EPS beat, cash flow back in the black — Tesla's Q3 brings margins home, not yet profit
Beat — above consensusTesla's Q3 revenue of $28.05B was flat year over year and in line with the $27.9–28.1B consensus, while adjusted EPS of $0.48 topped the $0.43–0.45 expected. Auto gross margin ex-credits rebounded to 18.1% from 16.3% in Q2, and with capex of $4.73B coming in well under the ~$7.9B expected, free cash flow swung back to a positive $1.53B. Operating income still shrank 56% to $718M on AI and Optimus spending, and when management cut 2026 capex guidance from $25B+ to about $20B late in the call, the stock reversed to a 2%-plus after-hours decline.
Earnings call · Oct 21 (US) · KST 10/22 06:30
Key Metrics
Segment Highlights
Automotive — margins rebound, revenue slips
Automotive revenue fell 3% to $20.56B. Deliveries of 486,532 (-2% YoY) beat the ~462,000 consensus, but average selling prices kept declining and regulatory credit revenue collapsed 73% to $113M. Tesla said battery cell availability constrained volume and that its passenger-vehicle backlog is the largest in four years. Auto gross margin excluding credits came in at 18.1%, a clear improvement from 16.3% in Q2 and 15.4% a year ago, which the company attributed to a higher APAC delivery mix and growing FSD subscription revenue. Active FSD subscriptions reached 1.74M (+67%), and FSD was attached to more than 75% of new North American deliveries.
Energy — revenue down 16%, margins jump on tariff refunds
Energy generation and storage revenue fell 16% to $2.88B, which Tesla said was partly due to tariff-related customer refunds. Storage deployments of 13.7 GWh (+10% YoY) were roughly level with Q2's 13.5 GWh. Segment gross margin, however, jumped to about 33% from 20% in Q2. CFO Vaibhav Taneja attributed the improvement primarily to tariff refunds recognized in the quarter. Megafactory Texas (planned capacity 50 GWh) began production, Tesla shipped its first Megapacks built with its own LFP cells from Nevada, and the business posted its highest-ever order rate and backlog.
Services & other — up 33%, record profit
Services and other revenue grew 33% to $4.61B, with a record $734M gross profit (+$86M QoQ) and margin climbing to 15.9%. The Supercharger network expanded 15% and 14% to 8,904 stations and 84,465 connectors. This segment was one pillar of the total gross margin's rise to 19.6% (18.0% a year ago, 16.8% in Q2), and it filled much of the revenue gap left by the automotive decline.
Key Quotes from the Call
"Fully autonomous miles driven are scaling exponentially. The Cybercab fleet in Austin has grown by a factor of 10 since launch about a month ago. The vast majority of the Cybercab stuff will be essentially Tesla-owned and transport-as-a-service."
Elon Musk · Robotaxi & Cybercab
"The vast majority of a million-unit production line has arrived in Fremont, but it will take at least a year — maybe a little longer — to get to that annualized rate of 20,000 units a week. There is still quite a lot of work to do on the AI and the robot for it to be truly useful for open-ended tasks."
Elon Musk · Optimus production
"Demand is growing rapidly for our vehicles — it is exceeding our production by a significant margin. We are aiming for 5 million vehicles a year by approximately 2030. The biggest driver of demand is Full Self-Driving."
Elon Musk · Demand & volume target
"We are revising our capex guidance to approximately $20 billion for the current year, funded primarily from our cash flows from operations. During the quarter we secured capacity to borrow up to $30 billion, but we do not plan to make any draws this calendar year. Our capex for the next few years would continue to expand."
Vaibhav Taneja (CFO) · Capex guidance
"Next month we'll be launching in Dallas, and then the same software just scales to lots of other cities. We suspect somewhere around Q2 next year is when we can turn on some of the first customer vehicles — we will be able to release the same software to customers as part of the V15 release."
Ashok Elluswamy (Head of AI) · Unsupervised FSD & Dallas
"I think I said everything that should be said yesterday. (On the SpaceX call a day earlier: both companies have formed special committees, there is a lot of merit to combining them, but treating a merger as a certainty would be overstating it.)"
Elon Musk · SpaceX merger
"Terafab will take tens of billions in capital equipment through early 2028, and north of $100 billion long term. Tesla and SpaceX still have to work out the relative split of who pays for what. Real production comes in late 2028 — lightning fast by fab standards."
Elon Musk · Terafab capital
"The software for Cybercab and Model Y is the same, and it is coming to customers soon with V15. What remains is making sure customer Model Ys can operate within our geofences and fleet management, and we suspect somewhere around Q2 next year is when we can turn on some of the first customer vehicles."
