Power Semis (motor drivers)

Power semiconductors are the rare component that scales ahead of the build ramp worldwide, there is no global deficit. Every actuator in a humanoid needs a motor-driver or gate-driver integrated circuit to switch its three-phase coils, about thirty such ICs per robot counting board-level power-management chips¹, and the numbers here track the robot-relevant part of the market: the motor-driver slice of the gate-driver IC business, which is roughly the twenty-eight percent of a one-and-a-half-billion-unit-a-year gate-driver IC market that drives motors²³, on the order of four hundred and twenty million ICs a year, a distinct slice of the broader two-hundred-and-forty-five-billion-unit analog-semiconductor universe behind it⁴. At about thirty ICs per robot, a worldwide fleet of roughly forty-seven million humanoids by 2040 needs about one-and-four-tenths billion driver ICs in total¹¹, yet cumulative global production over the same window delivers well over ten billion, leaving worldwide output about eight times ahead of need. The durable crossover, the year cumulative production drops below cumulative need, never arrives globally, and that no-deficit result is the robust finding: a mature-node analog fab base measured in billions of units a year simply does not bind at robot volumes. The real exposure is not raw silicon but concentration and grade: robot-grade gate-driver ICs are a thin, Western- and Japan-led slice, the automotive top five (STMicroelectronics, Infineon, NXP, Texas Instruments and onsemi) hold about seventy-seven percent, with ST alone near forty-one percent and no Chinese vendor in the top tier³. So building from one bloc alone is a different story. China is weak in robot-grade driver ICs, its power-semiconductor strength is in demand-driven automotive power devices, not these control ICs⁸, so China-only output falls behind the worldwide fleet around 2037. The United States is comparatively strong here, a rarity in this analysis: Texas Instruments is the world’s largest analog maker⁷, so US-only output nearly keeps pace and crosses only at the very end of the window, around 2040, and only just. Treat power semis as a pressure-release valve, not a volume chokepoint, the risk lives in who controls the robot-grade tier, not in whether enough wafers exist.

A motor-driver or gate-driver IC is the small power-control chip that sits between a robot’s controller and the windings of an actuator’s motor: it takes low-power switching commands and drives the gates of the power transistors that actually push current through the three-phase coils, with the protection, dead-time and current-sense logic that field-oriented control needs. Every actuated joint has one. A Tesla Optimus carries twenty-eight body actuators, fourteen rotary and fourteen linear, each a frameless torque motor that needs its own driver electronics¹, and counting board-level power-management ICs the model uses about thirty driver and power ICs per robot (a value of 0.03 in thousand-unit terms), roughly one driver per joint. Crucially this is the robot-relevant motor-control slice, the motor-driver segment, about twenty-eight percent of the one-and-a-half-billion-unit gate-driver IC market³, not the vast two-hundred-and-forty-five-billion-unit general analog-semiconductor universe of every op-amp, regulator and interface chip⁴, most of which has nothing to do with driving a robot’s motors. These control ICs are also distinct from the power devices, the IGBT and silicon-carbide transistors, that they switch.

Because a driver IC is built into a robot for life, the ICs the fleet needs are a cumulative stock, the total number of robots ever built worldwide multiplied by about thirty¹, not a per-year flow. Integrating the consensus shipment ramp¹¹, the worldwide installed fleet reaches about 585,000 robots by 2030, ten million by 2035 and forty-seven million by 2040, requiring roughly eighteen million, three hundred million and one-and-four-tenths billion driver ICs respectively. In the chart below, the black line is that worldwide cumulative need, the grey band is cumulative global production, and there is no red wedge: production stays far above need for the whole horizon.

Production is the only thing that differs between the Global, China and US views. The worldwide humanoid fleet and the ICs it needs stay identical in all three; only the supply line changes, asking whether one bloc’s fabs alone could equip the whole global fleet. This is the robot-relevant motor-driver / gate-driver IC slice, not the broad analog total, and real output follows an S-curve rather than a smooth ramp. Global production runs from about four hundred and twenty million ICs a year today²³ to roughly six hundred million by 2030, eight hundred and fifty million by 2035 and a modeled one-and-fifteen-hundredths billion by 2040, climbing on electric-vehicle, factory-automation and humanoid pull at or above the analog industry’s roughly seven-percent unit growth rate⁴, then plateauing on mature-node fab build cadence, not on any raw-material limit. The post-2030 figures are a modeled extrapolation, a humanoid-and-EV step-change above that base rate, not a reported forecast. The supplier base is Western- and Japan-led: Infineon, STMicroelectronics, Texas Instruments, onsemi, NXP and Renesas make the bulk of it²⁵, with the automotive top five holding about seventy-seven percent and ST alone near forty-one percent³. China climbs from a weak base, about twenty million ICs a year today, only around five percent of the slice, toward a modeled one hundred and ten million by 2040: it has no vendor in the gate-driver top tier, and its real power-semiconductor strength is in automotive power devices, where domestic champions like BYD Semiconductor (about twenty-six percent of China’s e-drive power-device market), CRRC Times and Silan dominate a vertically-integrated, demand-driven IGBT and silicon-carbide business⁸, not robot-grade control ICs. Its ramp rides the push for chip self-sufficiency and a mature-node foundry build-out, China’s legacy-node capacity share rising from twenty-nine to thirty-three percent by 2027⁹⁸. The United States holds far stronger ground than usual: it is the top single-country analog producer at about twenty-three percent⁶, Texas Instruments is the world’s largest analog maker and is investing more than sixty billion dollars in new Texas and Utah fabs⁷, so US output runs from about thirty-three million ICs a year toward a modeled one hundred and forty million by 2040. Even so, gate-driver ICs specifically are more European-concentrated than analog overall, ST alone leads them, so the US share of this particular slice sits nearer thirteen to fifteen percent.

