Meta and Tesla must show that $155bn of AI spending can pay its way

Meta and Tesla report third-quarter results after the US market close on Wednesday, 21 October. Between them, the two companies plan more than $155 billion of capital spending this year, mostly on AI. Meta's free cash flow fell 91% to $784 million in the second quarter. Tesla's turned negative, and the carmaker has since arranged debt facilities of up to $30 billion.

“Three months ago, investors were questioning whether Meta's AI spending would ever pay off, and Muse has gone a long way towards changing that conversation,” said Josh Gilbert, Lead Analyst APAC and Middle East at eToro. “Muse has changed the narrative, but there's still a long way to go.”

“The bigger focus is the bill, with capital spending guided at $130 billion to $145 billion this year after free cash flow shrank to just $784 million last quarter, so any early look into what costs may look like in 2027 will be watched closely,” Gilbert said. A year earlier, Meta's free cash flow was $8.55 billion. In the second quarter, revenue rose 28% to $60.8 billion, the fastest growth since late 2021, while free cash flow fell 91%.

“Meta will qualitatively guide to capex and total expense levels that are above the Street,” said Mark Mahaney of Evercore ISI. He estimates Meta's capital spending will rise by about $70 billion in 2027, an increase of about the same size as this year's, and that total expenses will grow about 40%. Mahaney expects analysts to cut their 2027 profit estimates as a result.

“Meta is still spending like a hyperscaler without a hyperscaler's business model, so investors will want to hear how Muse starts paying its way, along with any update on the compute sales Zuckerberg flagged last quarter,” Gilbert added. In July, Mark Zuckerberg said Meta planned to “grow a large business serving large customers” by renting out AI computing capacity. Alexandr Wang, who leads Meta's AI effort, has said spending decisions at this scale are made “long before there's proof points” that the technology works.

Meta's headline numbers will also include a legal cost. The company expects a $10 billion charge in the third quarter, linked to a settlement of up to $18 billion with US state attorneys general over concerns about young people's use of social media. Bernstein's Mark Shmulik forecasts third-quarter operating profit of $11.34 billion, about 40% below the $19.02 billion consensus. Analysts expect revenue of about $63.3 billion and earnings of $6.58 per share, according to TipRanks.

Muse has given investors a consumer story

“Shares have rallied almost 37% since the July result, with the Muse personal AI agent sitting at number one on Apple's App Store since late September,” Gilbert said. Muse launched on 8 September in the US and Canada. It browses websites, handles email and completes multi-step tasks such as bookings. Sensor Tower data puts downloads at about 5 million in the first 22 days. Bank of America estimated about 557,000 daily active US users on 19 September.

“App tracking data suggests downloads have levelled off since the launch surge, while time spent in the app is still climbing, which is the better sign of the two, because retaining users represents potential to truly monetise,” Gilbert noted. Muse has a free tier and paid plans at $20 and $100 a month. Meta also plans to take small fees on transactions the agent completes. Shay Boloor, market strategist at Truist, estimates Muse revenue of about $29 billion by 2030, with $11 billion of that from subscriptions. Hardik Shah of Morgan Stanley raised his price target to $820, citing Muse and business AI subscriptions.

Bank of America's Justin Post said Meta is “back on offense with AI”, but he named rising AI spending as the main risk. Shmulik is wary of how quickly Muse can make money and points to Facebook Shops and the metaverse as earlier bets that fell short. There have also been problems for users. Amazon blocked Muse from its shopping site on 20 September. A journalist reported that the agent read personal messages without being clearly asked to. OpenAI released its rival agent, Dots, on 29 September, available only on paid tiers at launch.

“After a 37% rally, a solid quarter is already priced in. Investors now want to see the money that's going out the door start coming back in,” Gilbert said.

Tesla sold more cars at thinner margins

“Tesla heads into Q3 earnings this week with the car business showing a pulse again. The big question, though, is whether those extra cars actually made any money,” Gilbert said. Tesla delivered 486,532 vehicles in the third quarter, compared with a company-compiled consensus of 461,974 from 24 analysts. That was still 2.1% below the 497,099 record set a year earlier, when buyers rushed to use the $7,500 US federal tax credit before it expired.

“Automotive revenue is still forecast to fall 7% on last year, with auto gross margin excluding credits modelled at roughly 16%, flat on the quarter, so any sign of pricing concessions will show up in margins quickly,” Gilbert continued. The Motley Fool calculates from Tesla's disclosures that automotive gross profit per vehicle fell to $7,213 in the second quarter, as prices weakened and production costs rose. Analysts expect revenue of about $28 billion and adjusted earnings of 44 to 45 cents a share, down from 50 cents a year earlier.

