Aya Research
DEEP RESEARCH
中EN

Tesla: The Market Values the Car Business at 12 Cents on the Dollar — Everything Else Is a Promise with a Deadline

Panoramic research report · longitudinal history + cross-sectional rivalry + synthesis Subject: Tesla, Inc. (NASDAQ: TSLA) Report date: 2026-07-05 · Data cutoff: Q2 2026 deliveries (reported 2026-07-02) + Q1 2026 financials + news flow through 2026-07-05 (Q2 earnings due 2026-07-22) Sources: company IR and SEC filings, Tesla quarterly updates, CNBC, Electrek, CnEVPost, Cox Automotive, NHTSA/California DMV records, Waymo/Baidu disclosures, ARK/Morgan Stanley/BofA/Wells Fargo research coverage, Yahoo Finance For information and research purposes only. Not investment advice.

Before we start

Every report in this series has audited the gap between a company's story and its receipts. Tesla is the report where the gap is the company.

Here is the arithmetic the market itself publishes. Bank of America's sum-of-the-parts model — a bullish one, with a price target above the current price — attributes 45% of Tesla's value to robotaxis, 19% to Optimus humanoid robots, 17% to Full Self-Driving software, 6% to energy storage, and 12% to the business that generates substantially all of the revenue: selling cars. Morgan Stanley's decomposition is structurally identical. Cathie Wood's ARK, the loudest bull, prices the stock at $2,600 with 90% of that value from robotaxis — and concedes that without them, its own model produces roughly $350. The most institutionally endorsed valuation view of Tesla, across bulls and bears alike, is that 85–90% of a $1.48 trillion market capitalization rests on businesses that produced approximately zero revenue last quarter.

Now the receipts, as of this week. The robotaxi business: roughly 20 unsupervised vehicles active in Austin (peak 25, per city filings), single-digit fleets in Dallas and Houston, and a Miami launch two days ago — against Waymo's 500,000 paid rides per week across ten metros, and Baidu's 350,000 in China. The robot business: zero Optimus units doing what Elon Musk himself calls "useful work"; the production-intent prototype has slipped to a promised summer reveal. The software business: real but disputed subscriber counts and roughly a billion dollars of deferred revenue. Meanwhile the 12%-weighted car business just printed its best second quarter ever — 480,126 deliveries, up 25% — and the stock fell 7% on the news, the market's bluntest possible statement that it no longer prices this company on cars at all.

Tesla is therefore the precise inversion of everything this series has audited. Meta was fined for spending without receipts while its core business accelerated; Tesla is rewarded — at 380× trailing earnings, roughly 196× forward — for promises, while its core business fights a two-front war against BYD's cost curve and its own CEO's political shadow. SK hynix taught us to read a record ADR as a referendum; Microsoft taught us to read a contract amendment as confession; Amazon taught us the permission slip. Tesla's confessional document is the $1 trillion compensation package its shareholders approved in November — twelve tranches whose milestones (20 million cumulative vehicle deliveries, 10 million FSD subscriptions, 1 million robots, 1 million robotaxis, an $8.5 trillion market cap) constitute management's own official probability map of the story. The market approved the map by 75%. This report audits the territory.

Part One is the longitudinal history — three genuine near-deaths, the greatest short squeeze ever recorded, and the political self-wound that broke the brand's largest market. Part Two is the board in July 2026: the car war, the robotaxi ledger, the FSD annuity, the robot promise, the energy business nobody prices, and the Musk complex. Part Three synthesizes: the snapshot with its quality asterisks, the compensation package as analytic object, scenarios, and the falsifiable claims. Series connections: Tesla's AI5/AI6 silicon runs through the Samsung Taylor fab our SK hynix report covered; its Cortex training cluster queues for the NVIDIA and memory supply our first reports priced; and its robotaxi contest with Waymo is the mirror of our Alphabet report's Section 11, argued from the other side.

Part One · Longitudinal: three near-deaths and a self-inflicted wound (2003–2026)

1. Two engineers, one check, and Christmas Eve (2003–2008)

Tesla was founded in July 2003 by Martin Eberhard and Marc Tarpenning — a fact worth stating plainly because the company's mythology has largely absorbed it. Musk arrived in February 2004, leading the $7.5 million Series A with $6.5 million of his own money and taking the chairmanship; a 2009 legal settlement licensed five men, Musk included, to call themselves co-founders. The distinction matters analytically, not just historically: Tesla is a company where narrative control has always been part of the capital structure. The founding wager itself deserves credit the skeptic's frame can obscure: in 2003, the consensus of every major automaker, every energy analyst and most of Silicon Valley was that lithium-ion vehicles were golf carts with delusions. Tesla's founders — and Musk's capital — were right about the single largest contrarian call in modern industry, seventeen years before the incumbents capitulated. Whatever this report concludes about today's promises, the company earned its hearing honestly.

The first near-death was the real thing. The 2008 crisis caught Tesla mid-Roadster with a collapsed funding round; the company's cash would have run out days after Christmas. The $40 million bridge financing closed at 6 p.m. on Christmas Eve 2008 — three days from insolvency — with Musk contributing essentially his last liquid dollar and borrowing from friends for rent. Every element of the Tesla equity cult — the founder who bet everything, the company that survives what kills others, the shorts who underestimate both — was forged in that week, and it has been the stock's operating system ever since. It is also, this series would note, the origin of a specific investor reflex that Part Three must price: the burden of proof at Tesla has historically fallen on the skeptic, because the skeptics were so spectacularly wrong at the two moments it mattered most.

Compare the founding near-deaths across this series, because the differences are diagnostic. Hynix in 2001 and Amazon in 2000 were saved by creditors and markets — external parties weighing collateral. Tesla in 2008 was saved by one person's refusal, weighing nothing but conviction. Companies carry their rescue mechanisms forward as culture: hynix builds contract cover, Amazon manages liquidity timing obsessively — and Tesla, rescued by belief, has institutionalized the production of belief. It is the only company in this series whose survival mechanism and whose product narrative are the same thing, which is the deepest reason its equity resists the receipt-auditing every other report in this series applies.