For most components this section names the crossover year; for power semis there isn’t one to name globally. Cumulative global production runs about eight times ahead of the fleet’s need at 2040, well over ten billion driver ICs produced against a one-and-four-tenths-billion requirement¹¹, so the durable crossover never arrives, the robust finding. A mature-node analog fab base measured in billions of units a year is simply too deep for robot demand, which amounts to less than a single year of global gate-driver output, to drain it²⁴. Where the picture changes is building from one bloc alone. China-only output crosses below the worldwide fleet’s need around 2037: it starts from a weak robot-grade base, no top-five vendor, around five percent of the slice, and even a fast ramp on domestic self-sufficiency cannot equip the whole global fleet from China’s control-IC fabs alone⁸. The United States, unusually for this analysis, is the strong bloc: with Texas Instruments and onsemi behind it⁷, US-only output very nearly keeps pace and crosses only at the very end of the window, around 2040, and only just, the latest single-bloc crossover in the whole component set. The contrast with the grinder-bound parts is the whole point: where roller screws or cross-roller bearings cross in the mid-2030s and meet well under half the 2040 need, power semis track the ramp globally the entire way, and even a single strong bloc nearly does. The exposure here is not a volume wall but a concentration risk, who owns the robot-grade tier, which is a reshoring-and-qualification problem, not a global supply gap. One honest caveat on those two bloc years: unlike the global no-deficit result, they are the lower-confidence finding, because they depend on the robot-grade carve-out, treating only the motor-driver slice, about twenty-eight percent of the gate-driver IC market, as the binding base³, and on modeled per-bloc production shares. Read the broad gate-driver universe instead and even the single-bloc views stay elastic; it is the robot-grade interpretation that makes a bloc cross. The global-never-binds verdict holds either way.

What makes the precision-mechanical parts tight does not apply here. There is no scarce grinder, no thin skilled-labor base, no multi-year qualification on a single European machine tool. Motor-driver and gate-driver ICs are built on broad, multi-sourced mature nodes, twenty-eight to ninety nanometres, the same legacy analog and power processes that already turn out hundreds of billions of chips a year⁴, far cheaper than leading-edge logic and not export-restricted at the wafer level. That is why the global verdict is elastic: capacity can be added on ordinary fab expansions rather than waiting on a rare tool. Two real constraints temper the “no barrier” verdict, and both are about timing and grade rather than raw volume. The first is fab lead time: a greenfield analog/power fab takes about two years to bring online, and the 2021 chip shortage, when cars waited a year for forty-nanometre-and-older analog, power-management and driver chips even though the silicon itself was abundant, showed that mature-node analog is fab-lead-gated, prone to sharp short-run spikes when demand steps up faster than lines can be built¹⁰. The second, and the one that actually shapes the bloc story, is grade concentration: robot-grade field-oriented-control driver ICs are a thin, Western- and Japan-led slice, the automotive top five hold about seventy-seven percent and STMicroelectronics alone near forty-one percent, with no Chinese vendor in the top tier³. China’s headline power-semiconductor strength, the “over forty percent” figures often quoted, is demand and device assembly (IGBT and silicon-carbide modules for its own electric vehicles), not production of robot-grade control ICs⁸. So the binding step, in the one scenario where it binds, is not a shortage of wafers but the qualification and reshoring of robot-grade driver-IC capacity outside the incumbent Western and Japanese base, a concentration the global market masks but a single decoupled bloc cannot. The verdict, globally elastic, no global crossover, is robust across a wide range of assumptions; it rests on the world keeping access to a deep, broadly-sourced mature-node fab base, which it does.

Each cobot joint has its own three-phase servo inverter, about six power modules per arm, and each industrial-arm axis its own servo amplifier, another six. So both the 2040 cobot and industrial-arm fleets add to the power-semiconductor draw. But like humanoids they ride on the vast mature-node analog/power-IC base that produces hundreds of billions of devices a year, so the addition is a rounding error and the link never binds. Quadrupeds join the same pool: each of their roughly twelve quasi-direct-drive joints carries its own inverter channel, about a dozen driver ICs per unit, and a quadruped fleet of about five million by 2040 is a still-smaller draw on that billions-a-year base, it never binds either. Drones are the sharpest illustration on this page that volume is not the same as binding: a direct-drive multirotor carries one electronic-speed-controller inverter channel per propeller motor, about four and a half per unit fleet-blended, so a cumulative drone fleet of roughly four hundred and seventy-eight million by 2040, units built, not a standing fleet, since about half are single-use military first-person-view craft, consumes on the order of two and fifteen-hundredths billion ESC channels, more than the humanoid fleet’s one-and-four-tenths billion driver ICs and the largest single draw of any embodiment here. Yet the broad power-discrete base ships tens of billions of units a year, so even that draw is absorbed without a ripple and never binds.