From 77 GHz EV radar to 1.6T AI data-center links: where new energy and AI are pulling high-frequency PCB demand — and the five challenges in the way.
Two engines — the new energy industry and artificial intelligence — are reshaping the demand picture for high-frequency PCBs at the same time. Both make extreme, precise demands on exactly what these boards do best: moving high-frequency, high-speed signals with minimal loss, delay and interference. The opportunities are real and large; so are the challenges standing in front of them.
New energy vehicles lead the industry in intelligence, and their ADAS and autonomous-driving functions lean on millimeter-wave radar and LiDAR — sensors running at frequencies like 77 GHz that need very low-loss laminates, with the RO4000 series as the standard answer. Add 5G-V2X communication and the data-hungry infotainment and T-Box systems, and a single vehicle becomes a bundle of high-frequency, high-speed boards — a large, high-value incremental market.
800V platforms and their ultra-fast chargers are deploying fast, and inside every charging pile the power distribution units, power management modules and communication control circuits must hold stable high-frequency signal control under high voltage and current. That calls for boards with strong heat resistance, high Tg and low dielectric loss — new territory for high-frequency PCBs in energy infrastructure.
Modern inverters and energy storage converters are no longer simple power devices; they run intelligent monitoring, group control and grid interaction, including virtual-power-plant participation. The communication, control and monitoring modules inside process heavy real-time data, which raises the bar for signal integrity and anti-interference capability — and pulls high-frequency boards into a sector that barely used them before.
Large-model training and inference run on server clusters whose GPUs, ASICs and FPGAs operate at extreme frequencies and data rates — PCIe 5.0/6.0, DDR5 memory and beyond. The motherboards and accelerator cards carrying this silicon are typically 10-plus-layer HDI builds on the highest-grade laminate, because every dB of distortion is computing power wasted. The supply squeeze around these materials is tracked in the AI computing supply analysis.
AI data centers live on internal data exchange, and the 400G, 800G and 1.6T optical modules plus high-speed switches that carry it are among the most demanding applications that exist — their boards specify ultra-low-loss and very-ultra-low-loss laminates. The technology split behind those categories is explained in high-speed versus high-frequency.
AI is moving from cloud to edge — AIoT devices, smart cameras, AR/VR headsets — and local inference raises the required communication rate between chips inside each device. Every one of those products is a smaller, cheaper high-frequency board, and the volume multiplier is enormous. The laminate requirements these terminals drive are catalogued in the AI-server laminate guide.
Material and process limits. PTFE-class substrates are soft, move with temperature and humidity, and make drilling, metallization and pattern transfer far harder than FR-4 — yield control is the first technical battle, and stacking HDI (multi-stage blind/buried vias, stacked vias) onto high-frequency materials multiplies the difficulty. Design barriers. High-frequency layout is electromagnetic simulation work — signal integrity, power integrity, EMC — and teams with that depth stay scarce. Supply concentration. High-end laminate capacity sits with a few foreign manufacturers, exposing buyers to price swings and geopolitical risk. Cost. Special materials, complex processes and yield losses keep manufacturing cost far above ordinary PCBs. Standards. Dk/Df measurement needs expensive instrumentation, and industry testing standards are still maturing — the same gaps the industry applications survey flags.
The new energy and AI waves are expanding demand for high-frequency PCBs in both volume and difficulty at once. For manufacturers, the winning posture is unglamorous: push material capability and process yield in parallel, build supply-chain depth to soften the import dependence, and grow the simulation-literate talent the designs now require. The undercurrents are real — and so is the size of the wave for whoever stays on the board.
Frequently Asked Questions
Because intelligence adds RF content: 77 GHz radar, LiDAR, 5G-V2X and high-speed infotainment each need low-loss high-frequency boards. A connected EV carries several times the RF board value of a conventional car.
High-frequency signal control must survive high voltage and high current in the same enclosure. That pushes requirements toward high Tg, strong heat resistance and low dielectric loss — a combination ordinary control boards never needed.
Switches and optical modules at 400G and above. Their boards specify ultra-low-loss and very-ultra-low-loss laminates, the most demanding (and expensive) tier of the material market.
Partly: qualifying second sources, choosing hybrid constructions that use less premium material, and locking volume pricing early all reduce exposure. Full independence is not yet realistic, but the exposure is manageable with planning.
Tell us your target segment — EV, charging, PV/storage or AI hardware — and we will match the material tier and process capability it demands.
Get a Fast Quote Fast Reply on WhatsApp