Tesla Beat Estimates While Profit Sank 43% — a Software Beat
Tesla Beat Estimates While Profit Sank 43% — a Software Beat
Tesla beat Wall Street on Wednesday, and lost 43% of its profit in the same quarter. Read that sentence twice — both halves are true, and together they tell you everything about what this company has quietly become. Revenue landed at $28.046 billion, roughly flat with last year, past LSEG's $27.88 billion consensus (other data shops printed $27.78–$28.34B, so treat the size of the "beat" as an estimate, not gospel), and non-GAAP earnings of $0.48 a share cleared the $0.43–$0.45 range — even as GAAP net income collapsed to $787 million, down 43% from $1.37 billion a year ago. GAAP operating margin: 2.6%.
Here is the twist that matters more than any of those numbers: Tesla's car business barely grew, and deliveries actually shrank. More than 75% of new North American buyers added Full Self-Driving to their cars, active FSD subscriptions jumped 67% year over year to 1.74 million, and the services business — software, supercharging, insurance, parts — surged 33% to a record $4.605 billion. Tesla didn't beat estimates by selling more cars. It beat them by becoming something closer to a software company wearing a car company's clothes.
And on the earnings call, Elon Musk barely wanted to talk about any of it. He wanted to talk about robots, robotaxis, and — when the subject turned to SpaceX — to talk about absolutely nothing at all.
Photo: Tesla’s Optimus humanoid robot on display at a Tesla store in Westfield Century City (September 2024). Tesla told investors on Wednesday that the first production-quality bots have started rolling off the line in Fremont. Photo: Alexis Doine, CC0, via Wikimedia Commons.
The Beat Came From Software, Not Cars
Start with the actual car business, because Musk didn't. Tesla delivered 486,532 vehicles — comfortably above the ~461,100 analysts expected, but still down about 2% from 497,099 a year ago. Model 3 and Y made up 478,237 of those; the "other models" bucket, which includes the Cybertruck, managed just 8,295, down from 12,364 in the second quarter. Production was 464,391, up 4% year over year — which means Tesla delivered 22,141 more cars than it built, its second straight quarter of drawing down inventory. Days of supply fell to 11 from 15.
We checked the quarter's inventory arithmetic against management's own claims: the same shareholder letter that boasts of Tesla's largest passenger-vehicle backlog in four years also confirms the second consecutive quarter of deliveries beating production. Drawdowns clear lots; backlogs build them. You can't celebrate both at once without an explanation, and Tesla didn't offer one.
What rescued the quarter was the subscription line. More than 75% of new North American deliveries included FSD — up roughly 20 percentage points sequentially — and active FSD subscriptions hit 1.74 million, up 67% year over year, with subscriptions now accounting for half of customers for the first time. Services and other revenue jumped 33% to $4.605 billion, a record gross profit of $734 million at 15.9% margin. Automotive gross margin excluding regulatory credits hit 18.1%, beating the 16.7% analysts expected. Musk, who said on the call that "the biggest driver of demand is full self-driving... we are heading towards significant growth," spent the whole quarter being right about that in a way that says nothing good about car demand: buyers aren't showing up for the cars, they're showing up for the software in them.
One caveat the bulls should swallow whole: regulatory credits, the zero-cost revenue Tesla collects from other automakers, collapsed to $113 million from $417 million a year ago — a 73% cliff. The margin beat survived it, but the tailwind that padded so many Tesla quarters is gone.
The Cash Flow "Beat" Was Really an Underspend
The other half of the beat is even less flattering. Free cash flow swung to $1.531 billion from negative $1.09 billion in the second quarter — but Wall Street had expected a cash burn, anywhere from negative $4.3 billion (WSJ) to negative $4.92 billion (Reuters). Tesla didn't generate that cash by earning it; it generated it by not spending it. Capital expenditure was $4.73 billion against roughly $7.5 billion expected, down from $5.79 billion in Q2. Thomas Monteiro of Investing.com put it in one clean sentence: "Tesla's surprisingly positive cash flow quarter did not come from stronger profits, but from spending billions less than planned."
Then came the line that should worry anyone pricing Tesla as an AI company. The company cut its 2026 capital-expenditure guidance to about $20 billion on the call. CFO Vaibhav Taneja added that Tesla has secured the capacity to borrow up to $30 billion but plans no draw this calendar year. Ryan Lee of Direxion read it plainly: "Lower capex suggests less of an 'all-in' approach to the AI arms race than one would expect from Elon."
Meanwhile, operating expenses jumped 39% year over year to $4.78 billion, with R&D up 69% to $2.76 billion, driven by Super Intelligence, Optimus, and the Cybercab. So Tesla is spending more on everything except the one thing that proves the buildout is accelerating. And tucked inside the GAAP net income figure was a quiet helper: about $290 million of unrealized bitcoin gains against a $142 million unrealized loss on Tesla's SpaceX holding — a net $146 million that accounts for roughly a fifth of the quarter's reported profit. The "profit" line, such as it is, is part crypto mark-to-market.