2. The IPO, the Google escape, and the first profit (2010–2015)

The June 2010 IPO — $17 a share, $226 million raised, the first American carmaker to go public in 54 years, up 41% on day one — priced a curiosity, not a company; the entire raise would fund roughly thirty hours of the company's 2026 capex program. The second near-death followed within three years: Model S demand wobbled in early 2013, and by the account in Ashlee Vance's biography, Musk had agreed terms to sell Tesla to Google for roughly $6 billion (plus $5 billion in committed capacity investment) with Larry Page — the deal dying only because Tesla scraped together its first-ever quarterly profit in May 2013 and the stock quadrupled. The company that today carries a $1.48 trillion valuation was, thirteen years ago, a signature away from being a Google subsidiary priced at less than one-half of one percent of that.

The counterfactual is not idle: it is this report's first calibration of how violently Tesla's terminal value has always swung on quarters, not decades. A single delivery beat in 2013 was worth, in hindsight, more than a trillion dollars of path. That is the same convexity the market is pricing today when it moves the stock 7% on a delivery print — the instrument has always behaved this way; only the notional has changed.

3. Production hell, "funding secured," and the greatest squeeze ever recorded (2016–2021)

The Model 3 ramp of 2017–18 was the third near-death — "about a month from bankruptcy," by Musk's own later account — and the crucible of the short war. It was also, seen from 2026, the founding demonstration of the company's genuine superpower: not invention (the Model 3 was a straightforward product) but manufacturing ramp under existential pressure — tents in the parking lot, lines redesigned weekly, an automation strategy publicly abandoned and rebuilt mid-crisis. Every subsequent Tesla operational miracle (Shanghai's twelve months, the 2020 volume surge, Megafactory Shanghai's seven months) is this capability re-executed. The reason the observation matters for Part Two: the robotaxi and Optimus promises are usually defended by analogy to these ramps, but the ramps were all manufacturing problems — domains where iteration speed and willpower compound. The current promises hinge on statistical software safety and regulatory admission, domains where iteration is rate-limited by miles, incidents and agencies, and where willpower has no purchase. The superpower is real; the analogy is category error. Tesla became the most-shorted stock in America; Musk's August 2018 "funding secured" tweet about taking the company private at $420 bought him a $40 million SEC settlement and the loss of his chairmanship; and then the survivors were paid: production stabilized, China arrived (the Shanghai Gigafactory went from groundbreaking to first deliveries in under twelve months in 2019 — still the operational benchmark nobody, including Tesla, has repeated), and the stock rose 743% in 2020 alone. Short sellers lost $38 billion in that single year. Tesla entered the S&P 500 as its largest-ever addition, and in October 2021 a Hertz order flipped the market cap above $1 trillion.

The squeeze's legacy is under-appreciated as a structural feature of the stock: it liquidated an entire generation of fundamental skeptics, professionally and financially, and taught the marginal investor that Tesla's valuation gaps close upward. The reflex survives among retail and momentum capital to this day. It has been rewarded, cumulatively, by a 300-fold return since IPO — and it is precisely the reflex that a promise-priced $1.48 trillion valuation now leans on, at a scale where the 2013 and 2020 rescues (a demand quarter; a production ramp) have no obvious analogue. Squeezes rescue stocks; only revenue rescues valuations.

The "funding secured" episode deserves one more analytical beat, because it established the disclosure regime under which everything in Part Two must be read. The SEC settlement was, by any Fortune-500 standard, extraordinary: a sitting CEO sanctioned for a materially false statement about the company's own capital structure, penalized, stripped of the chairmanship — and retained, with the shareholder base's enthusiastic blessing, precisely because the market had learned to price his statements as aspirations with option value rather than representations. That is the regime still in force. When Musk says a million robotaxis, the sophisticated holder hears a direction, not a date; when he says Optimus will be 80% of Tesla's value, the market neither believes it nor punishes it. This asymmetric-credibility equilibrium — statements that can inflate the multiple but never trigger accountability — is unique in large-cap markets, is the reason the promise inventory of Part Two could accumulate to 85% of a trillion-and-a-half-dollar valuation, and is, in the most literal sense, the product the shareholders voted to retain in November 2025.

4. The Twitter tax and the first real down-cycle (2021–2024)

The descent had two engines. First, distraction made liquid: Musk sold roughly $22 billion of Tesla stock in 2021 and another $15 billion-plus in 2022 to fund the $44 billion Twitter purchase — the largest founder monetization in market history, executed into a falling tape. Apply this series' capital-behavior lens without flinching: the most informed holder of Tesla equity sold ~$37 billion of it within eighteen months of the all-time high, to buy a social network. Every subsequent insider-conviction argument for the stock must be net of that datapoint — the confessional record shows the controlling shareholder monetizing the 2021 multiple, not underwriting it. Second, the business cycle arrived for a company priced as if exempt: the 2023 price war (six US cuts in a year) took automotive gross margin from a 29% peak toward the high teens; 2022 closed at -65%, the worst year in Tesla's public life, erasing more than $700 billion. Cybertruck — announced 2019, delivered late 2023 — became the era's emblem: 39,000 units in 2024, then -48% to about 20,000 in 2025, the steepest sales decline of any EV in America, its cheap variant cancelled. A 10% global layoff followed in April 2024. The "We, Robot" event that October — Cybercab, no steering wheel, sub-$30,000, 2026 production — was received as the pivot it was: the stock fell 8.8% the next day on detail famine, and the company's center of narrative gravity moved permanently from cars to autonomy.

