Tesla's cash-flow beat came from spending less on AI than planned
SummaryTesla reported $1.53 billion of free cash flow when analysts expected a $4.92 billion burn. The reason: it spent $4.73 billion on capital projects, not the roughly $7.5 billion expected, and cut its 2026 spending plan from more than $25 billion to about $20 billion.
· MODEL Editors · 3 min read
Some links may be affiliate links. Disclosure

A day after Meta showed its AI build-out eating all of its cash, Tesla showed the opposite move: it kept its cash by spending less than it said it would.
The headline numbers. In its Q3 2026 Update on October 21, Tesla reported revenue of $28.05 billion, flat from a year earlier (auto revenue down 3%, energy down 16%, services and other up 33%). GAAP net income was $787 million, down 43%; non-GAAP net income was $1.70 billion, or 48 cents a share, ahead of the 43-cent estimate CNBC cited from LSEG. Operating margin was 2.6%, with operating expenses up 39% on what CFO Vaibhav Taneja called on the call "SI development costs such as depreciation from additional compute, chip development and preproduction ramp costs" for Optimus, Megapack 3 and the Semi.
The surprise was cash. Operations generated $6.26 billion. Capital expenditures were $4.73 billion, more than double a year earlier but well below the roughly $7.5 billion analysts expected, according to LSEG data cited by Reuters. That left free cash flow of positive $1.53 billion, against an expected $4.92 billion cash burn, and after negative $1.09 billion in Q2.
The plan got smaller. On the call, Taneja said Tesla is "revising our CapEx guidance to be approximately $20 billion for the current year," down from the more than $25 billion it had guided. Tesla's 10-Q puts it as "in excess of $20 billion." Tesla has spent $13.01 billion through September, so the new figure implies about $7 billion in Q4. Neither the call nor the filing says which projects are getting less. Musk said capex is "ending up being a little lower than we anticipated" but will become "more meaningful next year," per Yahoo Finance.
Where the AI money is going. The update says Tesla plans to reach 400 MW of AI ("SI") training compute by the end of the year, from more than 225 MW across its Cortex 1 and Cortex 2 clusters in Texas. Optimus production has started on Tesla's lines in Fremont. Active FSD (Supervised) subscriptions rose 67% from a year earlier to 1.74 million, and Robotaxi has logged nearly two million cumulative unsupervised miles. Musk said on the call that open-ended Optimus work still needs "quite a lot of work" on the AI, which he expects to be solved around the end of next year (CNBC).
The catch in the cash. The 10-Q shows Tesla signed $30 billion of new credit facilities in September (a $20 billion delayed-draw term loan maturing in 2029 and $10 billion across two revolving credit facilities). Electrek reports none of it was drawn at quarter-end, and Taneja said Tesla doesn't plan to draw on it this year. Electrek's read of the filing found that $2.59 billion of the quarter's operating cash came from working capital, mostly bills not yet paid and customer deposits, including $588 million billed to SpaceX for Megapacks that have not been delivered. Electrek expects that timing to reverse when the invoices come due and the Megapacks ship.
How the stock took it. Reuters said shares rose about 2% after hours, then reversed to trade down about 2%; Yahoo Finance said they were down almost 5% before Thursday's open. They are down about 14% this year. Musk said again on the call that he had nothing to add on a possible Tesla-SpaceX merger.
Our take
Wall Street got the cash it wanted, but not the AI build-out Tesla has been pitching. A company that says its future is robotaxis, robots and its own AI just spent about $3 billion less than analysts expected in a single quarter, lowered its full-year plan by about a fifth, and lined up $30 billion of credit. Either projects are slipping into 2027 or Tesla is choosing to protect its balance sheet. Tesla hasn't said which.
What we did and didn't do: figures come from Tesla's Q3 2026 Update and its 10-Q, read October 22, 2026, plus the call transcript from The Motley Fool and reporting by Reuters, CNBC, Electrek and Yahoo Finance. Analyst estimates are as those outlets reported them. We didn't listen to the full call. Free cash flow is a non-GAAP measure, calculated by Tesla. This is not a recommendation to buy or sell anything. Not sponsored; no affiliate links. Not investment advice.
MODEL Editors
Sources
- Tesla Q3 2026 Update (Oct 21, 2026)
Tesla Investor Relations
- Tesla Form 8-K, Exhibit 99.1 (Q3 2026 Update)
U.S. SEC
- Tesla Form 10-Q for the quarter ended Sept 30, 2026
U.S. SEC
- Tesla cuts spending before potential SpaceX deal even as Musk bets future on AI, robots (Akash Sriram, Abhirup Roy)
Reuters (via Investing.com)
- Tesla beats on earnings and sales even as auto revenue declines (Lora Kolodny)
CNBC
- Tesla (TSLA) Q3 2026 Earnings Call Transcript
The Motley Fool
- Tesla (TSLA) quietly cuts 2026 capex guidance to $20B (Fred Lambert)
Electrek
- Tesla stock slips despite Q3 earnings beat, better-than-expected capex spend (Pras Subramanian)
Yahoo Finance
Not investment advice.
FREE · EVERY WEEK
The top AI tools this week, tested.
Every week MODEL's editors pick the AI tools worth your time, try them hands-on, and tell you what's good, what's not, and who it's for. Free, in your inbox.
Free. Unsubscribe anytime. Our picks are editorial and can't be bought.
The weekly three is editorial. A subscription does not buy a place in it.
Continue

Markets
Meta's AI build-out cost more than its business brought in this quarter
Meta beat revenue estimates with $64.6 billion, but capital spending of $34.91 billion outran the $32.79 billion its operations generated, leaving free cash flow at negative $2.12 billion. It is also creating a new class of non-voting shares to help pay for the AI race.
· MODEL Editors

Markets
Amazon can deliver almost anything in a day. Its new AI racks are waiting on electricity
Business Insider reports that AWS is taking delivery of billions of dollars of Nvidia GB300 server racks earlier than before, and that some sit in warehouses because the data centers don't have enough power to turn them on yet.
· MODEL Editors

Markets
Microsoft walks away from Britain’s biggest planned AI data centre
Nscale’s Loughton site in Essex was meant to hold 23,040 Nvidia GPUs for Azure. UK reports say its grid connection may not arrive until the 2030s. Microsoft says the decision covers a single site.
· MODEL Editors