How Many Circuit Boards Does a Robotaxi Need? L5 Electronics

On 3 September 2026, Tesla’s Cybercab began commercial operation in Austin, Texas, with plans for its first public display in Beijing and Shanghai later in the month. The vehicle is a two-seater with no steering wheel, no pedals, and no conventional mirrors. Registration data indicated roughly 45 units in Texas by the end of August, so the fleet is small. What makes the event notable is not the volume but the category: a purpose-built robotaxi carrying paying passengers.

Removing the driver’s controls has a direct consequence for electronics. In an assisted-driving vehicle, the human is the final redundancy layer. In a driverless one, perception, computation, communication, and actuation have to carry that responsibility between them. Every circuit board in that chain inherits a different reliability requirement than it would in a private car.

Removing the Steering Wheel Removes the Fallback

Published information describes the vehicle using eight high-definition cameras for environmental perception, with onboard computing performing real-time decision making. That architecture concentrates a great deal of capability into a small number of electronic assemblies.

The first effect on the board is in the sensor links. Cameras produce continuous high-bandwidth data streams, and each stream must reach the compute platform without corruption. That means serializer-deserializer links, automotive Ethernet, and their associated power and clocking networks coexisting in a confined space. High-speed differential pairs require continuous reference planes and stable impedance, so layer stackup, dielectric thickness, and copper weight variation all translate directly into link margin.

The second effect is that the compute platform must do more than process data. It must do so predictably. Real-time decision paths have timing budgets, and a board that introduces unexpected delay through marginal signal integrity or thermal throttling is not merely slower; it is outside its design envelope.

Where the Boards Actually Sit in a Driverless Vehicle

Rather than asking how many boards a robotaxi contains, it is more useful to ask what each one has to survive. In practical terms, the electronic content divides into several functional groups.

Central compute. The highest-value board in the vehicle, carrying the main processor, memory, storage, and the high-speed interfaces connecting them. This is where layer count, HDI structures, and impedance control matter most, and where thermal design is most constrained because the processor dissipates the greatest power in the smallest area.

Sensor front ends. Camera modules, radar, and ultrasonic assemblies each contain a small board handling the interface, timing, and power conditioning for the sensor itself. These boards are numerous, physically small, and exposed to the harshest environmental conditions in the vehicle, including direct sunlight, vibration, and moisture.

Actuation and drive. Steering, braking, and propulsion are controlled through power electronics carrying substantial current. These boards combine thick copper and thermal management with control-side communication, and they are the components whose failure has the most immediate consequences.

Power distribution and conversion. A driverless vehicle needs redundant supply paths, DC-DC conversion, and monitoring. This is often where the vehicle’s safety architecture becomes physical: two independent paths, each monitored, each capable of carrying the load.

Communication and connectivity. Teleoperation, fleet management, over-the-air updates, and passenger interface functions require a connectivity module and its supporting boards.

Each of these groups draws on a different manufacturing capability: high-layer-count HDI for compute, fine-line and often flex for sensors, thick copper for actuation, and controlled-impedance multilayer for communication. The vehicle as a whole therefore needs four different board technologies, all qualified to automotive standards.

Central Compute: High-Speed Signals and Power on One Substrate

Vehicle electronics previously distributed computation across many electronic control units. As driving intelligence increased, computation consolidated into domain controllers and now into central computing platforms. Consolidation increases the density of processors, memory, and power devices on a single board, and it merges two manufacturing problems that used to be solved separately.

On the signal side, high-speed differential networks need consistent impedance and uninterrupted return paths. Line width variation after etching, dielectric thickness variation after lamination, and copper roughness all affect insertion loss, and the effect compounds with channel length. On the power side, high-performance processors draw large transient currents, requiring thick copper planes, a dense decoupling network, and short, low-inductance paths from capacitor to die.

Accommodating both in one stackup is an exercise in compromise that has to be planned rather than discovered. Layer assignment, plane partitioning, and via structures determine whether the high-speed channels and the power network can coexist. That planning belongs in a layout review that includes fabrication constraints, because a stackup that cannot be built to the required tolerance is not a stackup.