5. The political self-wound (2024–2026)

Before the politics, mark the down-cycle's structural lesson, because the bulls' rebuttal to it powers the current multiple. The 2022–24 period proved that Tesla's automotive profit pool was cyclical after all — that the 29% gross margins of 2022 were a supply-demand artifact, not a software-like structural feature, and that when Chinese competition arrived the company would cut price like any automaker, because it is one. The bulls absorbed this not by disputing it but by reframing it: if the car business is cyclical and commoditizing, the investment case must live elsewhere — which is precisely when the robotaxi/Optimus narrative was promoted from optionality to thesis. The October 2024 "We, Robot" event and the 2025 comp package formalized the migration. In this series' vocabulary: when the collateral was marked down, the position was defended by enlarging the promise. That sequencing — narrative expansion as a response to fundamental compression, not a product of fundamental strength — is the single most important pattern for calibrating everything in Part Two.

What happened next has no precedent in large-cap history and deserves unsentimental measurement: a chief executive converted his company's largest growth market into a protest movement against himself. Musk's 2024 election alliance with Donald Trump, the DOGE government-efficiency tour of early 2025, the June 2025 public rupture (Trump threatening contract cancellations; Musk claiming electoral credit; an America Party flirtation since parked in favor of quietly funding midterm Republicans), and the subsequent reconciliation cycle produced measurable commercial damage concentrated exactly where Tesla's brand had been strongest. Europe: 2025 registrations -27.8% (to ~235,000) while the European BEV market grew 27%; Germany -48%, France -42%, the Netherlands -67%; thirteen consecutive months of decline through January 2026; BEV share from 18.2% (2023) to 8.9% (2025). US survey data: 67% of American adults would not consider a Tesla — 82% of Democrats, and, tellingly, 53% of Republicans, whose goodwill did not convert to purchases and dropped a further eleven points when Musk and Trump feuded. China, on a different mechanism (competition, not politics): retail share 7.8% → 6.0% → 4.9% across 2023–25, touching 3.06% in April 2026.

The analytical point is not moral; it is that brand impairment is the one balance-sheet writedown Tesla has never had to model, and it landed simultaneously with the product line aging (the Model Y platform is four-plus years old and, in China, carries nearly 90% of retail volume) and the US federal purchase credit expiring (September 30, 2025 — after which monthly US sales briefly fell below 40,000, the lowest in years). Three secular headwinds, one of them self-inflicted, arriving together: that is the demand backdrop against which Part Two's story businesses must be priced.

The counter-reading, which the Q2 print gives real support: brand damage of this kind may prove cyclical rather than structural — attached to a news cycle rather than to the product, decaying as the political temperature does. Europe's Q2 recovery (oil prices helping, the Musk-Trump feud cooling) is the first evidence; the deeper evidence is that Tesla's brand crisis never touched the metric that historically kills car brands, which is product quality perception. A brand wounded by its CEO's politics can heal on his silence; a brand wounded by its products cannot heal at all. The honest uncertainty is whether silence is an available input. Nothing in twenty years of observable behavior suggests it is — and that, rather than any poll, is why the impairment deserves a permanence discount somewhere between the bulls' zero and the bears' total.

6. The pivot to promises — and the quarter that complicated everything (2025–2026)

The 2025–26 period reorganized Tesla, formally, around the story — and the reorganization was thorough enough to read as a corporate restatement of purpose executed without a single SEC filing. The $1 trillion compensation package passed in November 2025 with over 75% support — twelve tranches to an $8.5 trillion market cap, with operational gates including 20 million cumulative deliveries, 10 million FSD subscriptions, 1 million Optimus robots and 1 million robotaxis. The Delaware-voided 2018 package ($56 billion) was restored by that state's Supreme Court in December and delivered in April. Tesla invested $2 billion in Musk's xAI (after a shareholder advisory vote too muddied by abstentions to bind), which subsequently folded into SpaceX ahead of a mooted IPO of the combined entity. Dojo — the in-house training-supercomputer bet — was shut down in August 2025 as an "evolutionary dead end," with compute strategy pivoting to NVIDIA (the Cortex cluster at ~81,000 H100-equivalents and growing) plus the $16.5 billion Samsung Taylor foundry deal for the AI5/AI6 inference chips. Deliveries fell 9% in 2025 — the second consecutive annual decline, surrendering the global EV crown to BYD — and revenue posted its first annual drop in company history; the stock, extraordinarily, finished the year up 11%, the purest single-year demonstration on record that this equity and its income statement have completed their divorce.

Then, two days ago, the complication: Q2 2026 deliveries of 480,126, up 25%, the best second quarter ever, powered by the $39,990 Model Y Standard, a genuine European recovery (oil prices helping), and a +32.8% China wholesale quarter. And the stock fell 7% on the print — its worst day in a year. Hold that pair of facts together, because it is the entire Tesla debate in one session: the car business can still surprise, and the market has stopped paying for it. What the market now pays for reports on July 22, and lives in Part Two.

(Why would a record quarter sell off 7%? The plausible mechanics, in declining order: the mix behind the beat — a $39,990 de-contented variant driving volume implies margin dilution the July 22 income statement will quantify, and the market front-ran the quantification; the "good news pulls forward nothing" problem — delivery beats no longer change the SOTP's 88%, so prints are sell-the-news events by construction; and positioning — the stock had rallied into the print. Whatever the weighting, the session belongs in the same exhibit case as Meta's April capex selloff and Amazon's February one: 2026's market grades every Mag 7 disclosure against the story's requirements, not the quarter's. Tesla's story requires autonomy numbers, and none were in the deliveries PDF.)

Part Two · Cross-sectional: the board in July 2026

7. The car business: appraising the 12%

Since the sell side prices the automotive business at 12–15% of the company, appraise it on its own terms — roughly $180–220 billion of implied value, about one BMW-plus-Mercedes, for the world's second-largest EV maker. (For calibration: Toyota, selling six times Tesla's volume at comparable group margins, carries about $290 billion. The 12% is not obviously wrong. That is precisely the problem for the other 88%.)

