Tesla Finally Puts a Number on FSD — and a Hundred Billion on the Robot
A demand problem that turned into a supply problem, an energy margin that leaned on a tariff refund, and an Optimus timeline that slipped another year while the only question the tape cares about stayed off the table.
TSLA · Earnings Call · 2026-10-21
Six quarters of arguing about demand ended on one sentence. In July 2025, Elon Musk told investors to brace for “a few rough quarters” — Elon R. Musk, Chief Executive Officer · 2025-07-23 as U.S. incentives lapsed, and Vaibhav Taneja spent his airtime on FSD education — noting that “half of Tesla owners who could use it haven't tried it even once.” — Elon R. Musk, Chief Executive Officer · 2025-07-23 On Wednesday the same CFO opened with two consecutive quarters above 480,000 deliveries, the “highest backlog in four years,” — Vaibhav Taneja, Chief Financial Officer · 2026-10-21 and deliveries "constrained primarily by production."
That is the real pivot in this report. Tesla is no longer arguing about whether people want the car. It is arguing about how fast it can build one, and how fast regulators will let it drive itself. The tape has not yet agreed: the shares are down about 6% over the last 90 trading days, 21% below the December peak of $489.88, and the print lands with the stock already in a mild drawdown.
The attach rate just became a P&L line
The most consequential number in the quarter was not one Tesla printed. It was the one it nodded at. Pierre Ferragu estimated FSD annual recurring revenue "approaching $1 billion" and asked whether that order of magnitude was right. “Yeah, I think the numbers which you shared are correct,” — Vaibhav Taneja, Chief Financial Officer · 2026-10-21 Taneja replied, adding that Tesla will "start sharing more metrics around FSD, not just the revenue." That matters because the attach rate stopped being a marketing statistic this quarter. Over 75% of North American deliveries signed up for paid FSD at delivery, roughly 20% above Q2, and the paid base crossed 1.7 million — half of it subscription for the first time. Three quarters ago the same disclosure was a 25% penetration bump and a lot of metaphors about a cat that can sing and dance. Now it is a recurring-revenue line, with the churn rate conspicuously deferred. And yet: ~$1 billion of ARR against $28.2 billion of quarterly revenue is a rounding error on the income statement. At 16x price-to-revenue, nobody is paying for the annuity. They are paying for the option, which is exactly why the timeline slippage later in the call is the part that should worry holders.The energy margin is a tariff refund wearing a costume
Energy storage posted its second-best deployment quarter at 13.7 gigawatt-hours and gross margin jumped from 20.4% to 33.1% — the kind of number that gets circled. Taneja's own attribution: “primarily attributed to tariff refunds recognized in the quarter and warranty true-downs resulting from alignment with our vendor.” — Vaibhav Taneja, Chief Financial Officer · 2026-10-21 A one-time cash recovery and a supplier reimbursement, in other words, dressed as operating leverage. Automotive tells the same story in miniature. Margins ex-credits went from 16.3% to 18.1% on "higher mix of deliveries in APAC" — geography, not cost — while management conceded that "commodity price increases and interest rate changes continue to add to our cost." Group gross margin in the last filed quarter sat at 16.8%; nothing in this call suggests that structural number has moved. Tariff refunds were the single hottest keyword in the entire market two quarters ago, and they keep being booked: Hasbro and Investor AB both flagged them in this week's batch of reporters. Tesla is riding a market-wide, one-time cash flow. To management's credit they did not hide it, and the proceeds are going back out the door into U.S. capacity. But an investor extrapolating 33% energy margins is extrapolating a refund.A $20 billion year, a $30 billion revolver, and a $100 billion blank check
Capex guidance moved to roughly $20 billion, funded "primarily from cash flows from operations." The last filed quarter shows why that framing matters. Capital expenditure was $5.8 billion, up 142%, against negative free cash flow of $2.2 billion and R&D of $2.4 billion, up 49% — about 8% of revenue, a ratio this company did not carry a few years ago. Then the Terafab. Musk put real numbers on it for the first time, and they are not small:He then conceded the split between Tesla and SpaceX is unsettled — "the sort of relative split is between SpaceX and Tesla in terms of who pays for what at Terafab." Compare that to January, when Taneja answered the same capex question with “we have over $44 billion of cash and investments on the books. So we'll use our internal resources.” — Vaibhav Taneja, CFO · 2026-01-28 Six months later Tesla has secured capacity to borrow up to $30 billion. That is a change of financing posture, not merely of ambition, and it is happening with the equity 21% off its high.It's certainly in the tens of billions... And then it will, it'll be a lot more, it'll probably be north of $100 billion long-term.
The S-curve keeps moving right, and the merger stays out of scope
Optimus is the tell. In October 2025 Musk described “a million units Optimus production line, hopefully with the production start towards the end of next year.” — Elon Musk, Chief Executive Officer · 2025-10-22 In January he said Tesla “wouldn't expect to have any kind of significant Optimus production volume until probably the end of this year.” — Elon Musk, CEO · 2026-01-28 Wednesday: early production has begun in Fremont, but the line will not reach its 20,000-units-a-week annualized rate for “at least a year, maybe a little longer,” — Elon Musk, Chief Executive Officer · 2026-10-21 and customer deliveries land "around the end of next year." Same product, same conviction, one more year. The bottleneck is now named out loud: the hand. Electromechanical dexterity entered this company's top keywords for the first time, straight from the engineering lead, and Musk's own explanation — "humans have self-repair... robots don't" — was the least promotional thing he said all evening. Meanwhile the one catalyst that could plausibly re-rate a $1.5 trillion market cap was ruled out before the first analyst spoke: no questions on a SpaceX merger.Robotaxi is the one place where rhetoric and arithmetic agree. Unsupervised autonomous miles grew 8x in the quarter, the Cybercab fleet is up tenfold since launch, and management's own markers — "noticeable" in Q2, "a big deal" by Q3 or Q4 of 2027 — are refreshingly far out. Energy, by contrast, leans on data-center appetite the tape has stopped paying for: the market's AI data center basket is down roughly 5.6% over 90 days with 48 decliners against 6 advancers, even as Tesla says it is "almost sold out of all Megapacks for 2027." That is the shape of this report: an operations story that finally turned, a margin story propped up by a refund, a capital story that just acquired a hundred-billion-dollar question mark, and a robot story that slid a year — all while the stock sits 6% lower over three months. Tesla's demand problem is solved. Its credibility discount is not.I think I said everything that should be said yesterday, tap dancing through a minefield there.