Layer count increases are often the visible response, but the underlying requirement is different. More layers are a means of fitting high-speed routing, power distribution, and control networks into a limited footprint. Once the count rises, pressed thickness control, lamination registration, drilling accuracy, and barrel plating reliability all become harder to hold, and consistency in volume becomes the real challenge. Those constraints are worth stating explicitly in a PCB manufacturing capability review before the design is frozen.

Utilization Changes the Reliability Requirement

A commercial robotaxi differs from a private car in how it is used. A private vehicle may operate for an hour or two a day. A revenue-generating robotaxi is expected to run for most of the day, charging between trips, accumulating far more thermal cycles, vibration hours, and power-up events in a single year.

That changes which failure mechanisms matter. Solder joint fatigue accumulates with thermal cycling, and the joints most exposed are those under large ball grid array packages, where the thermal expansion mismatch between package and board creates the greatest strain. Connectors experience insertion and vibration wear. Camera modules, mounted at the extremities of the vehicle, absorb vibration and thermal shock that internal electronics never see. Power devices cycle between high and low current hundreds of times per day.

Because these are wear-out mechanisms rather than random defects, the manufacturing response is layered inspection and traceability. Solder paste inspection catches volume errors before they become joints. Optical inspection verifies placement. X-ray inspection resolves hidden joints under packages where no camera can see. Electrical test confirms that the assembled board matches the design. A complete PCBA test strategy combines these rather than relying on any single method, and a quality management system records the results against the individual unit so that a field event can be traced back to its manufacturing history.

Traceability deserves emphasis in this application. When a vehicle accumulates more operating hours in a year than a private car does in a decade, an intermittent defect that would surface slowly elsewhere can appear within the warranty period. Being able to bound the affected population precisely, rather than inspecting or recalling everything, is a commercial advantage as much as a technical one.

Why Board Count Is the Wrong Metric

It is tempting to estimate the electronics content of a robotaxi by counting boards. That metric misleads in two directions.

First, sensor modules contain many small boards, so the count rises without much value being added. A camera module’s interface board is a low-cost item regardless of how many exist. Second, the value concentrates in a few boards that are difficult to build: the central compute platform, the sensor fusion assembly, and the solid-state power electronics. Those boards carry the layer counts, the fine lines, the impedance tolerances, and the thermal requirements that determine what a manufacturer must be capable of.

The more useful framing, and the one the industry is converging on, is functional density: how much computation, communication, and power delivery is integrated onto each critical board, and whether that integration can be reproduced reliably at volume. A robotaxi with forty modest boards and a robotaxi with twelve sophisticated ones may have similar total content, but the latter places far greater demands on the supply chain, particularly where AI-capable PCBA manufacturing is concerned.

It is also worth maintaining perspective on timing. Cybercab’s Austin operation remains limited in scale, and regulators in the United States have begun examining the self-certification process for the vehicle’s safety compliance. It is premature to read the launch as an immediate surge in automotive PCB demand. What the vehicle does demonstrate is a direction: once the steering wheel is gone, the vehicle depends entirely on computation, perception, and electronic actuation. When that architecture spreads, PCB value will rise not through the number of boards per vehicle, but through what each critical board must integrate and how consistently it can be manufactured.

Frequently Asked Questions

Do driverless vehicles use more circuit boards than conventional cars? Not necessarily far more in count, but the boards that matter carry far greater complexity. Value migrates from many simple controllers to fewer, denser computing and power boards.

Why does central computing increase PCB difficulty? It places high-speed interfaces, dense processor and memory packages, and high-current power delivery on one substrate, so signal integrity and power integrity constraints must be satisfied simultaneously.

What fails first in high-utilization robotaxis? Wear-out mechanisms dominate: solder joint fatigue under large packages, connector wear from vibration, and thermal cycling of power devices. These accumulate far faster than in a private vehicle.

Which inspection methods matter most? No single method suffices. Paste inspection, optical inspection, X-ray for hidden joints, and electrical test each cover different failure modes, and traceability ties the results to the individual unit.

Is a robotaxi launch an immediate driver of automotive PCB demand? Not at current volumes. It signals an architectural direction whose effect on demand depends on how widely purpose-built autonomous platforms are adopted.