Tesla (TSLA.US) Q3 Financial Report: Saved Money on the Balance Sheet, but the AI Story Is Drifting Further Away

According to Zhitong Finance APP, Tesla (TSLA.US) released its Q3 earnings, causing its stock price to drop more than 3% after hours. Judging by the numbers alone, the report does not look bad: profits, cash flow, and capital expenditures were better than Wall Street expected, with revenue only slightly below forecasts. The real turning point happened during the earnings call—annual capital expenditure guidance was lowered from $25 billion to $20 billion. Management described “spending less” as discipline, but several analysts interpreted it as project delays.

Mixed Q3 Results: Cash Flow Improvement from Spending Pace, Not Profits

The report shows Tesla’s Q3 revenue was $28.05 billion, flat year-on-year and slightly below the market expectation of $28.13 billion. Adjusted earnings per share were $0.48, higher than the expected $0.44; adjusted EBITDA was $3.9 billion, topping the $3.66 billion forecast.

Cash flow was the most closely watched part of this report. Last quarter, free cash flow was negative $1.092 billion, so the market expected a negative $499 million this quarter; the actual result was a positive $1.53 billion, flipping from negative to positive. Breaking it down, operating cash flow was $6.263 billion, just slightly higher than last year’s same period—a modest increase. The real difference came from capital expenditures: $4.732 billion, which, although up YoY, was still $3.2 billion less than the forecasted $7.95 billion. The earning call accordingly reduced full-year capital expenditure guidance from $25 billion to $20 billion. The CFO described this as “efficient capital deployment management.” In other words, the cash flow beat expectations mainly due to spending habits; the outperformance was because of less spending, not better profit generation.

The market isn’t fully convinced. Dave Mazza, CEO of Roundhill Investments, bluntly stated: for most companies, spending less and generating more cash is good, but Tesla is an exception—the market doesn’t price it like a traditional automaker. Capital expenditure is the only hard metric for investors to judge how quickly Optimus and Cybercab are built. They raised guidance by $5 billion six months ago, and now cut it by $5 billion, “This doesn’t look like discipline, it looks more like future ramp-ups are being postponed.”

Analyst opinions diverge here. StoneX’s Mickey Legg is cautious: Q3 underspending “reflects both some sliding projects and a scaled-down overall plan.” Morningstar’s Seth Goldstein is more optimistic: with lower capital intensity, Tesla can “self-finance, rather than needing to issue debt or shares.” But his optimism assumes no project delays—a point not confirmed by the call.

The income statement paints a different picture. Revenue was flat YoY but slightly below expectations. Operating income was $718 million, down 56% YoY; operating margin was 2.6%, narrowed by 322 basis points from last year. Gross margin rose by 160 basis points YoY to 19.6%, adding about $440 million in gross profit. But operating expenses rose $1.35 billion YoY to $4.776 billion, not only offsetting the margin improvement but causing operating profit to fall by $906 million compared to the previous year.

GAAP net profit was $787 million, down 43% YoY; non-GAAP net profit was $1.697 billion, down just 4% YoY. The roughly $900 million gap between them mainly comes from stock-based compensation, digital asset, and SpaceX equity investment unrealized gains/losses. In other words, much of the big GAAP decline comes from volatility unrelated to the automotive core business.

The cash flow statement looks better than the income statement, but the income statement is more honest than the cash flow statement. Cash flow improvement is real, but it comes from the pace of spending, not better core operations. The income statement didn’t catch up—expense growth far outpaced any gross margin recovery.

Automotive and Energy Storage: Shrinking Inventories and Buffers

Data shows Q3 automotive revenue was $20.56 billion, down 3% YoY. Deliveries were 486,532 vehicles, down 2% YoY, but above the market expectation of 463,000. This “beat” should be taken with a grain of salt: production was 464,391 vehicles, so deliveries exceeded production by 22,141 vehicles—meaning some Q3 deliveries came from Q2 unsold inventory.

Regional performances provided mutual offsets, not synchronous recovery. The U.S. weakened due to the expiration of federal EV tax credits—Cox Automotive estimates Tesla sold about 144,000 vehicles in America during Q3, down nearly 20% YoY but still accounting for about 54% share of the US EV market, higher than all other brands combined. In Shanghai, wholesale sales in September reached 95,366 vehicles, the best month of 2026, with exports up 58% and domestic sales down 9%. In Europe, registrations grew 43.3% from January to August.

Product structure is more noteworthy. Non-Model 3/Y vehicles only delivered 8,295 units, nearly halved YoY. Roadster is still in the “design and development” phase, production capacity undetermined, and mass production not expected until 2028. Deepwater’s Gene Munster calls it an “engineering showcase;” Morningstar expects annual sales to be not even 2,000 units. It’s a brand symbol, not a volume product.

