Tesla (TSLA) — The Price Moved the Finish Line

Tesla beat the estimate, and this time the profit came mostly from operations rather than from a paper gain on SpaceX. Our fair value moves from $91 to $93.88; the quarter's operating results account for $0.27 of the change. The share price has risen 21% since July, and what it requires of robotaxis and humanoid robots has grown by a fifth.

TON618 Capital Research · As of 21 October 2026 · Download PDF · Download model

In July we asked how far the share price sat from what Tesla's businesses earn. We measured the distance in cars and robots: twenty-one unsupervised Teslas on public roads against about four million robotaxis by 2035, and a few hundred humanoid robots against twenty-six million a year. Both ends of that ruler moved in the third quarter. The near end moved by evidence: the count of cars rose from tens to hundreds, real progress though not yet thousands. The price moved the far end further. This note re-marks the July valuation on the quarter's figures, shows exactly which inputs moved it, and re-solves what $386.47 now requires.

Terms used throughout, defined once. GAAP earnings follow US accounting rules. Non-GAAP earnings are Tesla's own measure, which adds back stock-based compensation, the marks on its bitcoin and SpaceX holdings and certain tax items. Below the operating line: profit that does not come from selling cars, energy products and services, such as interest on cash, investment marks and tax items. Free cash flow: operating cash flow less capital spending. Working capital: money tied up in or released from receivables, inventory, payables and customer deposits. Residual: the part of enterprise value (equity value less net cash and the SpaceX stake) left after the base-case value of the businesses that earn money today. Those businesses are cars and services, energy storage and Full Self-Driving (FSD) subscriptions. Reverse DCF: a discounted-cash-flow model run backwards from the price, to ask what future the price assumes. Unsupervised: a Tesla carrying riders with no employee aboard. Growth stages: periods in the forecast, each as long as the thing driving its growth; every length is argued in the stage table. Fair value: the probability-weighted sum of three scenario values at the July note's segment discount rates (10.1% for cars to 13.7% for Optimus).

Revenue was $28.05B, -0.2% on the year. Automotive revenue fell -3%, with deliveries of 486,532 down -2%. Energy revenue fell -16% to $2.88B, which the deck attributes partly to tariff-related customer refunds, even though deployments rose 10% on the year to 13.7 GWh. Services and other revenue grew +33% to $4.61B. Gross profit rose $440M on the year, and operating expenses rose +39%. Research and development alone rose +69% to $2.76B. Operating income was therefore $718M, against $1.62B a year ago, a margin of 2.6%. The July note's description still holds: the margin did not collapse; Tesla spent it.

The July note's earnings test drew the most attention, so we repeat it on the same basis. Of the second quarter's $0.32 of GAAP earnings per share, $0.22 was the SpaceX mark and $0.08 a California tax release; July measured that as the share of GAAP net income not covered by operating income, 64%. The third quarter is 9% on the same measure. That measure sets after-tax net income against pre-tax operating income, so we add a like-for-like check: non-operating income (interest, investment marks, currency) was 70% of pre-tax income in the second quarter and 25% in the third. Both say the same thing. The earnings moved most of the way back to operations. In detail, GAAP earnings of $0.22 break down as follows. The SpaceX mark took off $0.03 (a $142M pre-tax loss) and a tax accrual took off $0.01; a $288M gain on Tesla's bitcoin added $0.06. What remains, $0.20, is operating income plus net interest, less a $302M currency loss, after tax. Interest income of $445M on the cash pile, about $0.10 a share after tax before interest expense, is still a large part of it.

That settles one question and opens another. The beat was measured on non-GAAP earnings: $0.48 against a $0.45 consensus. Non-GAAP earnings add back stock-based compensation, $0.28 a share after tax this quarter, or 58% of the figure. The add-back is standard and consensus is built on the same basis, so the beat is real on its own terms. But share-based pay is a cost that dilutes owners every quarter, and the house expenses it. On GAAP, the quarter's operating income after a 21% tax was worth about $0.16 a share. In July, the question was whether the earnings were operating. This quarter they mostly were, but there were not many of them.