Goodbye Model S and X — Hello, the Robot Line
The single most symbolic line in the quarter didn't come from the financials at all. Tesla has stopped production of the Model S and Model X at Fremont, and their lines are being replaced by the company's first large-scale Optimus production line. The two halo sedans that built the brand are being succeeded, on the same factory floor, by humanoid robots. No outlet led with it, which is strange, because it is the cleanest possible statement of what Tesla thinks it is now.
Optimus production has started in Fremont, yielding what the shareholder letter calls "the first production-quality bots that we have manufactured." Musk then did something uncharacteristic: he tempered expectations, in the same breath as promising the moon. "I would expect the AI and electromechanical design and the reliability and usage life issues to be solved by the end of next year," he said, before adding: "Time to expect when we will be delivering Optimus to companies outside of Tesla will be end of 2027. I do want to temper expectations around Optimus."
For the record, that is not next year. When headlines shorthand it to "robot sales next year," they mean the very end of 2027 — and even then, only to outside companies. Musk paired the caution with his usual swagger: "I will be stunned if anyone has a robot competitive with Optimus when we ship in 2027." Read the two sentences together and you get the Musk dialectic in miniature: the schedule is always later than it sounds, and the confidence is always total.
Photo: An Optimus humanoid robot mingles with guests at Tesla’s 10/10 robotaxi event at Warner Bros. Studios, October 2024. On Wednesday’s call, Musk said design and reliability issues should be solved by the end of 2026, with external deliveries at the end of 2027. Photo: Steve Jurvetson, CC BY 2.0, via Wikimedia Commons.
The Cybercab Isn't a Car — It's a Fleet Musk Wants to Own
The robotaxi operation is scaling fast and it is not, structurally, a car business. Cybercab launched into the Austin robotaxi fleet in early September; paid rides began September 3, and the fleet has grown roughly tenfold in about a month, from under 40 vehicles to more than 300, with Dallas next. Tesla now runs robotaxis in six cities — Austin, Dallas, Houston, Miami, Orlando, and Tampa — from early morning to late evening, and unsupervised robotaxi miles grew eightfold quarter over quarter to nearly 2 million, on sustained double-digit weekly growth. "Robotaxi miles driven scaling exponentially — hard to overstate at this point," Musk said.
Then came the disclosure that reframes the entire Cybercab thesis. The car is designed with 50% fewer parts than any Tesla ever made — vehicle engineer Lars called it "the easiest product to scale" — with installed Texas capacity already above 125,000 units a year. But Musk doesn't plan to sell them to you. "Cybercab is going to be more of a transport as a service as opposed to a car that you would want to own," he said, "the vast majority of the Cybercab fleet will be Tesla-owned." Individual purchase? "2027 ish," per Lars. That is a transport-as-a-service company talking, not an automaker — and it explains why Musk also declared: "Unless we are stopped by government, I think Tesla will do the majority of passenger miles." (We tracked the Cybercab's first paid rides in Austin earlier this month; the "vast majority Tesla-owned" disclosure is new.)
FSD Slipped Again — Quietly This Time
For anyone keeping count of autonomy timelines, the call delivered another quiet delay. Ashok Elluswamy, Tesla's VP of AI software, said the company expects to start adding private customer cars to the robotaxi fleet "somewhere around Q2 next year." Back in April, Musk had said the consumer unsupervised version would "probably" arrive in Q4 2026. The date has moved, again, by two quarters — and nobody on the call made a headline of it. Elluswamy said it the way people do when they know the track record: "we suspect," he said, choosing his verb like a man walking on ice.
Overseas, FSD is now approved in nine EU countries, with more than 100,000 customers and over 160 million kilometers logged; Tesla has agreed to rename the product "Tesla Assisted Driving" in Europe, with an EU vote due in December and Musk hoping for EU-wide approval by the first quarter of 2027. Europe, notably, is where Tesla just posted record deliveries — Canada, Australia, Japan, and Taiwan also set delivery records, and Berlin hit record production — a genuine bright spot in a quarter where the US lost its tax credits and demand softened.
"Tap Dancing Through a Minefield"
The question everyone was waiting for got the most famous dodge of the evening. Special committees have now been formed at both Tesla and SpaceX boards to evaluate a combination — revealed the day before on SpaceX's own call, where Musk said there was "a lot of merit to combining the companies." Asked about it on Tesla's call, Musk refused to engage: "I think I said everything that should be said yesterday. Tap dancing through a minefield there."
What did emerge: Musk is not on either committee, and the deal needs 50% of Tesla shares — "quite a high hurdle." (We covered the earlier October merger speculation in detail when it first surfaced; the special committees and Musk's on-the-record refusal are the new facts.) Musk did, however, want to talk about Terrafab, the joint Tesla-SpaceX chip plant: "not going to be less than $10 billion of capital equipment... north of $100 billion long-term." "The Terrafab will be the most valuable building in the world," he said. Real production is expected in late 2028.
