Tesla Q2 2026: Record Deliveries, Robotaxi Expands to Florida, Optimus Enters Production

Tesla’s second quarter of 2026 tells two simultaneous stories — one about a vehicle business finally breaking records on deliveries, and one about a technology company burning through capital so fast that its operating profit fell 57% despite a $28 billion top line. The two stories are not contradictions; they are what a transition from automaker to AI-and-robotics company looks like from the inside, in real time.

The quarter produced 480,126 deliveries — Tesla’s best quarter ever — and $28.24 billion in revenue, up 26% year-over-year. But adjusted earnings per share came in at $0.33 against a Wall Street consensus of $0.49, a miss of nearly 33%. Operating income fell to approximately $400 million, down from roughly $930 million a year ago. Research and development spending jumped 49% to $2.37 billion. The AI infrastructure build is expensive, and Tesla’s financials are now structured around absorbing that cost.

Robotaxi: Seven Cities, 2.5 Million Miles

The operational highlight of the quarter was Tesla’s Robotaxi expansion. The fully unsupervised service, which began limited public operations in Austin in 2025, expanded to Miami, Orlando, and Tampa during July 2026, bringing the total number of active U.S. markets to seven. Cumulative paid Robotaxi miles reached 2.5 million.

The Florida launches are noteworthy for a reason beyond geography: Florida’s regulatory environment, road network, and climate present different challenges than Texas — more humidity, more pedestrian-dense beach corridors, more erratic weather. Successful operation in Florida would strengthen the case that Tesla’s Robotaxi system generalizes across environments rather than being tuned to specific road conditions.

Full Self-Driving subscriptions reached 1.48 million active users by the end of Q2, a 56% increase year-over-year. In North America, more than 55% of new Tesla vehicle purchases included FSD at the point of sale — a figure that, if sustained, means the software revenue layer is now structurally embedded in Tesla’s vehicle business. Cumulatively, Tesla has logged nearly 12 billion miles of FSD Supervised driving, with European deployments exceeding 50 million kilometers.

Optimus: From Lab to Factory

The Optimus humanoid robot program crossed a threshold in Q2: Tesla announced that initial manufacturing runs are expected to begin in Q3 2026 at a dedicated production facility being built inside Gigafactory Texas. Elon Musk described one of Tesla’s “greatest engineering and AI challenges” as training robots to perform generalized tasks without explicit programming — the shift from robot-as-mechanical-tool to robot-as-autonomous-agent.

The Optimus program represents Tesla’s largest bet on AI applied to the physical world. Unlike the Robotaxi, which adapts a vehicle that Tesla was already manufacturing, Optimus requires building a new manufacturing infrastructure from scratch. The company has not disclosed unit volume targets or production costs, but the Q3 manufacturing launch is the first step toward transitioning from prototype demonstrations to commercial scaling.

AI Infrastructure: Cortex Doubles

Tesla’s AI compute capacity in Texas more than doubled during the quarter. The Cortex 1 AI training cluster reached 90 megawatts of sustained capacity, and Cortex 2 came online at 115 megawatts — together representing 205 MW of dedicated AI training infrastructure on a single campus. This compute is used to train the neural networks that power FSD, Robotaxi routing, and Optimus.

More surprising was an announcement that received relatively little attention: Tesla disclosed plans for TERAFAB, an internal chip fabrication initiative, with a dedicated launch event planned for later this year. The move would, if realized, represent Tesla’s most aggressive step yet toward vertical integration in AI silicon — following a path that has worked for Apple (which designs its own silicon) and is being pursued by Amazon and Google.

The strategic logic is straightforward: Tesla’s Robotaxi and Optimus ambitions require enormous amounts of inference compute, deployed in vehicles, robots, and data centers. Designing custom chips for specific workloads — rather than buying general-purpose GPUs from Nvidia — could dramatically lower the unit cost of intelligence at scale.

Energy: Strong Volume, Compressed Margins

Energy storage deployments hit 13.5 gigawatt-hours in Q2, up 41% year-over-year — a strong operational result. But gross margins in the energy segment contracted sharply, from 39.5% to 20.4%, driven by tariff impacts and rising component costs. The compression suggests that Tesla’s energy business, which had been a bright spot for margin expansion, is facing headwinds that the vehicle business already knows well.

What the Market Made of It

Tesla’s stock reaction was measured. The revenue beat was visible, but the EPS miss and operating income collapse were harder to dismiss. Investors who had been expecting a profits recovery found instead a company still in the middle of absorbing a multi-year technology buildout — one that includes AI training compute, Robotaxi infrastructure, Optimus production lines, and now internal chip fabrication.

The counterargument — and it is the argument that Tesla bulls have been making for years — is that none of these investments are speculative at this point. The Robotaxi is operating commercially in seven U.S. cities. Optimus is entering production. FSD subscriptions are growing at 56% year-over-year. TERAFAB, if it follows through, would give Tesla structural cost advantages in AI inference that could matter enormously at scale.

The question is not whether Tesla is building something real. It clearly is. The question is how long the transition absorbs earnings before the new revenue streams are large enough to offset the investment, and whether the operating leverage that Tesla’s vehicle business once exemplified will reemerge on the other side. Q2 2026 doesn’t answer that question — but it does establish, unambiguously, that Tesla is no longer primarily in the business of selling cars.