The cash statement is thinner still. Operating cash flow was $6.26B and capital spending $4.73B, so free cash flow was $1.53B, after −$1.09B in the second quarter. Most of it came from working capital. Payables and accrued liabilities rose $3.99B and deferred revenue $1.25B, against a −$1.82B build in prepaid and other assets. Net of all six working-capital lines, $2.59B of the quarter's operating cash was released rather than earned. Before that release, the business generated $3.67B of operating cash and spent $4.73B: free cash flow of −$1.06B. The second quarter on the same basis was −$2.50B. Payables rising with production is normal; payables rising by $3.99B in a quarter when revenue was flat is a timing benefit that tends to reverse.

The capital plan is the second point. In July, Tesla guided 2026 capital spending to more than $25 billion, growing for two to three years more. Spending through September was $13.0B, so reaching $25 billion would have needed $12.0B in the fourth quarter, more than twice any quarter on record. On the call the chief financial officer cut the guide to "approximately 20 billion for the current year," citing "managing our capital deployment efficiently"; that implies about $7.0B in the fourth quarter. He did not describe a smaller plan. Capital spending "for the next few years would continue to expand" across robotaxi fleet, Optimus capacity, solar manufacturing, AI compute and a semiconductor fab. The fab, which Tesla calls Terafab, is the new line item: "tens of billions" of equipment through early 2028 by the chief executive's estimate, probably more than $100 billion over the long term, with the split between Tesla and SpaceX still to be settled. Management said the related spending becomes more meaningful next year and reaches the tens of billions in 2028. The CFO also said Tesla has now "secured capacity to borrow up to $30 billion," which it does not plan to draw this year; in July that capacity was being arranged, and the 10-Q should show the terms.

The house reads the cut as a deferral rather than a saving. The CFO's own explanation is efficiency, and tripwire 2 is written to decide between the two. The July base case carried $26.9 billion of 2026 capital spending, easing to $24 billion by 2028; a plan that spends $7 billion less this year and tens of billions more on a fab in 2028 has a different shape, and the house does not re-model it on a spoken range. Net cash rose to $35.9B from $34.2B at the end of June. The July position on funding holds: the money is there, but the business no longer funds its own plan, and a quarter of positive free cash flow built on payables does not change that.

Start with the size of the move. The stock closed at $386.47, up +3.4% on the day, and traded at $374.29 at 8:29 p.m. ET, -3.2%. Options had priced a move of about 6.1% either way (the straddle to 23 October, which includes one extra day). The evening price is within a dollar of Tuesday's $373.71 close, so it handed back the day's run-up into the report and little more. We record it as one data point. A beat that trades lower is not unusual for Tesla. It fell in the session after each of its two most recent beats, by 3.5% in January and 3.6% in April.

Against expectations, the beat was narrow. FactSet's consensus, as cited by Electrek, was $0.45 of non-GAAP earnings per share on $28.06B of revenue; Investing.com's was also $0.45. Tesla delivered $0.48 on $28.05B: revenue in line, earnings three cents ahead. The brokerage feed's estimate was lower, at $0.40, which is why the brokerage feed showed an eight-cent beat. On either figure it is a beat.

Five things in the written materials and on the call cut against the headline number, and each can be checked.

1. Profit fell, and the company says the cost line keeps rising. Operating income was $718M, against $1.62B a year earlier, on flat revenue. The CFO said the "trend of growing OPEX would continue till we commercialize these new products," naming AI compute depreciation and pre-production costs for Optimus, Megapack 3 and the Semi.
2. The margin beat leaned on items that will not repeat. Energy gross margin of 33.1%, against 20.4% in the second quarter, came "primarily" from tariff refunds and a warranty true-down agreed with a cell supplier, the CFO said. The second quarter's figure was itself depressed by a $240 million warranty charge. The automotive improvement came from a richer mix of deliveries in Asia-Pacific, against commodity and interest-rate headwinds.
3. The free cash flow was working capital (§2). Before a $2.59B working-capital release, the quarter consumed −$1.06B.
4. The capital plan got a new, unpriced line. The 2026 capex cut came with a fab costing "tens of billions" through early 2028, with Tesla's share of the cost not yet agreed with SpaceX, which the chief executive also runs.
5. The merger went unanswered. Investor relations said the company would "not be addressing any questions today on a potential merger with SpaceX"; the chief executive said he had said everything the day before.