The bear facts are structural. BYD reclaimed the global BEV crown in Q2 (557,090 vs 480,126) and out-delivered Tesla across the first half; its vertical integration (cells to ships) supports a cost position Tesla matches only in California-and-Texas manufacturing efficiency, and its price war — average discounts at record 10%, industry capacity of 55 million units against 23 million of demand, the chairman himself calling it a "brutal elimination round," and BYD's own profits falling for the first time since the pandemic — sets the global clearing price for EVs. (An elimination round brutal enough to wound BYD is, for series readers, the automotive edition of the memory industry's pre-consolidation dynamics: overcapacity grinding all players until exits create pricing power. The difference: memory consolidated to three players over thirty years; China's EV sector still counts dozens, and its government subsidizes persistence. Tesla is fighting the 1990s DRAM industry, before the mercy of consolidation.) China share has roughly quartered in three years. Europe is a brand-repair project measured in years. The US credit is gone, the domestic EV market shrank 28% in Q1, and the regulatory-credit annuity — $11.8 billion of nearly pure profit over a decade — is legislated toward zero by 2027 after Washington's CAFE changes. The product line is old, and its one new high-volume entry (Model Y Standard) is a de-contenting exercise that cannibalizes mix.

The bull facts are real but narrower. Q2 proved elastic demand exists at the right price (+25%). Post-credit America consolidated toward the strongest player: Tesla's US BEV share rose to 57.5% as the market shrank — the "last man standing" dynamic in a subsidy washout, a pattern our memory-cycle reports would recognize as consolidation-by-downturn; while the US EV market fell 28% in Q1, Tesla's own US deliveries fell only 4.6%, meaning the credit cliff functioned as a stress test its weaker rivals failed harder. The energy business (Section 11) increasingly deserves separation from "cars" entirely. And manufacturing remains excellent: gross margin ex-credits recovered to 19.2% in Q1 despite everything above — though the quarter's optics flatter, containing one-time warranty and tariff-refund benefits that Electrek and others correctly flagged.

The China file deserves its own paragraph, because it is where the two wars meet. Xiaomi's YU7 — launched with 240,000 locked orders in eighteen hours, briefly China's best-selling vehicle outright, now undercutting the Model Y by ~$4,300 with 50km more range in its standard trim — is the sharpest expression of a general condition: Tesla now competes in the world's largest EV market against consumer-electronics companies with better software reputations locally, faster refresh cycles, and cost structures built on the same supply chain. The Q2 wholesale recovery (+32.8%) is real but flatters — it counts exports from Shanghai, while retail share in China itself touched 3.06% in April, a fifth of its 2023 level, with the aging Model Y carrying ~90% of what remains. No brand-repair program fixes this one; it is product cadence and price, against competitors who iterate annually. The bull answer — that China was always going to normalize and Tesla's margin discipline beats chasing share into a 10%-discount price war — is defensible, but it concedes the growth market that once anchored the growth multiple.

Net appraisal: as a standalone automaker, Tesla is a strong-margin, aging-portfolio, two-front-war company plausibly worth its sell-side weighting — which is to say, the 12% is fairly priced, and therefore does no work in the investment case. The equity is decided entirely by what follows.

8. The robotaxi ledger: 42 cars versus half a million rides

State the competitive scoreboard without adjectives. Waymo: ~500,000 paid rides per week (March 2026), ten-plus metros, roughly 3,000–3,600 vehicles, 577 registered in Texas alone, $126 billion externally-priced valuation, targeting a million weekly rides by year-end. Baidu Apollo Go: 350,000+ weekly rides, 27 cities, expanding to Europe and the Gulf. Zoox: operating Las Vegas and San Francisco, Uber-app integration coming. Tesla: 42 fully driverless vehicles registered in Texas per city filings (~20 active in Austin after a peak of 25), single-digit fleets in Dallas and Houston, Miami launched July 3 in a 10–14-square-mile geofence, no ride counts disclosed, and the CEO's own guidance that scale awaits the FSD v15 rewrite in late 2026 or 2027.

The valuation cross-check, which our Alphabet report set up and this one must complete: the market prices Waymo — operating at three-orders-of-magnitude greater scale — at $126 billion, while pricing Tesla's robotaxi aspiration, per the sell side's own 45% attribution, at roughly $650 billion. The implied claim is that Tesla's future robotaxi business is worth five Waymos today, discounted from a starting position of 42 cars. There are internally consistent ways to hold that view (fleet-cost asymptotics, manufacturing scale, no-depot economics), but they all require believing Waymo's operational lead is worthless — that permits, safety records, consumer habit and city-by-city regulatory capital do not compound. Nothing in the history of regulated transport supports that; everything in the history of regulated transport suggests the incumbent operator's moat is the regulatory record. One of these two prices is wrong by a multiple, and unlike most such disputes, this one publishes weekly ride counts.

The bull case is not stupid, and this series' Alphabet report stated it fairly from the other side: Tesla's approach — cameras only, no lidar, no HD maps, a fleet manufactured at automotive scale on its own lines — carries a cost structure (Morgan Stanley: ~$0.81/mile against Waymo's ~$1.40) and a scaling model (every Model Y is a potential robotaxi) that, if the software converges, makes Waymo's depot-and-sensor model look artisanal. The Cybercab, purpose-built with no controls at a claimed sub-$30,000 build cost, would compound the advantage. That is the "if" on which 45% of the company's value rests. Baidu's Apollo Go deserves more weight in this comparison than Western coverage gives it: 350,000 weekly rides on a hybrid sensor stack, expanding internationally (Switzerland, the Gulf, a London pilot with Uber), it is the second existence proof that lidar-inclusive architectures scale commercially — and a reminder that if Tesla's robotaxi ever reaches China, the market-share fight will be against an incumbent with a four-year operating head start and state-adjacent regulatory standing, in a country where Tesla's FSD approval remains partial.