Ashok Elluswamy · FSD Unsupervised in customer cars
Guidance & Outlook
- 2026 capex guidance cut from 'more than $25B' to about $20B, funded primarily from operating cash flow; the newly secured $30B borrowing capacity will not be drawn this year. Capex becomes 'more meaningful' in 2027 and 'noticeable in the tens of billions' in 2028 (Musk).
- Cybercab is Austin-only today but will reach other cities before year-end (Taneja). Model Y Robotaxi launches in Dallas next month (Elluswamy). Tesla is trying to get Cybercabs to customers by the end of 2026 (Lars Moravy). The Robotaxi fleet should start contributing to financials in 2027.
- Unsupervised FSD for some customer vehicles is expected around Q2 2027 via the V15 update. The $0.30-per-paid-mile target is not yet met, but management sees a path as utilization improves (Taneja).
- Optimus: the Fremont line has begun producing the first 'production-quality' bots. Reaching a 20,000/week run rate will take a year or more; deliveries to outside customers start around end-2027, beginning with dangerous or repetitive industrial work. AI5 should go into Optimus around April 2027 if on plan (~10x AI4 compute).
- SI training compute to grow from ~230 MW in September to 400 MW by December. Austin Fab equipment commissioning planned for 2027. Semi entered volume production in Nevada (25,000/yr installed capacity); Cybercab installed capacity is 125,000+/yr. Targeting ~5 million vehicles a year by around 2030.
- Official outlook: focus on maximum factory utilization, with deliveries depending on demand, supply chain readiness and allocation to Tesla's own fleet. Hardware profits are expected to be accompanied by an acceleration of SI, software and fleet-based profits. The trend of growing opex continues until the new products are commercialized (Taneja).
Market Reaction
The stock closed the regular session up 3.41% at $386.47 and initially added gains toward $390 after the release. Once the capex guidance cut ($25B+ to about $20B) landed late in the call, it gave those gains back and turned to a 2%-plus decline — reported down more than 2% as of 6:09 p.m. ET and at $377.88 (-2.22%) after hours. The EPS beat and positive free cash flow were welcomed, but the lower spending was read as a signal of slower Optimus and Cybercab ramps. The stock closed down 3.08% at $374.58 on October 22.
Deep Dive
Earnings quality — half the 'beat' is money not spent
Adjusted EPS of $0.48 beat the $0.43–0.45 consensus but was still below last year's $0.50, and GAAP net income was just $787M (-43%), or $0.22 per share. Operating income fell 56% to $718M (2.6% margin) from $1.62B — gross profit actually rose $440M to $5.49B, but operating expenses swelled 39% to $4.78B, with R&D up 69% to $2.76B. Net income also carries a $288M unrealized bitcoin gain and a $142M unrealized loss on the SpaceX stake.
Much of the 'surprise' came from the cost side. Capex of $4.73B was only ~60% of the ~$7.9B expected, flipping free cash flow from an expected -$4.5–4.9B to a positive $1.53B, and adjusted EBITDA of $3.90B topped the $3.66B estimate. Revenue has now been flat at ~$28.0B for two straight quarters. The honest read is a quarter where investment slipped, not one where earnings power improved.
Auto margin & demand — the delivery beat did come with margin
The preview's first question was whether the 486,532-unit delivery beat came without price cuts. The answer is a qualified yes. Auto gross margin ex-credits rebounded 180bp to 18.1% from 16.3% in Q2, topping the Zacks estimate of 17.3%. Tesla credited a higher APAC delivery mix and FSD subscription revenue (1.74M active subscriptions, +67%). FSD attached to more than 75% of new North American deliveries signals structurally rising software revenue per car.
Still, ASPs kept falling and automotive revenue slipped 3% to $20.56B. The company said battery cell supply constrained volume and that its passenger-vehicle backlog is the largest in four years, though Electrek noted that falling ASPs are an odd fit for a supply-constrained automaker. Musk said demand 'is exceeding our production by a significant margin' and set a target of ~5 million vehicles a year by around 2030. Whether ASPs 'improve' in Q4, as Taneja suggested, will decide if the margin rebound sticks.
Energy — revenue down, margin up: two sides of a tariff refund
Energy revenue of $2.88B (-16%) was as weak as the preview feared and well below the ~$3.9B analysts expected; Tesla said tariff-related customer refunds reduced revenue. Deployments of 13.7 GWh (+10%) were level with Q2.
Segment margin, by contrast, jumped from 20% in Q2 to about 33%, which Taneja attributed primarily to tariff refunds recognized in the quarter — so part of this margin is one-off and should not be extrapolated. The structural signals are elsewhere: Megafactory Texas (50 GWh) started production, the first Megapacks with Tesla's own Nevada LFP cells shipped, and order rate and backlog hit record highs. We found no reported comment on the mid-October departure of the energy business head.
