began funding work on in July 2026.
Evidence base
- Europe’s share of global bare-board PCB production is reported to have fallen from 16% in 2000 to roughly 2.3% in 2022, with regulatory barriers constraining capacity growth; an EU PCB earmark resolved under EDIP on 23 June 2026 ref.
- A continuous-variable thermodynamic computer built from RLC unit cells on a printed circuit board, run as a parallelised Langevin sampler, was published in Nature Communications on 15 April 2025; the board form was pre-silicon and acknowledged as unscalable at the 8-cell prototype stage in January 2024 ref ref.
- A wideband PCB-based reconfigurable intelligent surface operating around 100 GHz was demonstrated using AlGaAs SP3T bare-die switches and optimised bond-wire interconnections, with fabrication limits, switch size and parasitics named as the binding constraints (published 24 May 2026).
- A monolithic single-board RF Paul trap for electrons, requiring no assembly, achieved trapped-electron lifetimes of 2.13 ms and secular frequencies up to 90 MHz (22 July 2026).
- PCB-QA, described as the first PCB design question-answer dataset, comprises 480 manually created question-answer pairs from 8 open-source hardware projects and tests whether LLMs can read schematics and netlists in native formats (24 June 2026).
- Two coordinated NSF SaTC awards of $299,999 and $300,000, dated 25 July 2026, fund modelling and mitigation of reflective backscatter side channels in which a remote low-power radio can recover typing from reflections off internal cables and PCB traces, through walls and without line of sight.
Open questions
- Does the EDIP earmark translate into measurable new European bare-board capacity, or does environmental permitting keep the 2.3% share flat?
- Can the data-scarcity problem in PCB design automation be solved without proprietary vendor datasets, given that PCB-QA currently spans only 8 open-source projects and 480 question-answer pairs?
- Do board-based reconfigurable surfaces at around 100 GHz survive the move from bond-wired bare-die proof of concept to manufacturable panels, or does switch integration force a semiconductor process?
- How expensive is mitigation of reflective backscatter leakage from PCB traces, in shielding cost or layout constraint, once the NSF ReBac work quantifies the channel?
Assessment drafted 2026-08-31 from up to 14 KB sources using the technology-scorecard framework; scores are a draft read pending review.