The evidence file on the "if," honestly summarized: NHTSA's engineering analysis EA26002 (upgraded March 2026, the last step before a recall demand, covering ~3 million vehicles) documents FSD failures in low-visibility conditions — glare, fog, dust — that are precisely the scenarios lidar exists to solve; the robotaxi fleet has reported 14 collisions in ~800,000 miles (roughly one per 57,000 miles, an order of magnitude worse than Waymo's peer-reviewed injury-rate data); a Katy, Texas pedestrian fatality opened a new probe in June; and Musk's own 100-billion-supervised-miles framing keeps moving the convergence date. None of this proves vision-only fails — the honest statement is that Waymo has demonstrated safe scale on expensive hardware while Tesla has demonstrated cheap hardware without safe scale, and the $1.48 trillion question is which constraint relaxes first. Two years of runway remain on the promise before the comp package's own milestones (1 million robotaxis) begin to date-stamp it.

One more asymmetry the debate rarely prices: the cost side of the bull case decays with time even if the technology converges. Waymo's next-generation platform cuts its vehicle cost from ~$125,000 to ~$32,000 — the sensor-stack premium that anchors Tesla's $0.81-versus-$1.40 per-mile advantage is a depreciating moat, shrinking with every lidar cost curve and every Waymo hardware generation, while Waymo's operational lead (permits, safety data, city relationships, consumer habit) compounds. Tesla's economic edge is largest now, when it cannot yet deploy it, and will be smallest when it finally can. The same clock runs on the fleet-conversion argument: the millions of customer Model Ys that could "become robotaxis overnight" carry HW4 compute that Musk has already conceded may not suffice for unsupervised operation — the installed-base option, 45% of the SOTP's collateral, quietly requires a hardware retrofit program nobody has costed publicly. These are not fatal objections. They are the reason "if the software converges" understates the conjunction actually required: the software must converge while the cost edge still exists and on the hardware already sold. Three conditions, one deadline, priced as one.

9. FSD: the annuity that keeps being pre-sold

The one story business with actual revenue deserves its own honest ledger. FSD generates real money — roughly $1 billion of deferred revenue scheduled to recognize over the next year — and real strategic value: Europe approvals began (Netherlands, April 2026, spreading under UN R171), China granted supervised approval in May, and the subscription conversion (Tesla killed the $15,000 outright purchase in February, going subscription-only) converts a lumpy option into an annuity. Subscriber counts are contested across sources — one credible reading puts ~476,000 subscribers and ~$546 million ARR in Q1; another claims 1.28 million; the divergence itself tells you disclosure is inadequate for a business carrying 17% of a $1.48 trillion valuation in the sell side's own models.

Two structural observations. First, the subscription-only pivot is double-edged as a confession: it signals confidence (recurring revenue beats one-time at scale) and it abandons the upfront cash that once financed development — a company guiding free cash flow negative does not lightly give up $15,000 checks unless the take-rate at that price had stalled. Second, FSD's terminal value is not independent: it is the same neural network as the robotaxi bet, priced twice in the SOTP — once as consumer software (17%) and once as the robotaxi platform (45%). If v15 converges, both legs pay; if it doesn't, both were the same leg all along. Sell-side decompositions that treat them as diversified line items are double-counting a single technological event, and investors should read 62% where the models say 17-plus-45.

The regulatory geography, though, is FSD's genuine 2026 bright spot and deserves fair weight: the Netherlands' RDW approval under UN R171 (April) opened a European pathway four countries have already followed, with all-Europe targeted by summer; China granted supervised approval in May with full rollout targeted for Q3. Each new jurisdiction expands the subscription TAM at near-zero marginal cost and — more subtly — builds the international regulatory record that a future unsupervised case will cite. If the v15 promise has a hedged version, this is it: even a supervised-forever FSD, sold monthly across three continents into a 7-million-vehicle installed base, is a real software business worth a meaningful fraction of the 17% — the one fraction of the story stack that pays out partially on failure.

10. Optimus: the 19% shipped at zero

The robot receives four sentences of facts, because that is what the facts support. Optimus has missed every publicly stated timeline since 2021; as of January 2026 Musk acknowledged no unit performs "useful work," with a few hundred deployed in learning roles; the production-intent V3 prototype, promised for Q1, has slipped to a summer reveal alongside a Fremont line (converted from Model S/X) targeting 50,000–100,000 units in 2026 and a claimed million-unit-per-year line by year-end — claims Musk himself hedges as "literally impossible to predict" given ~10,000 unique parts. The competitive field is not waiting: Figure's units already work commercially in BMW's largest plant at ~$25/robot-hour; Unitree shipped ~5,500 units in 2025 at consumer prices. BofA weights this business at 19% of Tesla — roughly $280 billion, or about one entire Toyota — for a product line with zero commercial deployments. It is the purest promise in the Mag 7, its first falsifiable test (the Fremont reveal) is weeks away, and this series will simply note the base rate its own timeline history establishes.

To be fair to the 19%, state the bull logic in its strongest form: humanoid robotics is plausibly the largest product market ever conceived (labor itself), Tesla is one of perhaps three entities on earth with the manufacturing scale, actuator supply chain, real-world training data and capital to attack it, and the same skeptics' arithmetic — "zero units, missed timelines, competitors ahead" — described Tesla's automotive position in 2015 against an industry it subsequently humbled. The rebuttal to the rebuttal: cars were a known product with known demand awaiting a better version; general-purpose humanoid labor is an unproven product category whose unit economics, safety regime and demand curve are all conjecture — and the competitor now playing Tesla's 2015 role (fast, cheap, iterating: Unitree) is Chinese. The honest position is that 19% is not an analyzable number; it is a reservation of narrative space, and the market's willingness to hold it there — through five years of missed dates — is Section 3's asymmetric-credibility regime doing its heaviest lifting. The Fremont reveal this summer converts it, one way or the other, from theology to a unit count.

