Flat revenue, halved operating profit
Start with the two numbers that do not sit comfortably together. Gross profit actually improved: total GAAP gross margin rose to 19.6%, and automotive gross margin excluding regulatory credits improved to 18.1%, up from 16.3% in the second quarter and 15.4% a year earlier, even as automotive revenue declined.
Then the spending arrived. Operating expenses rose 39% to $4.776 billion, and R&D alone went from $1.630 billion to $2.761 billion in a year — a $1.131 billion increase that exceeds the entire $440 million year-over-year improvement in gross profit. But R&D is not the whole explanation, and the deck does not present it as one. Tesla attributes the operating-income decline to a combination: higher spending on SI and other R&D projects, stock-based compensation (including the 2025 CEO Performance Award) and SG&A, lower regulatory credit revenue, and lower energy gross profit — partly offset by improved vehicle mix, higher ancillary sales driven mainly by FSD subscriptions, growth in services gross profit and a positive currency effect. The deck does not disclose how much of the R&D increase belongs specifically to autonomy or robotics.
That is a choice, not an accident, and the deck is open about what the money is for: SI training compute Tesla plans to take to 400 MW by the end of 2026, a Cybercab fleet, an Optimus line, a Semi line. But a choice still has a price, and this quarter the price was more than half the operating profit of a company whose revenue did not grow.
The supporting cast points the same way. Regulatory credit revenue, once a reliable profit top-up, fell to $113 million from $417 million. Automotive revenue fell 3% and energy revenue fell 16%. Energy storage deployments nevertheless rose about 10% to 13.7 GWh; separately, Tesla reported its highest-ever energy storage order rate and backlog, a company claim given without a comparable backlog figure in that passage. The growth inside the quarter came from services, up 33% to $4.605 billion — supercharging, insurance, parts — with gross profit at a record $734 million. Useful, profitable, and not the AI story.
The 75% is the strongest number in the deck
Against that, the FSD figures are genuinely hard to argue with, as long as they are read as what they are. Tesla says the FSD (Supervised) attach rate reached over 75% of new vehicle deliveries in North America, a nearly twenty percentage point sequential increase. Active FSD subscriptions reached 1.74 million, up 67% in a year; Tesla's subscription metric counts upfront purchases and monthly subscriptions and excludes free trials. What the deck does not disclose is how much of the inclusion rate came from pricing, promotions or bundling rather than buyers choosing FSD at full price. The 75% measures reported inclusion. It is not, by itself, proof of full-price willingness to pay.
The robotaxi numbers need the opposite handling. Tesla reports nearly two million cumulative unsupervised Robotaxi miles as of the end of the quarter, an eightfold increase in the quarter, with the Austin service now running from 6 a.m. to midnight after Cybercab rides began on 3 September. That establishes expanding activity in one limited service area. It does not establish a national commercial rollout, and the deck discloses no Robotaxi revenue, unit economics or cost per mile with which to judge it as a business.
Optimus sits in the same category. Production has begun in Fremont on Tesla's high-volume lines, and Tesla calls the output its first production-quality bots. There is no unit count, no customer, and no separately disclosed revenue figure attached. A production start is a milestone. It is not yet a business, and this deck gives no basis for pricing it as one.
What to watch
The fourth quarter tests whether the spending curve bends. R&D at $2.761 billion a quarter, capex at $4.732 billion and planned compute expansion all assume the software side starts paying at scale: FSD subscription growth holding after the inclusion jump, robotaxi expanding beyond Austin, and Optimus moving from a production start to countable units. Free cash flow recovered to $1.531 billion this quarter, after a negative $1.092 billion in the second, which buys time. It does not settle the question the deck raises: whether the AI buildout earns before the car business is asked to fund another year of it at this pace.
Sources
- [1] Tesla, Q3 2026 Update, 21 October 2026, Exhibit 99.1 to Form 8-K (SEC filing) — the primary source, read directly for this article: financial summary and statements (revenue, margins, operating income, R&D, segment revenue, regulatory credits, cash flow, cash balance), operational summary (deliveries, production, storage deployments, FSD subscriptions and the metric's definition), and the SI, Robotaxi and Optimus commentaryRead source
- [2] TeslaNorth, "Tesla Q3 2026 Earnings: US$28B Revenue as FSD Attach Rate Tops 75%", 21 October 2026 — secondary report of the same Update deck, used as contextRead source
- [3] TradingKey, "Tesla Q3 Operating Profit Plunges 56% YoY, Dragged Down by Higher Operating Expenses Including R&D", 21 October 2026 — secondary report of the same Update deck, used as contextRead source