Against those, the call offered dates: Cybercab service in Dallas next month; the first customer-owned cars joining the geofenced robotaxi fleet around the second quarter of 2027; robotaxi "a big deal by Q3 or Q4 of next year"; Optimus delivered to outside companies from "around the end of next year"; an aspiration of five million vehicles a year by about 2030, above the July bull case's 2.96 million. None is a reported figure, and the house does not price a date; each that matters is a tripwire in §8. The coverage that night was consistent with this list. Stocktwits reported a 2% gain on the release that reversed as the capex cut and the slower Optimus timetable came out on the call; Electrek's headline was that margins rebounded but profit still fell. We do not attribute the move to any single item.

The July note summed five businesses, each on its own drivers and discount rate, weighted 30 / 45 / 25 across three scenarios. Its data capture kept the drivers and the business values but not the year-by-year cash flows behind them, as that note disclosed. We rebuilt the cash flows from the stored drivers, setting one cost parameter per business and scenario so the rebuild reproduces every July value at the July date, then tested it against figures it was never fitted to: July's grid of single discount rates applied to all five businesses. At 12.81% the rebuild's base case gives $43 against July's $41; at 8% it gives $88 against $96, and the bull case $520 against $539. That is close enough to re-mark from, and slightly less sensitive to the discount rate than the original. One calibration is worth stating: the July bull-case robotaxi value is larger than its stated drivers produce under the rebuild's timing, by about $4.1B a year after tax, and the rebuild keeps the July bull as published rather than trimming it.

Then we change only what the quarter changed, one input at a time, so every step is a row in the attribution table. Discount rates, scenario weights and every long-run assumption stay as published in July, including the 2030 and 2035 fleet, robot, delivery and margin anchors. The robotaxi and Optimus paths are not moved. The fleet count moved by an order of magnitude, but from a base so small that the July base case's 2030 fleet of about 130,000 cars is still the relevant test. Optimus has no customer.

Read plainly: $1.92 of the change is the passage of a quarter, which brings the future cash flows nearer. $0.42 is the rise in net cash and $0.69 is the SpaceX stake, which Tesla carries at $2.87B and which the July sum left out. The share count took off a few cents. Of the quarter's own figures, lower 2026 deliveries took off $0.04. A 2026 automotive margin of 17.9% instead of the July base case's 16.5% added $0.30. FSD penetration already at 17.1% added $0.01. Fair value is $93.88: bear $28.47, base $56.28, bull $240.03. Even the bull case is 38% below the price, so every scenario we draw still lands below it, as in July. July's headline was a range, $69–91; its lower end came from a 40,000-draw simulation that was not stored and is not re-run, so this note re-marks the probability-weighted figure, the range's top and the figure the house logged.

How long each business grows faster than normal. The lengths are the July anchors, now argued from what drives each stage. The model workbook's Stages sheet carries the year-by-year cash flows.

Why these lengths. Every stage ends where a physical constraint the company has disclosed would bind: installed vehicle capacity, installed Megapack capacity, the Cybercab line, or the fleet itself for software. Where a stage depends on something that has not happened, such as Optimus's first sale, its start is not dated, and the tripwires track it instead. Two sensitivities: if the robotaxi fleet reaches its 2030 size two years later, fair value is $91. Terminal growth of 2.5% or 3.5% gives $91 or $97. Terminal value is 65% of the base case's operating value.

At $386.47 the market values Tesla's equity at $1,528B. Less net cash and the SpaceX stake, that leaves $1,489B of enterprise value. The base-case value of the businesses that earn money today is $177B: cars and services $117.8B, energy $15.2B, FSD $43.9B. The residual is therefore $1,312B, or 88% of enterprise value and $332 of the share price. In July, at that note's $328 modeling anchor, it was $1,088B and $275. The quarter added a few billion dollars to the businesses that earn money; the residual is $224B larger.

