Tesla Stock: The $25 Billion Burn That Changes Everything

The headlines from Tesla's second-quarter earnings say what looks like a turnaround. Record deliveries, 25% year-over-year growth, revenue up 26%. The stock had already sold off nearly 27% year-to-date before the report, which is exactly the kind of price action that makes inflection hunters sit up.

But the market is still pricing the old story—a cash-generating electric car company with AI upside—while the numbers now point to something entirely different. Tesla has become a cash-burning AI infrastructure builder, and the core car business can no longer fund the ambition.

The delivery rebound. The margin collapse.

Q2 2026 deliveries hit 480,126 vehicles, roughly 74,000 above Wall Street's consensus estimate. That's the strongest second quarter in company history and the first year-over-year growth after two straight years of decline. The auto business is stabilizing.

The problem is what it costs to get there. Operating margin collapsed to 1.4%, down from 4.1% a year ago and 4.2% in Q1. Automotive gross margin excluding regulatory credits fell to 16.3%, well below the 18.7% the street expected. The $1.11 billion in GAAP net income included a $1 billion unrealized gain on SpaceX equity—an event that doesn't show up in operating cash flow and won't recur.

Margins are down because average selling prices have been cut to move volume after the federal $7,500 EV tax credit expired, regulatory credit income dropped to $146 million from $439 million a year earlier, and operating expenses surged 47% to $4.35 billion. The margin compression isn't one bad quarter. It's what happens when a company that once commanded pricing power starts fighting BYD for every sale while spending nearly half as much again on AI research.

The cash-flow proof point

This is the metric that changes the reader's judgment. Capital expenditure in Q2 alone jumped 142% year-over-year to $5.79 billion. Management guided to over $25 billion in total capex for full-year 2026—nearly triple the $8.5 billion spent in 2025, with spending expected to stay elevated over the next two to three years.

Free cash flow flipped to a deficit of $1.09 billion. That's the first negative quarterly free cash flow in over two years, reversing a $1.44 billion surplus from Q1. The math is blunt: at a $25 billion annual run rate, the business burns roughly $2 billion per quarter in capital spending on top of normal operations. The car business that once generated cash is now consuming it.

The money is going to Cortex 2 AI compute clusters, Cybercab tooling at Gigafactory Texas, first-generation Optimus production lines at Fremont, and a semiconductor plant called the Terafab built in partnership with SpaceX. None of these projects generates meaningful revenue today. Tesla has $43.5 billion in cash reserves, which buys time. But cash reserves are an input, not proof.

What the market is still pricing

At $328, the stock trades at 341 times trailing earnings and roughly 410 times forward earnings. The market cap is $1.3 trillion. Consensus EPS estimates for 2025, 2026, and 2027 have been cut 77%, 70%, and 71% respectively since October 2022. A relative P/E valuation using peer automotive multiples implies a fair value near $24—meaning the stock trades at roughly 14 times its peer-justified price.

The AI/robotics narrative is doing all the heavy lifting. Here's where that narrative actually stands. Tesla's robotaxi network has accumulated nearly 2.5 million cumulative paid miles, of which just over 380,000 have been unsupervised—across seven U.S. metro areas, none of them operating in dense city centers. Waymo has driven more than 200 million fully autonomous rider-only miles. FSD subscriptions grew 56% to 1.48 million, which is strong, but each subscription still requires a human driver behind the wheel. Cybercab production just began at an annual capacity of 125,000 units, and Musk himself said the vehicle needs more chassis-specific driving data before deployment can scale.

AInvest's aggregate signal sits at Hold, with a fundamental rating near the ceiling of its scale. That's the market's way of saying the business looks good in isolation while the valuation says something else entirely.

The three-year math

The bear case over a three-year horizon doesn't require the AI bet to fail entirely. It only requires it to take longer than the balance sheet can comfortably fund while the car business delivers margins at their new floor.

On the auto side: even if deliveries hold at Q2's 480,000-vehicle run rate—that's roughly $105 billion in annual automotive revenue. At a 5% operating margin, the car business generates about $5.25 billion in operating income. At 7%, roughly $7.4 billion. Apply a generous 15 times earnings multiple to the resulting $3-4 billion in net income, and the car business alone is worth $45-60 billion. That's 3-5% of the current market cap.

The remaining $1.2+ trillion must come from robotaxis, FSD software, humanoid robots, and energy storage. Energy is the most credible component—13.5 GWh deployed in Q2, up 40% year-over-year—but even aggressive assumptions only push that segment into the $20-30 billion range.

Which leaves robotaxis and Optimus to justify the vast majority of the valuation. If those businesses don't reach meaningful scale within three years, and the car business can't raise margins above the low single digits while burning $25 billion annually in capex, the stock faces a reckoning that has nothing to do with sentiment and everything to do with cash.

What changes the view

Two things would narrow the gap between price and proof. First, free cash flow returning to positive territory on its own, without the car business carrying the AI spend. That would mean margins are recovering despite capex, and the operating engine is strong enough to fund growth. Second, robotaxi milestones that move beyond pilot zones: sustained unsupervised deployment in dense urban areas, intervention rates that beat Waymo's at scale, and actual trip-margin economics that show the ride-hailing model can be profitable per mile.

Neither of those things is visible yet. The 10% weekly growth rate in robotaxi mileage is promising in isolation, but 380,000 unsupervised miles is a small base. The Cybercab production line is real—125,000 annual capacity is a meaningful number compared to competitors building at tens of thousands—but Musk acknowledged the vehicle needs more data before scaling.

The tripwire

Discipline over ego. The condition that invalidates a bullish view is straightforward: if free cash flow remains deeply negative through 2026 and Q1 2027, and operating margins don't recover to at least 3%, the business has structurally lost its ability to self-fund while the AI projects remain unfunded bets. At that point, a $1.3 trillion market cap requires faith in outcomes that the financials haven't started to confirm.

The opposite is also true. If margins hold near the 5% mark despite the capex surge, and FCF stabilizes, the market may be wrong about the near-term pain. But the proof path has to show up in the cash flow first. Simple forward multiples on the car business alone don't get you to $328. They get you somewhere closer to $30. The remaining $300 must be earned.

I can be wrong again. The robotaxi rollout could accelerate faster than the current trajectory suggests. But the setup right now is a company burning $25 billion a year on unproven technology while the core car business trades at a single-digit operating margin. That isn't an inflection. It's a bet. And bets at this valuation can consume cash reserves that look large until they don't.