Tesla’s weakest manufacturing metric is no longer battery chemistry or drivetrain efficiency it is empty space on the factory floor. After two consecutive years of delivery declines, the company’s global footprint increasingly resembles the traditional auto industry problem it once claimed to have escaped: too much installed capacity chasing too little demand.
Tesla said annual sales fell 6.7% last year and production totaled 1.65 million vehicles. Against more than 2.35 million units of stated annual capacity across Shanghai, Fremont, Berlin and Austin, that implies about 70% utilization roughly in line with the North American industry average and below what analysts typically regard as healthy. The shortfall matters because an auto plant’s economics are unforgiving: depreciation, tooling refreshes, maintenance, utilities and staffing do not scale down gracefully when a line slows.
The pressure points are visible in the newer sites. Europe deliveries were down sharply, leaving Berlin with a wide gap to its rated output. In Texas, the Cybertruck’s underwhelming volumes have turned what was sold as a marquee capacity expansion into a partially stranded asset. That mismatch is why consultants and analysts keep returning to the same diagnosis: Tesla’s volume base is being asked to ride on products that are no longer new. As Tu Le of Sino Auto Insights put it, “They’re relying on a six- and a nine-year-old product to hold almost two million units – too many units of capacity. That’s an impossible task.”
There is a tactical release valve: build something else. That is the industrial logic behind Elon Musk’s decision to repurpose parts of Fremont for Optimus humanoid robots and to talk openly about extreme ramps. “We’re going to launch on the fastest production ramp of any large, complex manufactured product ever, starting with building a one-million-unit production line in Fremont. That’s line one,” Musk said, adding a second line target of 10 million units per year in Austin. Even if those numbers remain aspirational, the underlying strategy is clear: keep high-fixed-cost assets busy by shifting from vehicle-only throughput to a broader manufacturing portfolio.
Robotaxis sit in the same bucket high promise, low present-day absorption of capacity. Musk has said Cybercab production starts in April, but a steering-wheel-less vehicle runs directly into the reality that U.S. passenger vehicles must meet FMVSS equipment rules unless a manufacturer secures an exemption; NHTSA said Tesla has not applied for any exemptions for the Cybercab. In parallel, the cost case for robotaxis is still shaped by utilization and operations labor, including remote support. One peer-reviewed model found robotaxi labor can remain material at $1.02 per mile, challenging the simplistic narrative that autonomy automatically deletes human cost.
Tesla does have other levers energy storage deployments have been strong in recent quarters, and the company has signaled a Nevada line designed for 50,000 Tesla Semis a year. But the decisive constraint remains the same: an automaker can out-engineer many things, yet it cannot outrun the arithmetic of plant utilization for long. Tesla’s distinctive move is not that it has excess capacity; it is that it is attempting to treat that capacity as a platform for machines beyond cars.