Tesla Profit Falls as Autonomy Bet Triggers Cash Burn

Tesla (TSLA ) spent its way to record revenue and a shrinking profit in the second quarter of 2026, as the money behind Elon Musk’s pivot from carmaker to robotics-and-autonomy company pushed it into its first cash burn in more than two years.

The company reported operating income of $398 million for the quarter, down 57% from a year earlier, cutting its operating margin to 1.4% from 4.1%. Revenue rose 26% to a record $28.2 billion, helped by 480,126 vehicle deliveries, also a second-quarter record. What the top line hides is that the capital funding Tesla’s move into robotaxis, humanoid robots and its own chips is now visibly outrunning what the car business earns.

The cost of the pivot

Capital expenditure more than doubled from a year earlier to $5.79 billion, a 142% jump, financing an autonomy build-out that spans self-driving software, the purpose-built Cybercab, the Optimus humanoid robot, and a set of chip and compute projects. That spending drove free cash flow to negative $1.1 billion, the first quarterly cash burn since early 2024, and trimmed Tesla’s cash reserves by $1.2 billion to $43.5 billion. Musk lifted the 2026 capital-spending plan above $25 billion in April, close to triple the $8.5 billion Tesla spent in 2025.

How healthy the profit looks depends on where you read it. GAAP net income of $1.1 billion slipped only 5%, but that figure was propped up by a $1 billion unrealized gain on Tesla’s stake in SpaceX. Strip that and other one-offs out, and adjusted net income fell 17% to $1.15 billion, well short of the roughly $1.95 billion analysts had expected, according to the Financial Times. Two pressures did most of the damage: revenue from selling regulatory credits to other automakers fell to $146 million from $439 million a year earlier, and automotive margins thinned as Tesla leaned on price cuts and low-rate financing to keep volume moving.

Not everything is under strain. Energy storage deployments hit a second-quarter record of 13.5 gigawatt-hours, up more than 40%, and the services unit posted record gross profit, giving Tesla two smaller businesses that generate cash while the autonomy programs consume it.

What the robotaxi map actually shows

For a company valued largely on autonomy, the disclosure that matters most sits below the margin lines: the robotaxi coverage table. Tesla says its Robotaxi service now operates in seven metro areas after launching in Miami, Orlando and Tampa in July, alongside Austin, Dallas, Houston and the San Francisco Bay Area. How those services actually run is the part the count leaves out.

By Tesla’s own accounting, the Bay Area service still operates with a safety driver behind the wheel, not driverless. The six Texas and Florida metros are labeled “ramping unsupervised,” and the two cities left from Tesla’s original 2026 target, Phoenix and Las Vegas, remain in “preparations.” Even the flagship Austin service, running for more than a year, still fields only about 20 vehicles, Electrek reported, a fleet that has barely grown.

The software carries the same distance between name and capability. Tesla’s “Full Self-Driving” is labeled, in the company’s own materials, as FSD (Supervised): a driver-assistance system that needs a human ready to take over at any moment, not an autonomous one. Active FSD subscriptions grew 56% to 1.48 million, and the feature won its first European approvals during the quarter, starting in the Netherlands and extending to Lithuania, Estonia, Denmark and Belgium. But the edge cases that separate assisted driving from unsupervised driving are the hard part, and Tesla’s own labels still place every one of its markets short of it.

Optimus, chips and what to watch

The spending is buying capacity that is not yet product. Tesla has pulled the Model S and Model X lines out of its Fremont plant to install the first-generation lines for Optimus, its humanoid robot, with the earliest units set aside for internal training rather than sale. It began building Cybercabs at Gigafactory Texas and started giving employees autonomous rides on the campus, both steps ahead of any public robotaxi launch of the vehicle. And it is pushing money into compute and silicon: the Cortex 2 training cluster is online, and an in-house chip fab in Austin is in early construction.

Whether that outlay becomes the high-margin autonomy revenue Tesla’s valuation assumes depends on execution the company has repeatedly delayed. The quarter sets the terms of the bet plainly: record cars, thinning margins and a cash outflow underwriting programs whose real-world footprint, measured against human-driven benchmarks, remains mostly supervised and mostly ramping. Investors sent the shares down more than 2% in after-hours trading.