Cash flow & capex — back in the black, and the question '$20B' raised
Free cash flow swung from -$1.09B in Q2 to a positive $1.53B. Operating cash flow of $6.26B was flat year over year; the difference came from capex falling to $4.73B from $5.79B in Q2. Cash and short-term investments rose $1.9B to $45.4B.
Late in the call, Taneja cut full-year capex guidance from 'more than $25B' to about $20B. With $13.0B spent through nine months, that implies roughly $7B in Q4 — on par with Tesla's largest quarter ever. Musk said capex 'is actually ending up being a little lower than we anticipated,' but the market read it as a question mark over the pace of the Optimus and Cybercab ramps, and the stock gave back its after-hours gains. Tesla secured $30B of borrowing capacity it says it will not draw this year, and flagged expanding capex in 2027 and spending 'in the tens of billions' in 2028. Positive free cash flow is welcome, but if the price is delayed investment, opinions will split.
AI execution check — Robotaxi has numbers now, Optimus needs time
Robotaxi produced its first 'exponential' numbers: cumulative unsupervised miles grew eightfold in the quarter to nearly 2 million, the Austin Cybercab fleet grew tenfold in a month, and hours expanded to 6am–midnight. The Say questions got answers — expansion beyond Austin is 'Dallas next month' (Elluswamy) and 'other cities before year-end' (Taneja); Cybercab customer deliveries are 'targeted by year-end' (Moravy), though Musk stressed the vast majority will be Tesla-owned transport-as-a-service. On the $0.30-per-paid-mile target the answer was 'not yet, but we see a path,' and unsupervised FSD for customer cars slipped again, to Q2 2027.
Optimus, by contrast, reset expectations lower. Fremont has started building the first production-quality bots, but reaching 20,000 a week will take 'at least a year, maybe longer,' and deliveries to outside customers are pencilled for around end-2027. Asked the Say question on manufacturing cost per unit, Taneja said only that costs are 'trending closer' to where Cybercab's will end up, with no figure. On the SpaceX merger, Musk had said on the SpaceX call a day earlier that special committees are formed, the combination has 'a lot of merit,' and certainty 'would be overstating it'; on this call he closed with 'I said everything that should be said yesterday.' Through the annual meeting (by Dec. 6), the merger remains the stock's biggest swing factor.
What We Were Watching For
- Automotive margin recovery — auto gross margin ex-credits fell to 16.3% in Q2, and Zacks expects 17.3% for Q3. Whether the delivery beat (486,532, +5.3% vs. company-compiled consensus) came without heavier discounting is the first test.
- A repeat of Q2's earnings shock? — Q2 adjusted EPS of $0.33 badly missed consensus ($0.50–0.55) and the stock fell 14.52% the next day. Q3 EPS estimates have been cut ~18% over 90 days, from $0.55 to $0.43–0.45, so the bar is lower.
- Cash flow and capex — Q2 free cash flow turned negative at -$1.1B on $5.8B of capex. With roughly $25B of 2026 AI, robotaxi and Optimus spending, investors want to see FCF back in positive territory.
- Robotaxi economics — the Cybercab fleet passed 300 in Austin within a month of its September launch, and service hours now run to midnight. Goldman says robotaxi and Optimus updates will matter more to the stock than EPS — revenue per paid mile and the next cities (Dallas) are the key questions.
- Energy margins — Q3 storage deployments were 13.7 GWh. Watch whether energy margins recover from Q2's sharp drop, and whether management addresses leadership after the mid-October departure of the energy business head.
- Any word on a SpaceX merger — Barclays thinks a deal could be announced before the annual meeting (by Dec. 6), but the company has avoided official comment. How far management goes on the call could drive volatility.
Investor & Press Questions
When will Cybercab operations expand beyond Austin?
Retail investors (top Say question)When will customers be able to purchase a Cybercab?
Retail investors (top Say question)How much will it cost to manufacture each Optimus robot?
Retail investors (top Say question)Call Q&A
How does the Cybercab fleet scale beyond Austin, and what is the regulatory path outside Texas with NHTSA's safety review still open?
Tesla launched in Austin only because it had a limited number of cabs; nothing fundamentally confines Cybercab to Austin. Dallas launches next month, and the same software then scales to other cities alongside the Model Y Robotaxis. Cybercab's AI was trained largely on data from the Model 3/Y platforms, so it is scaling well without a huge fleet, and the goal is to 'swarm' Austin. The NHTSA review and the regulatory path outside Texas were not addressed.
When will the Cybercab be available for customer purchase and ownership?