We translate the residual into physical output the same way the July note did. Split evenly between the two programs, at economics set in their favor ($0.50 of robotaxi operating profit a mile over 55,000 miles a year; a 20% Optimus operating margin on a $28,000 robot), it requires about 4.8 million robotaxis operating and 31 million Optimus units a year by 2035. In July it was 4.0 million and 26 million. If robotaxi alone must carry it, the fleet is 9.7 million. The tracker's broadest count, about five hundred, is about one ten-thousandth of what the price now asks for.

The whole-company cross-check moved further, because free cash flow fell while the price rose. To justify today's enterprise value at the July company cost of equity (12.81%) and 3% growth thereafter, trailing free cash flow of $3.30B must compound at 67% a year to $326B in 2035. In July the figure was 53% from $5.76B. Had trailing free cash flow stayed at the July level, the rate would be 56%. The price alone moved it, and the weaker trailing cash moved it further.

The discount-rate objection is the one the July note answered with its 4.7% solve. The rebuild solves it from cash flows rather than from a grid: 4.4% at the July $328 anchor, already just under that day's 4.71% Treasury yield, and 4.2% at today's price, against a yield that has risen to 5.13%. Holding our base-case forecast fixed, the price implies an equity risk premium of about −0.9 points: an owner of the equity would be paid less than a Treasury bond. No investor underwrites that on purpose, which is the point. The price does not rest on a friendlier rate applied to our forecast. It rests on a different forecast, and our own bull case is that forecast: it reproduces today's price at about 9.4%, an ordinary rate for an equity. The price is our bull case at full weight and an ordinary discount rate, not a lower rate applied to our base case. At 8% for every business, a rate a reader could fairly argue for given $35.9B of net cash, fair value is $183 (base $90, bull $530), and the price is still 111% above it; at 10% it is $124.

Tesla agreed in January to invest in xAI. After SpaceX acquired xAI in February, Tesla bought about $2.0 billion of SpaceX Class A stock in March in its place. With SpaceX now listed, the mark swings each quarter with a traded price: a $1,005M gain in the second quarter and a $142M loss in the third, leaving a carrying value of about $2.87B. The third-quarter mark does not reconcile to the quoted price on the obvious basis: the stake's value fell about 4.7% while SpaceX's share price fell 11.7% between the quarter-end closes ($170.86 to $150.86). The 10-Q will state the measurement basis, and we read nothing into the gap until it does. SpaceX closed 21 October at $159.31, 5.6% above its September close, so on a quoted-price basis the fourth quarter would start with a gain. The stake is too small to move fair value; what matters is the quarterly GAAP noise it puts into earnings.

The combination matters more than the stake. On 20 October, on SpaceX's call, Elon Musk said both companies have formed special committees of independent directors to consider a merger, that no terms have been agreed, and that a Tesla vote would need a majority of all outstanding shares. Tesla's proxy for its 3 December annual meeting, filed the day before, has three items: three directors, the advisory vote on pay, and the auditor. It contains no shareholder proposals and no combination item, and Tesla says it knows of no other matters. A combination therefore cannot reach this meeting on the current ballot. It would need a supplemental proxy adding an item, or, as with SolarCity in 2016, a separate special meeting with its own proxy. On Tesla's own call the next day, the company declined merger questions. Our SpaceX note of 20 October carries the merger probabilities. This note does not value a combination, and the tripwire below is written on the filings that would show one coming.

The July note armed two triggers that would reverse its call. Both carry a January 2028 outer date and either could fire earlier; neither has. Each is graded against its own words.

Each tripwire states its basis, its tolerance and what the model does with the answer, so the next update applies a rule written now.