“Energy storage deployments of 13.7 GWh also fell about 14% short of forecasts, removing one of the steadier supports for the earnings story,” Gilbert said. Analysts had expected 15.9 GWh. Outside the US, Electrek reported that exports from Shanghai are making up for weak retail sales in China, and that BYD outsold Tesla by about 276,000 EVs in one quarter.

Tesla is borrowing to fund its next phase

“As previously guided, our free cash flow ended up being negative for the quarter,” Tesla CFO Vaibhav Taneja said in July. Second-quarter free cash flow was negative $1.1 billion, against a positive $146 million a year earlier. Capex rose 142% to about $5.8 billion. Consensus compiled by TradingKey puts third-quarter free cash flow at negative $4.51 billion.

“After a week when the cost of financing AI infrastructure rattled markets, investors will want a much clearer picture of when that spending starts paying its way,” Gilbert said. Elon Musk has said Tesla will spend on capital projects “as fast as we can without it being too wasteful”. Taneja has said capex “will grow for the next two to three years.” Keith Fitz-Gerald, principal at the Fitz-Gerald Group, said that “profitability is being sacrificed for infrastructure”. He expects a payoff in 12 to 36 months.

“Full Self-Driving (FSD) is growing, with 1.48 million active subscriptions at the end of June and more than 55% of new North American deliveries including it, but it remains a limited earnings contributor for now. The robotaxi rollout is moving ahead slowly, while Optimus robots still seem like a far-flung ambition,” Gilbert said. Tesla's robotaxi service operates in seven major metro areas, six of them with no safety driver. It passed 1 million unsupervised miles on 3 September. Cybercab production has begun in Texas, and public rides started in Austin in early September.

“New models are a more tangible promise, but the ‘flying' Roadster unveiled in a piece of pure theatre last week isn't due to enter production until 2028, and a niche sports car was never going to be the answer to a delivery problem,” Gilbert added. Gene Munster of Deepwater Asset Management called the demonstration “a ridiculous display of engineering” with “no commercial value” beyond embarrassing traditional carmakers. Gary Black of the Future Fund assumes 3,000 Roadsters a year at $100,000 profit each. On that basis, the car would add about six cents a share to his 2029 earnings forecast of $5.40.

“The impressive Roadster event puts the Tesla/SpaceX relationship on the table in a tangible way,” Ives wrote on X. A Barclays note suggested a merger could be announced before Tesla's annual shareholder meeting on 3 December, although neither company has confirmed any plans. Tesla is the only Magnificent Seven stock trading lower so far in 2026, and its shares are about 25% below their December 2025 record close of $489.88.

Credit markets have started pricing AI risk

“CoreWeave is absolutely the canary in the coal mine,” said Max Frumes, global head of distressed and leveraged finance at credit data firm 9fin, speaking to Reuters. “This is going to be a turning point in the pricing and availability and structure of credit.” CoreWeave sold bonds in June at a yield of 9.675%. Last week, yields on that debt rose above 15%, and its shares fell 15.1% between Monday's open and Thursday's close. The other neocloud operators, Nebius and Iren, fell by similar amounts. The SOX semiconductor index dropped for six sessions in a row.

“CoreWeave is a bellwether for the riskier cusp of hyperscaler sentiment,” John Atkins, who covers corporate bonds at PitchBook, told Axios. He added that “people are taking a dim view of speculative borrowing that will need to be refinanced at punishingly high rates.” The effects reached well beyond the US. Firmus, a Nvidia-backed data centre operator in Australia, was one of several companies to delay their stock market listings.

“AI financing costs aren't dramatic overall, but there are worrying themes at the edges,” said Jim Reid, global head of macro and thematic research at Deutsche Bank. He linked this to a shift towards a capex cycle “funded out of debt rather than simply free cash flow.” Dan Ives, partner and technology analyst at Yorkville Ives, expects 5% to 10% of planned data centre projects never to be built.

One evening will test the whole AI supply chain

“Meta's capital spending is someone else's revenue,” Gilbert said. Any change in Meta's 2027 budget will go straight through to the chipmakers and data centre builders behind this year's market gains. eMarketer expects Meta's global advertising revenue to reach $243.46 billion this year, ahead of Google's for the first time. That advertising income is what pays for the buildout. About 98% of Meta's revenue last quarter came from ads.

“A few months ago, I would have said people were just too optimistic around AI,” Larry Holzenthaler, senior portfolio manager of fixed income at Catalyst Funds, told Reuters. Meta's results will show how much is left for shareholders once a consumer AI product has to pay for its own infrastructure. Tesla's will show whether higher car sales can help fund borrowing for driverless cars and robots. Both answers matter well beyond the two companies, to the bondholders, chip suppliers and retail investors who have backed the AI trade.