Also on the Call
The rest of the deck reads like a company placing bets across the board. The Tesla Semi is in volume production in Nevada, with the top variant rated at 500 miles at 82,000 lbs. Megafactory Texas began production — 50 GWh planned — and the first Megapacks have shipped with LFP cells made in Nevada. Energy storage order rates and backlog are at their highest-ever levels, even as energy revenue fell 16% year over year on tariff-related customer refunds, and storage deployments of 13.7 GWh came in below the ~15.9 GWh expected. Energy gross margin jumped to 33.1% on tariff refunds and warranty true-downs, per Taneja.
Tesla produced and delivered its 10 millionth vehicle, and Musk set the 2030 target: five million vehicles a year — a goal he says the company has "a good shot at achieving." Training compute is targeted at 400 MW by the end of the year. And the next-generation Roadster — unveiled on October 15, as we reported — is slated for sale in 2028.
What the Stock Did: a Pop, Then a Fade
The market's reading came in two acts. Shares closed the regular session up about 3%, near $386.47, popped another 1–2% at the release — and then faded during the call to down 3.5% after hours, ending the evening near $377.15. The stock is down roughly 14% year to date with a market cap near $1.4 trillion. The trajectory is the tell: the numbers bought a pop; the story bought a sell.
Our Take: Stop Reading the Beat, Start Reading the Balance Sheet
Here is our judgment, plainly stated: Tesla's third quarter was colossally misread by anyone who stopped at the word "beat." A company that delivers fewer cars than a year ago, posts a 2.6% operating margin, and funds its "positive" cash flow by spending $2.8 billion less than planned is not having a great quarter — it is having a great narrative quarter, and there is a difference. The autonomous driving timeline has slipped, yet again, from "probably Q4 2026" to "somewhere around Q2 2027," continuing a decade-long record of shoddy autonomy forecasting that the stock somehow never prices in. And a robot that doesn't reach outside customers until the very end of 2027 is doing heavy lifting for a $1.4 trillion valuation today.
None of that means the pivots are fake. A 75% FSD attach rate is a genuine transformation; 1.74 million subscriptions is real recurring revenue; a tenfold robotaxi fleet in a month is real execution; and turning the S/X lines into an Optimus line is exactly the kind of ruthless reinvention that justifies a growth multiple. But notice the pattern: every bull case is dated 2027 or later — Optimus deliveries, Cybercab-for-sale, unsupervised FSD, Terrafab silicon — while every hard number that matters in 2026 (deliveries down, GAAP profit nearly halved, credits cliff, capex guidance cut) points the other way. The botched habit of promising the future to paper over the present is Tesla's oldest tradition. This quarter's numbers just made it easier to see.
Sources: Tesla Q3 2026 shareholder letter and 8-K (Exhibit 99.1); Q3 2026 earnings-call transcript live-reported by Shacknews' Asif Khan; Reuters; The Wall Street Journal; CNBC; TechCrunch (Kirsten Korosec); Electrek; TeslaNorth; TradingKey; TeslaNewswire; The Circuitry; Barron's; AskTraders; ZeroHedge; AllMotorNews.
Tesla–SpaceX merger: your questions, answered
Will Tesla merge with SpaceX? On SpaceX's investor call on October 20, Musk called a merger "a good idea" with "a lot of merit"; both boards have formed special committees, but no terms are agreed and a deal needs 50% of Tesla shareholders — "quite a high hurdle," Musk said. On Tesla's own call he added nothing: "I think I said everything that should be said yesterday."
Why did Tesla cut its 2026 spending plan? Tesla now guides around $20 billion of capex for the year, down from the revised $25 billion target; third-quarter capex of $4.73 billion missed the $7.95 billion analysts expected. Analysts called it a positive only if the ramp is delayed, not cancelled.
Why did the stock fall after an earnings beat? Revenue of $28.05 billion beat the $27.9 billion estimate and EPS of $0.48 beat $0.43, but auto revenue fell 3%, energy fell 16%, and free cash flow plunged 62% to $1.53 billion; shares fell 3.2% in extended trading after a 3.4% regular-session gain to $386.47.
Update, Oct. 22
The merger talk moved past rumour the night before earnings. On SpaceX's investor call, Musk said combining Tesla and SpaceX would be "a good idea" with "a lot of merit," and disclosed that both boards have formed special committees to explore terms — none agreed yet. Any deal needs approval from 50% of Tesla shareholders, which Musk called "quite a high hurdle." Asked again on Tesla's call, he shut it down: "I think I said everything that should be said yesterday." Separately, Tesla cut its full-year capex plan to about $20 billion from the revised $25 billion target after third-quarter capex of $4.73 billion undershot the $7.95 billion consensus; free cash flow fell 62% year over year to $1.53 billion. The stock fell 3.2% in extended trading.
Comments
Post a Comment