Tesla Says Demand Now Outruns Production and Puts Optimus Customer Deliveries at Roughly One Year Out, While Terafab Needs "North of $100 Billion"
Q3 2026 earnings call, October 21, 2026: record backlog, a Cybercab ramp in Austin, and a heavier capital commitment
Tesla used its third-quarter call to argue that the company has moved from a long stretch of flat results to a period of constrained supply and rising investment. Management said vehicle demand exceeds production "by a significant margin," that the automotive backlog is the highest in 4 years, and that Megapacks are almost sold out for 2027. The company also gave its first detailed timeline for Optimus and a first cost framework for Terafab. Both point to a capital-intensive next few years. Elon Musk declined to discuss a potential SpaceX merger, saying he had said everything that should be said the day before and that he was "tap-dancing through a minefield."
Optimus: Musk Resets the Timeline and Names the Hand as the Hardest Problem
The most useful new disclosure was a more sober account of Optimus. Early production has started on the high-volume line in Fremont, and most of a 1 million unit per year line is installed. Musk said it will take "at least a year, maybe a little longer" to reach an annualized rate of 20,000 units a week. He framed the ramp as "10,000 things that have to go right," moving "as fast as the slowest, least likely of the 10,000." He said outright that "there are still parts of the design that are changing."
Musk expects the AI, the electromechanical design and robot service life to be resolved around the end of next year, which is when Tesla would deliver Optimus to outside companies. The first customers are likely to be industrial, covering dangerous, repetitive work that runs around the clock, with home sales coming later as production scales. Management said the hand remains the main hardware obstacle. Konstantinos Laskaris called it more complicated than the locomotion actuators and noted that bespoke gears are still in development. Musk said it is hard to build a hand that survives wear, stays soft enough to handle delicate objects, and protects its sensors through repeated use.
The software path differs from the car program. Tesla has no fleet of robots feeding data back, so Musk said the company needs "on the order of 10,000 or 20,000 bots" doing self-play in an "Optimus Academy." Ashok Elluswamy described a reinforcement learning loop that starts from human demonstration, then adds self-play in simulation and in the real world. Musk said the AI5 chip, which offers roughly an order of magnitude more compute than AI4, should be integrated into Optimus around April 2027. He also said Tesla may release an Optimus video in 1 to 2 months showing non-preprogrammed behavior. Vaibhav Taneja said internal Optimus cost tracking is going well, but it is "too early" to claim the target has been reached. Tesla is also building an additional robot factory in Austin.
Robotaxi: Rapid Mile Growth, but Revenue Is Still a 2027 Story
Cumulative unsupervised miles grew 8x in the quarter, with weekly growth in the double digits throughout. Musk said the Austin Cybercab fleet is up 10x since launching about a month ago. Dallas is slated for next month, and Tesla currently operates in 6 cities. Management was clear that the financial contribution is still ahead. Musk said Robotaxi "won't move the needle this quarter," should become noticeable in Q2 2027, and should be "a big deal by Q3 or Q4 of next year."
Unit economics are not yet at target. Asked about the $0.30 per mile cost goal, Taneja said: "Are we at our goal of $0.30 today? No." He said he sees a path there as utilization improves and fleet operations such as cleaning and pickup are refined. Elluswamy said the constraints are mostly technical today but should become regulatory in the Q1 and Q2 2027 time frame. Service hours are being extended step by step, with a 2:00 a.m. close coming soon and 24/7 operation and higher-speed roads targeted by year-end, under what he described as a zero-incident operating principle.
On the customer side, Elluswamy said the first customer-owned vehicles could join the Robotaxi fleet around Q2 2027 through the v15 software release. Lars Moravy said Tesla is trying to get Cybercabs to customers by year-end, though current output goes to Tesla's own fleet, and Musk said the vast majority will stay Tesla-owned. Moravy said the Cybercab has 50% fewer parts than any other Tesla vehicle and that the line's listed capacity of 125,000 units understates what it can do. Tesla is currently building only at the rate the service can absorb. Musk predicted that Tesla will do "a majority of all passenger miles on Earth in any country where it is allowed to operate freely." He added the caveat "unless we are stopped by governments, which is certainly a possibility."
FSD: Attach Rates Are Rising, and Management Confirms the Scale
More than 75% of North American deliveries included a paid FSD subscription sign-up at delivery, about 20% higher than Q2. Management noted that every delivered car includes a month of free FSD. Paid customers exceed 1.7 million globally, and 50% of them are now on subscriptions for the first time. New Street's Pierre Ferragu estimated annual recurring revenue approaching $1 billion. Taneja said his numbers were "correct" and that Tesla will disclose more metrics as the business scales. He declined to give churn data, citing free trials in newly added countries. Tesla is also offering incentives on certain European cars to build a base ahead of broader FSD approval. Musk said he would be surprised if EU-wide approval slips beyond Q1.
On competition, Elluswamy argued that the long tail makes the problem so hard that "anyone with their sane mind should not attempt" it. He said no rival will match unsupervised FSD at scale "in several years." Taneja said Chinese OEMs lack comparable capability but acknowledged that China approval is "hard." Tesla is working with regulators there and has published an unsupervised Robotaxi safety report. Elluswamy also said Tesla is replacing third-party maps with an AI-native navigation system that will begin with the v15 release.
Financials: Margin Gains, Some of Them Low Quality, and a Larger Funding Cushion
Deliveries topped 480,000 for a second straight quarter. Automotive gross margin excluding credits rose from 16.3% to 18.1%, which management attributed to a higher APAC mix. It flagged commodity prices and interest rates as headwinds that it is partly offsetting through financing programs. Energy storage deployments were 13.7 GWh, the second best quarter on record. Energy gross margin rose from 20.4% to 33.1%, but the improvement was helped by tariff refunds and warranty true-downs tied to vendor cell-cost reimbursement, so the level should not be treated as run-rate. Management said ASP and margin should improve further on data center demand. Services and other margin reached a record 15.9%. Tesla also began Megapack deliveries using in-house LFP cells.
Operating expenses continue to climb on AI compute depreciation, chip development, and preproduction costs for Optimus, Megapack 3 and Semi. Taneja said the trend continues until those products are commercialized. Net income included a $142 million mark-to-market decline on the SpaceX holding, offset by a gain of about $290 million on Bitcoin. Free cash flow was over $1.5 billion and cash and investments exceeded $45 billion.
Tesla lowered 2026 capex guidance to about $20 billion but warned that spending will keep rising as the company funds the Robotaxi fleet, Optimus capacity, solar manufacturing, AI compute and a fab. It secured capacity to borrow up to $30 billion and does not plan to draw on it this calendar year. Musk said the 2026 capex outcome is "a little lower than we anticipated."
Terafab and Capacity: Large Capital Commitments, Distant Payoff
Musk said the Austin research fab will begin commissioning equipment in the first half of next year, with a first test chip around May or June, "just to see if we can make any chip at all." The aspiration is useful chips at volume around Q3 2028. Capital needs are "certainly in the tens of billions" through early 2028 and "probably north of $100 billion long term." The split between Tesla and SpaceX has not been settled. Taneja said most orders for the first equipment set are already placed. Musk said capex impact is modest now but becomes meaningful next year and runs to "tens of billions in '28."
Tesla is targeting 5 million vehicles a year by about 2030 and has applied for permits to expand Berlin. It has also started a new Semi factory in Reno. Musk said the Megapack and "Megapod" distributed power and compute opportunity is still early and should become clearer around the end of Q1 or in Q2.