11. Energy: the second curve nobody prices at 6%

Inverted neglect: the business the SOTP models weight least is the one compounding on receipts. Tesla Energy deployed 46.7 GWh of storage in 2025 (+49%), generating ~$12.8 billion (+27%) at gross margins that touched 29.8% — a record, and structurally above the automotive business. The Shanghai Megafactory went from groundbreaking to production in seven months; Houston's Megapack 3 plant targets 50 GWh annually; the LG Energy Solution deal ($4.3 billion, Michigan-made LFP cells) answers the 82.4% effective tariff on Chinese cells that cost ~$200 million in a single quarter. Headwinds are real — Q1 2026 deployments actually fell 15% before Q2's 13.5 GWh recovery, and BYD passed Tesla as the world's largest storage supplier in 2025, with CATL and Sungrow crowding the segment — but this is a genuine, growing, margin-rich business riding the same datacenter-and-grid buildout every report in this series has priced from another angle. At utility multiples on 2026 run-rate economics it plausibly justifies most of its 6% weighting by itself; at "AI infrastructure" multiples, considerably more. It is the quiet asset in the file — and the only Tesla segment whose customers sign the kind of contracts this series calls receipts.

For series readers, note who the customers are: the same hyperscalers and grid operators funding the four tenants' data-center buildout — Megapacks increasingly ship to the power infrastructure of AI campuses, which makes Tesla Energy a genuine (if second-order) beneficiary of the same capex wave this series audits from Meta through Amazon. It is the one channel through which the AI trade pays Tesla receipts rather than the reverse. A strategist inclined to own "Tesla's execution culture without Tesla's promise multiple" keeps returning to this segment and finding it un-investable separately — which is, perhaps, the point of keeping it consolidated.

12. The Musk complex: key-man risk as an asset class

Every governance thread in this report converges on one person, and the exposure now runs in both directions. Inventory the Muskonomy as of July 2026, because Tesla is now one holding in a portfolio: SpaceX (absorbing xAI and X at a combined ~$1.25 trillion private mark, IPO mooted — a company our series priced separately, whose Starlink business alone out-earns Tesla's net income); xAI (Grok, the Colossus clusters, $2 billion of Tesla's cash as an investor); X (the distribution machine); Neuralink and Boring (options); and the political franchise (parked, not sold). Tesla's claim on the portfolio manager's attention is enforced by exactly one instrument — the comp package — against four ventures that are earlier, needier, and in at least one case (SpaceX pre-IPO) carrying a richer personal payoff on the same effort. That is the actual key-man structure: not "what if something happens to Musk" but "what does Musk's own portfolio math say about marginal hours," and the observable allocation (engineers migrating to xAI, the Bloomberg reporting on his 2026 calendar) already answers it directionally. The dependence: Tesla's valuation is substantially a bet on Musk's execution of autonomy and robotics, yet his attention is divided across SpaceX (absorbing xAI and X, pre-IPO), politics (the America Party flirtation, now parked in favor of funding Republican midterm campaigns), and the feud-reconciliation cycle with the White House that whipsaws the brand. The entanglement: Tesla invested $2 billion in xAI over a functionally failed advisory vote; engineers migrate between Musk entities; related-party transactions run to hundreds of millions; and the bylaws now require 3% ownership to bring derivative claims — a threshold that insulates nearly everything. The executive bench keeps thinning (the manufacturing chief fired, the sales chief gone, the Cybertruck and Model Y program leads resigning within days of the comp vote). And the compensation package itself is best understood as the board's answer to the attention problem: $1 trillion of contingent equity as a retention-of-focus instrument — the most expensive attention purchase in corporate history, ratified by shareholders who concluded that Tesla without Musk's focus is worth less than Tesla diluted 12% for it.

The chip strategy, for series readers, sits inside this complex: Dojo's shutdown ended the pretense of full-stack AI independence; training now rents from NVIDIA (Cortex: ~81,000 H100-equivalents, headed to 100,000 — another tenant queueing for the memory complex's HBM); inference bets on AI5/AI6 through TSMC and the Samsung Taylor fab whose $16.5 billion contract our SK hynix report flagged as Samsung foundry's anchor win. Tesla is, in the AI economy's four-sided market, a mid-sized tenant with a silicon roadmap — real, but a consumer of the infrastructure trade, not a driver of it.

Dojo's death deserves a coda, because it is the one place where this company's promise-inventory was actually marked to market — by management itself. Dojo was, for three years, the same class of asset as Optimus is today: a Musk-announced moonshot (a custom training supercomputer to break NVIDIA dependence) that analysts dutifully carried in SOTP models, that Morgan Stanley once argued was worth $500 billion, and that was quietly shut down in August 2025 as an "evolutionary dead end," its team scattered, its function outsourced to the vendors it was built to displace. No impairment ran through any income statement, because promises don't sit on balance sheets — but $500 billion of sell-side option value evaporated without a line item, and the stock barely noticed. That is the accounting regime of the story stack: promises are added at announcement and removed without ceremony, and only the survivors are visible at any moment. An investor pricing today's 85% should know that the inventory has already had a major writedown, handled precisely this way — invisibly.

Part Three · Synthesis: the map, the territory, and the deadline

13. Where it stands: the July 2026 snapshot

The engine, five quarters:

Quarter Revenue Deliveries Gross margin Auto GM (ex-credits) Credits GAAP net income
2025 Q2 $22.5B (-12%) 384,122 17.2% ~15.0% $439M —
2025 Q3 $28.1B (+12%) 497,099 18.0% 15.4% $417M —
2025 Q4 $24.9B (-3%) 418,227 (-16%) 20.1% 17.9% ~$540M $840M (-61%)
2026 Q1 $22.4B (+16%) 358,023 (+6%) 21.1%* 19.2%* $380M (-30% QoQ) $477M
2026 Q2 (reports 7/22) 480,126 (+25%, record Q2) — — — —

(*Q1 margins include one-time warranty adjustments and ~$250M of tariff refunds; underlying margins run lower. The regulatory-credit line — $11.8 billion of cumulative near-pure profit over a decade — is legislated toward roughly zero by 2027.)

Read the table's texture: revenue growth (+16% in Q1) now runs ahead of delivery growth (+6%) on energy and services mix — a favorable composition shift — while the margin line's apparent recovery (21.1%) needs its one-time asterisks removed to see the underlying ~19% that a 10%-discount Chinese price war keeps pressing. The credits column is the quiet countdown: from ~$540M to $380M in two quarters, on its way to approximately zero by 2027 — call it $1.5–2 billion of annual pure profit exiting a company that earned $477 million total last quarter. The bears' cleanest arithmetic in the entire file is that line: at current run-rates, the credit cliff alone exceeds GAAP net income, meaning the reported profitability of the 12%-weighted collateral business is, on a two-year view, substantially an artifact of a regulatory regime that has already been repealed.