Energy storage deployments were 13.7 GWh, up 10% YoY, but energy revenue fell by 16%. Management partially attributed this to tariff-related customer refunds and acknowledged a decline in energy gross profit. This once-important profit buffer is shrinking. Services and other businesses were highlights: revenue rose 33%, gross margin at 15.9%, driven by charging, insurance, and parts, but the scale is too small to offset the declines in automotive and energy storage.

Another overlooked detail: regulatory credit income continued to fall, a drag on both revenue and operating profit as listed by management.

AI: Where Did the $20 Billion Go, and Where’s the Payback Timeline?

The biggest topic of the call was Optimus, but the focus wasn’t on progress—it was on bottlenecks.

When asked about the biggest current design issue, Musk replied bluntly: the hands. He explained that humans have self-healing abilities, whereas robots do not; human fingers are soft and embedded with sensors. This explains why pulling robots off the production line does not equal commercialization—the hardest hardware issue isn’t solved yet.

The timetable was also extended. Musk said it would take another year to deliver robots to customers. This matches previous statements about needing at least another year, maybe longer, to hit “one million unit annualized production or 20,000 per week.” In other words, Optimus is unlikely to make substantive revenue contributions before 2027.

The contradiction is here: if Optimus is a key recipient of the $20 billion in CapEx, why is the ramp-up just beginning, hand problems unresolved, and CapEx already being trimmed? Joe Gilbert of Integrity Asset Management said: lowering capital expenditures “is inconsistent with their ambitions,” and wouldn’t be surprised if the figure is raised again next quarter. Zacks analyst Andrew Rocco tried a neutral interpretation: the cut might reflect deployment timelines—the company is in a major transition, and the key is for management to balance future growth and current cash flow.

Cybercab’s situation is more concrete and more urgent than Optimus’s. By mid-October, over 300 Cybercabs were in service in Austin, an eightfold increase in just 30 days, and Dallas was confirmed as the next stop; but Tesla disclosed neither how many vehicles are operating without safety drivers nor the paid ride volume or unit economics. Regulatory hurdles, unaffected by capital expenditures, are a hard constraint: NHTSA issued 21 special queries about the steering wheel- and pedal-less Cybercab. Tesla requested an extension; NHTSA granted only 30 days, with an October 30 deadline. There’s another timing issue worth noting—Tesla plans to deliver Cybercabs to consumers by the end of 2026, but unsupervised FSD won’t reach consumer vehicles until Q2 2027. This means buyers will receive Cybercabs for up to half a year without either a steering wheel or usable autonomous driving software.

Terafab is a longer-term bill. Musk said equipment costs “are in the tens of billions of dollars,” and total long-term costs “over $100 billion,” shared between Tesla and SpaceX. Most major orders for the first set of equipment have already been placed, aiming for the first test chip output in May or June next year.

This means the long-term capital expenditure pressure won’t disappear just because this year’s spending is cut to $20 billion—this year’s reduction looks more like deferred expenditure than cancellation. Khurshid from Karobaar Capital points to the core issue: If spending is merely postponed, the improved cash flow is temporary; if long-term project timelines remain, future CapEx will return.

Merger: Board Committees Established, Zero Incremental Information

Just a day earlier on the SpaceX call, Musk confirmed both companies have set up special independent director committees to evaluate the merger and said Tesla’s side requires approval from 50% of all outstanding shares—“a fairly high bar.” The market was waiting for more details on Tesla’s earnings call.

The result: nothing. The earnings report made no mention, neither did analyst Q&A; management has now twice refused to discuss it. That in itself is a signal. The merger has retreated from “committees established” to “no new information,” and the 50% shareholding threshold means the deal could drag on for quite some time. The previous day’s excitement cooled quickly the next day.

Analysts also picked up on the call’s atmosphere. SLC Management’s Mullarkey said Musk “seemed less engaged—he seems bored with the automobile industry and has been for quite some time.” This vibe doesn’t change the numbers, but it does impact market confidence in management’s ability to balance future growth and current cash flow.

Summary

To be accurate, this is not an across-the-board earnings beat. Revenue missed. Profits, cash flow, and capital expenditures topped expectations, but breaking it down, the profit outperformance contained accounting items and cost structure effects; the cash flow and CapEx beats came from “spending less.” None of it resulted from improved core business performance.

This shaped the nature of the report: short-term numbers are presentable, but the narrative supporting valuation isn’t strengthened. Free cash flow improvement came from lowering capital expenditure, but the CapEx reduction is seen by many analysts as a sign of project delays or plan shrinkage, not operating leverage. Optimus’s hand issue, a one-year-out delivery timeline, Terafab’s $100-billion long-term bill—all imply the AI business payback schedule is postponed, and spending pressures remain. Merger hype faded overnight, market sentiment cooled.

The 3% post-market drop is less about rejecting the earnings and more a mild vote that “these good numbers aren’t enough to support the current story.” Next quarter’s test will be more severe: when inventory buffers are used up and CapEx can no longer be trimmed, where will profits come from? When Optimus and Robotaxi timelines keep moving back, what will support the AI premium?