Tesla robotaxi rollout tests AI growth beyond car fleet

Tesla robotaxi growth is advancing cautiously as the company pairs city-by-city launches with wider AI plans for Optimus and distributed compute.

Tesla management used its second-quarter earnings call to frame autonomy as a safety-led rollout, not a sudden national switch. According to the supplied ARK commentary, executives said the unsupervised robotaxi service is operating in seven cities while Tesla adds locations one at a time.

Seven cities shape the rollout

Ashok Elluswamy, Tesla's VP of AI, said the fleet has accumulated more than 380,000 unsupervised robotaxi miles, with no notable incidents reported by management. He also said unsupervised miles have grown at a double-digit weekly rate since the start of 2026.

The next test is whether the city-by-city model can shorten approval cycles without weakening safety controls. Elluswamy said the time needed to open a new market should “continue to trend toward zero,” a path that could eventually support approvals across entire states.

Tesla executives also said the current live fleet is testing Full Self-Driving v15, the software architecture management expects to underpin scaled unsupervised service. According to the commentary, the version has achieved 40% of planned improvements, leaving most of the stated technical gains still ahead.

Optimus shifts toward Fremont

Tesla's humanoid robot effort is moving more slowly than the robotaxi network, but the company has tied it to real factory plans. In its shareholder letter, Tesla said Optimus production is expected to begin later this year at Fremont, using manufacturing lines that previously built the Model S and Model X.

Elon Musk acknowledged that scaling Optimus manufacturing will be difficult, a rare point of caution in a business line often discussed in large-market terms. ARK's research view is that Optimus could approach human-level performance near the end of this decade, though that is an investment firm's assessment, not a company milestone.

The link between Optimus and robotaxis is Tesla's attempt to reuse software, data infrastructure and manufacturing discipline across machines that operate in the physical world. If that reuse works, Tesla could spread AI development costs over cars, robots and fleet services rather than treating each product as a separate bet.

Megapod pushes compute outward

Tesla's AI plan now reaches beyond vehicles and robots into distributed computing. During the earnings call, Musk described a Megapod architecture that pairs Tesla's AI4 chip with x86 compute in a Megapack-style system that could be deployed “anywhere in the country or outside the country.”

ARK said Tesla's Supercharger network has seven gigawatts of available power, which it presented as a potential hedge against electricity limits facing centralized data centers. The mechanism is straightforward: inference workloads could move closer to the grid's edge if local power and hardware are available.

Tesla is also building Starlink into Cybercab, according to the commentary, with uses ranging from in-cabin entertainment to navigation, customer service and fleet management. Musk suggested the same terminals could operate as mobile relay points for nearby phones and WiFi devices, extending the car's role from transport asset to connectivity node.

The scenarios now split by execution. If robotaxi approvals accelerate and FSD v15 delivers the remaining planned gains, Tesla could strengthen its core company economics while forcing the autonomous-vehicle sector to respond faster; if safety reviews slow expansion, capital stays tied up longer and rivals gain time.

If Optimus production at Fremont proves repeatable and Megapod compute finds buyers, Tesla's AI story becomes less dependent on car deliveries; if either effort stalls, the wider industry keeps treating autonomy, robotics and edge compute as separate markets rather than one integrated platform.