Full-year 2025: 1.64 million deliveries (-9%, the crown lost to BYD), revenue $94.8 billion (-2.9%, the first annual decline ever). 2026 consensus: ~1.65 million deliveries — roughly flat — though Q2's beat may force upgrades. The balance sheet is genuinely strong: $44.7 billion of cash against ~$16 billion of debt. The cash flow is genuinely deteriorating by design: Q1's +$1.4 billion FCF comes with management guidance that remaining 2026 quarters run negative, as capex nearly triples to ~$25 billion (six factories, the AI5 fab plans, Optimus lines, Cybercab, Cortex). Note the proportion against the series' tenants: $25 billion is an eighth of Meta's program on a fortieth of Meta's operating income — relative to its earnings power, Tesla is running the most aggressive capex-to-profit ratio in the Mag 7, funded by balance-sheet drawdown rather than operating gush, with none of the receipt mechanisms (backlog, contracts) the market demands of everyone else. Valuation: $1.48 trillion — tied with Meta for sixth in the Mag 7 — at ~380× trailing and ~196× forward earnings (correcting the stale screen data; the multiple has expanded as earnings shrank), against sell-side anchors of $125 (Wells Fargo), ~$400–471 (Morgan Stanley, BofA), and $2,600 (ARK, 2030). No dividend, no buyback; the shareholder-return instrument is, and has only ever been, the multiple itself.

14. The compensation package as confession: management's own probability map

This series reads capital-markets behavior as disclosure, and Tesla's November 2025 compensation package is the richest such document in the Mag 7 — because unlike an ADR or a buyback, it enumerates. To pay out fully, the following must occur within roughly a decade: market capitalization of $8.5 trillion (nearly six times today's); 20 million cumulative vehicles (roughly 12 million more than delivered to date, implying sustained volume growth the last two years flatly contradict); 10 million FSD subscriptions (twenty-one times the conservative current count); 1 million robots delivered (from zero); 1 million robotaxis in operation (from ~42); and $400 billion of adjusted profit milestones (an order of magnitude above the current run-rate).

Set the package beside its own prehistory for scale. The 2018 package — $56 billion, voided by Delaware, restored by its Supreme Court in December 2025 and delivered in April — was itself the largest compensation instrument ever written, and its milestones (a $650 billion market cap that seemed absurd in 2018) were all achieved, which is the strongest single datapoint the bulls own: the last time this board wrote a "impossible" milestone contract, the milestones fell. The 2025 package multiplies that bet eighteen-fold in notional and shifts its character: the 2018 gates were mostly financial (market cap, revenue, EBITDA — things a great car company could deliver), while the 2025 gates are mostly categorical (robots, robotaxis — things that require new industries to exist). Same instrument, different collateral: the 2018 package was a leveraged bet on execution; the 2025 package is a leveraged bet on invention. The board, at least, is consistent — it prices Musk's invention capacity the way the market does, and it has now put a decade clock and a public scoreboard on it.

Read as a confession, the package says three things. First, the board and the controlling shareholder-CEO believe — or at minimum are willing to be paid in — outcomes that require every story business to succeed simultaneously; there is no tranche structure for "the cars recover and energy compounds." Second, the milestones date-stamp the promises: a decade-long package whose early tranches vest on a $2 trillion market cap converts "eventually" into a schedule, against which the ~20-car robotaxi fleet and the zero-useful-work robot count become annually measurable shortfalls. Third — and this is the tell the 75% approval vote ratified — the shareholders themselves have accepted that the equity is a venture portfolio wearing an automaker's income statement, and priced Musk's continued attention as its single largest asset. Every prior report in this series found companies whose disclosures understated their positions (Amazon's excluded backlog, Microsoft's unmarked stake). Tesla is the inversion: the official documents overstate — they encode the maximal story as compensation math — and the analytical task is subtraction rather than discovery. (The FSD subscription-only pivot, Section 9, is the minor confession beside the major one: a company confident in imminent autonomy does not convert $15,000 upfront payments into $99 monthly ones; it converts monthly payments into upfront ones.)

15. Three scenarios for the year ahead

Anchors: ~$95–100B 2026 revenue consensus, ~$25B capex, negative FCF guided, ~196× forward at ~$393.

Scenario A — v15 converges. The FSD rewrite ships late 2026/early 2027 and demonstrably works: robotaxi fleets scale from dozens to thousands across the launched metros, ride counts get disclosed (disclosure itself would be the tell), Optimus's Fremont line produces real units into real jobs, and the Q2 delivery momentum holds through the low-cost lineup. The SOTP models' 85% suddenly has receipts; the stock re-rates toward ARK's territory and the first comp tranches vest. This is genuinely possible — the engineering organization has repeatedly done what specialists called impossible, from Shanghai's twelve months to the 2020 ramp — and it is the only scenario in this series whose payoff is measured in multiples rather than percentages. Its probability is the entire debate; its timing requirement is now contractual.

Scenario B — the grind of promises (modal). V15 slips or lands ambiguously; robotaxi cities multiply while fleets stay in the dozens ("expansion theater" — new geofences as press events); Optimus reveals another prototype; the car business stabilizes at Q2's level on cheap variants and energy compounds quietly; credits bleed out of the P&L. Earnings stay thin, the multiple stays theological, and the stock oscillates violently around flat — 2025's pattern (+11%) extended. The instrument's character in B: maximal volatility, zero yield, catalysts always one quarter away. B can persist for years precisely because Section 3's reflex — gaps close upward — keeps a buyer under every dip; it ends only when A or C forces it. Note what B does to the comp package's internal clock, though: the early tranches need a $2 trillion market cap, and a multi-year grind at $1.5 trillion means the most expensive attention-retention instrument ever written pays its intended recipient nothing — at which point the attention it was purchased to retain has, by revealed preference, three other companies bidding for it. B is stable for the stock and unstable for the governance; that asymmetry is its exit mechanism.