Moravy said Tesla aims to get Cybercabs to customers by the end of 2026, though all output so far goes to Tesla's internal Robotaxi fleet; it will start small this year to learn about delivery, ownership and fleet management in customer hands, then scale. Musk added that Cybercab will be offered primarily as transport-as-a-service with the vast majority Tesla-owned, while the Model 3, Y and Roadster are the cars customers would typically own.
Where do the cost per paid Robotaxi mile and the cost per Optimus unit stand?
CFO Taneja said Tesla is not yet at its $0.30 per paid mile target but sees a path there through scaling operations: utilization in its main markets is improving week by week, and pickup, drop-off and cleaning costs are being refined as the fleet grows. Cybercab is only in Austin today and will reach other cities before year-end. On Optimus, internal cost tracking is going well as the design matures and costs are trending closer to where Cybercab costs are expected to land, but with design iterations ongoing it is too early to call the target met. No dollar figure was given.
When can we expect FSD Unsupervised in customer cars?
The software for Cybercab and Model Y is the same and is coming to customers soon as part of the V15 release. What remains is ensuring customer Model Ys can operate within Tesla's geofences and fleet management so they can run fully autonomously; the first customer vehicles could join the Robotaxi fleet around Q2 2027.
What are Terafab's capital needs and key milestones?
Musk said Terafab will need 'tens of billions' in capital equipment through early 2028 and could exceed $100 billion over the long term; how the cost is split between Tesla and SpaceX is still to be determined. Taneja said most orders for the first set of research-fab equipment have already been placed, and Musk added that real production in late 2028 would be 'lightning fast by fab standards.' Prepared remarks targeted commissioning in H1 2027, a first test chip around May–June 2027 and volume chips in Q3 2028.
With unsupervised miles growing ~15% weekly toward ~10 million by year-end, how does that translate into revenue and gross profit — and what is the Cybercab line's ramp curve?
Taneja said continued exponential growth would make Robotaxi a significant part of revenue and profit in 2027; Musk said it won't move the needle this quarter, should become noticeable around Q2 2027 and could be a big factor by Q3–Q4 2027. No miles-to-gross-profit figure was given. On the Cybercab line (listed capacity above 125,000 units a year), Taneja said the design is easy to scale and the main constraint is regulatory approval; Moravy said the listed figure reflects the current running rate, actual capacity is significantly higher, and output currently matches the internal Robotaxi fleet's needs.
By my math FSD annual recurring revenue is approaching $1 billion — is that right, and what about retention and churn?
Taneja said the ~$1 billion figure is about right and that Tesla will share more FSD metrics as the business scales. Churn and retention figures will come later, since free trials and new-country launches muddy the numbers. He noted that the 75%-plus of North American deliveries signing up for paid FSD at delivery (up about 20% from Q2) comes with every North American delivery including one free month of FSD. Globally there are over 1.7 million paid FSD customers, and for the first time 50% are subscriptions.
Robotaxi pricing is below rideshare — is lower pricing a long-term plan to take market share?
Taneja said rideshare carries a human driver while Tesla builds its own vehicles, so the cost basis is fundamentally different, and reiterated the goal of taking over all forms of transportation. Musk said governments could block Tesla to protect domestic industries, but wherever it is allowed to operate freely he predicted, 'confidently,' that Tesla would handle a majority of passenger miles. No pricing figures were given.
I estimate paid FSD attach in North America at 30–40% — could competitors partnering with Nvidia or Qualcomm pressure FSD pricing, and what about China?
Taneja said the software problem is not easily solved, and Elluswamy added that the first 90% is easy to demo but the long tail is extremely hard, so he doubts competitors will match Tesla for several years. Showroom customers increasingly ask for FSD test drives rather than vehicle test drives. On China, Taneja said Chinese OEMs don't currently have the same lead, Tesla is working with regulators toward broader FSD approval, and it has published safety reports.
How is Tesla working with suppliers on semiconductor capacity, and is it taking more aggressive steps for a larger ramp — what is the capex impact?
Musk said the impact is not massive right now and capex is actually ending up a little lower than expected this year; it becomes more meaningful next year and reaches the tens of billions in 2028.
Consensus Estimates
Adjusted EPS of $0.43 (Zacks) to $0.45 (MarketBeat; AlphaStreet, 24 analysts) on revenue of $27.8B (Zacks) to $28.1B (AlphaStreet). EPS estimates are down ~18% over 90 days from $0.55; individual estimates range from $0.30 to $0.70.
Sources
Earnings Archive
Q2 2026 Q1 2026Summaries are prepared by the Tesla Briefing editorial team and may not capture every nuance of the original reporting. You are solely responsible for your own investment decisions.