1. The annual meeting, and the SpaceX combination. Basis: Tesla's EDGAR filings from now to 3 December. A DEFA14A or revised proxy that adds a combination item to the annual meeting, a PREM14A or DEFM14A for a special meeting, or an 8-K under Item 1.01 announcing a definitive agreement fires it. Committee statements, leaks and prediction-market moves do not. If the meeting is held on the 19 October ballot and the Item 5.07 results 8-K reports only the three proposals, the house records no definitive agreement by early December and the SPCX note's merger probabilities stay where they are. If it fires, a standalone merger note follows, and Tesla's fair value is re-marked on the exchange ratio, not on the committee.
2. Fourth-quarter capital spending and the 2027 guide. Basis: the Q4 release's capital-spending line and the call. The guide implies $7.0B for the fourth quarter (about $20 billion for the year less $13.0B spent); within $1.5 billion of that, the deferral reading holds. The main read is a first 2027 figure, or a Terafab budget with the Tesla–SpaceX split. Any agreement on sharing the fab's cost is a related-party transaction and must appear in an 8-K, the 10-K or the next proxy. If Tesla's share is $20 billion or more through 2028, the base case's free cash flow from 2027 to 2029 is cut by that amount from the cars-and-services leg, which funds the company; the fab's payoff gets no value until it makes useful chips at volume, which the chief executive's stated aspiration puts around the third quarter of 2028. Cost is charged before output because the cost will be contracted and disclosed before any chip exists, while a chip's value depends on yields and a customer the house cannot yet observe. A 2027 guide at or below the July base case's path, with Tesla's Terafab share under $20 billion, records the 2026 cut as a saving.
3. Free cash flow before working capital. Basis: operating cash flow less the six working-capital lines less capital spending, each quarter, from the cash-flow statement. Third quarter −$1.06B, second quarter −$2.50B. If two consecutive quarters come in at zero or better while capital spending is at or above the guide, the base case's cash conversion is confirmed and its 2027 free cash flow is raised to the run-rate. If the payables release reverses by more than half in the fourth quarter, the house records the third quarter's free cash flow as timing.
4. The automotive margin excluding credits. Basis: Tesla's non-GAAP reconciliation. The base case's path is 17.9% in 2026 rising to 18.2% by 2030. Two consecutive quarters at or above 19% move the base 2030 anchor to the bull's path (20.2%). Two at or below 16% move it to the bear's 15.5%. One quarter moves nothing.
5. The independent fleet count. Basis: the community Robotaxi Tracker's "in service, unsupervised" count and its 30-day sighting count, read on each print date. Today: 511 and 299. The armed trigger fires only at thousands, with a third-party safety record. Short of that, a 30-day sighting count above 2,000 by the first-quarter 2027 print, a pace consistent with the base case's 2030 fleet, holds the base case. A count still below 1,000 at the second-quarter 2027 print moves the base case's robotaxi stage from four years to six. A count above 2,000 with a third-party safety record moves the robotaxi stage toward the bull's path. The asymmetry is deliberate: a falling count shows slower deployment on its own, while a rising one needs a safety record to show the deployment is safe.
6. A third-party safety record. Basis: an independent per-mile analysis of NHTSA crash reports for the unsupervised fleet, or a state regulator's published rate. Tesla's own Safety Report page does not count. This is the second half of the fleet trigger, and it is tracked separately so progress on one half is not read as the whole.
7. Optimus: a customer, a price and a date. Basis: any Tesla disclosure or verified report of an external sale. Tesla told JPMorgan sales could start "as early as" the second half of 2027. No disclosure by the second-quarter 2027 print leaves the volume stage undated and the base case's Optimus leg at a cost. A disclosed external sale fires the armed trigger and moves the base case's volume stage to start in the year of the sale.
8. The price. Fair value $93.88. A close at or below it resolves the call, and the house re-underwrites. The residual is re-solved at every update; a residual below 70% of enterprise value, by either route, would be the first sign that what earns money is catching up with what is priced.

Fair value $93.88 a share (bear $28.47 / base $56.28 / bull $240.03 at 30 / 45 / 25), from $91 on 24 July, against the $386.47 close: an expected return to fair value of -76%. The quarter improved the quality of Tesla's reported earnings and lifted the core margin. It did not add enough to the businesses that earn money to change their value by more than $0.27 a share, and its free cash flow rested on working capital. In the same three months the price rose 21%. The distance the July note measured is therefore larger. It is now about 4.8 million robotaxis and 31 million robots a year by 2035, and on our base-case forecast it implies a discount rate below the risk-free rate. The two triggers that would reverse the call both moved, and neither is close. The Fund holds no position in TSLA or SPCX.