Scenario C — the deadline arrives. Some combination of: NHTSA's EA26002 escalates to a recall or operational restrictions on FSD/robotaxi just as scale-up begins; a high-profile robotaxi fatality freezes expansion (the Katy investigation is the template); the credit cliff plus price-war margins produce an earnings tape too thin to ignore at 380×; BYD's cost curve forces another price cut cycle; or the Musk attention complex produces a governance event (an xAI/SpaceX consolidation proposal involving Tesla would be the classic). The multiple does what promise-multiples do when the deadline passes unmet — Wells Fargo's $125 implies roughly -70%, and unlike the 2022 drawdown there is no efficiency lever (costs are already the moat) and no buyback backstop. C's honest probability is material: this is the only company in the series where the modal sell-side model (both bull and bear versions) requires businesses that do not yet exist.

C also carries a quiet second-order effect for series readers: a Tesla promise-collapse would be the AI era's first large-scale demonstration that story multiples can die without a recession — a repricing template the market would immediately shop against every other unaudited claim in the index, Meta's capex faith and the private AI labs' marks included. Conversely, Scenario A — verified autonomy at automotive scale — would be the single largest validation event the physical-AI narrative could receive, with read-through up the chain to compute demand (robotaxi fleets are inference farms on wheels) that our tenant reports' Scenario A's would inherit. Tesla is uncorrelated with the four-sided market in daily price and maximally correlated with it in narrative endgame: it is where the story economy's credibility gets settled first, in public, on a schedule.

16. The valuation paradox, Tesla edition — the series' control case

Every prior report found a version of the same paradox: strong receipts, punished multiple (the memory complex), or contested receipts, discounted multiple (the tenants). Tesla completes the series as the control case: no receipts, maximal multiple — proof that the market's receipt-auditing regime of 2026 is selectively applied, suspended for exactly one name on the strength of one man's execution mythology and the squeeze-scarred reflexes of its shareholder base.

Why does the suspension persist? Three structural supports, none of them fundamental. Flow: Tesla is the retail era's signature holding and the most-traded optionable stock on earth; its price formation runs through weekly options and thematic ETFs to a degree that mutes fundamental arbitrage — shorting a 380× multiple has been a career-ending trade twice, and the survivors' absence is itself a price support (the market microstructure equivalent of Section 3's liquidated skeptics). Index: at sixth in the S&P's weighting, every passive dollar buys Tesla's promises automatically, no audit performed. And narrative infrastructure: no other company's story is retold daily by its CEO to 200 million followers on a platform he owns — vertical integration of the multiple itself. These supports explain persistence; they do not repeal arithmetic. They specify its schedule: flow-supported multiples do not decay gradually, they gap on disillusionment events, which is why the risk matrix below weights discrete catalysts over erosion. The comparison table is the argument: Meta pays 15.8× forward for a 33%-growth ad machine with $17 billion quarterly net income; Tesla pays ~196× for a shrinking-revenue automaker with $477 million — a 12× multiple premium for a 35× earnings deficit, justified entirely by futures the company's own compensation schedule dates to the 2030s. The full family for the record (July data): NVIDIA $4.72T (29.8×/15.3×), Apple $4.53T (37.3×/32.1×), Alphabet $4.39T (27.5×/24.7×), Microsoft $2.90T (23.3×/20.2×), Amazon $2.61T (~29×/~24.5×), Meta $1.48T (21.2×/15.8×), Tesla $1.48T (~380×/~196×) — one of these rows is not a valuation, it is a belief with a ticker. Within this series' framework the conclusion writes itself: Tesla is the single largest un-receipted claim in the AI economy, larger in promise-dollars than Meta's capex program, carried at a valuation that assumes the promises while the operating business pays none of the carrying cost. What the framework cannot supply is the probability that the promises land — that is an engineering judgment the market has crowdsourced to one individual's track record — and honesty requires stating both that the track record is extraordinary and that its two greatest vindications (2013, 2020) were rescues of product promises, while the current promises are of a different kind: statistical safety cases and humanoid economics, domains where charisma has no derivative. The multiple prices the pattern repeating. The scoreboard — 42 cars, zero robots — prices the base rate. July 22, and every quarter after, arbitrates.

17. What would have to be true, the risk matrix, and five signals

One more piece of honest bookkeeping before the list: what would make this report wrong in the bullish direction. If FSD v15's architecture genuinely delivers the 10× parameter scale-up with commensurate capability — and the June rollout of v14 to the older HW3 fleet suggests the deployment machinery works — then the fleet-count signal could move faster than any regulator-gated model predicts, because Tesla's marginal robotaxi is a software flag on an existing car, not a manufactured vehicle. The dozens-to-thousands transition that took Waymo five years could, in the best case, take Tesla five quarters. That is the specific, falsifiable form of the bull case this report considers strongest, and the fleet counts below are designed to detect it early in either direction.

For Scenario A to pay from ~$393, most of the following must hold:

  1. FSD v15 ships within two quarters of guidance and produces a step-change in disengagement/safety data visible to regulators, not just release notes;
  2. Robotaxi fleets grow by two orders of magnitude within 18 months (dozens → thousands) with disclosed ride volumes;
  3. Optimus's Fremont line produces units performing commercial work in 2026 — not prototypes, deployments;
  4. The car business holds Q2's trajectory through the credit-cliff year (the collateral cannot crack while the bets mature);
  5. NHTSA's EA26002 resolves without operational restriction.

Falsify (1) or (5) and the robotaxi timeline — 62% of the sell side's own value attribution — moves right by years at 196× forward.

Risk matrix (probability × severity):

Signal design notes, as throughout the series: ordered by arrival, chosen for independence — the earnings print reads the collateral's true margins, the Fremont reveal reads the robot promise, the v15 window reads the software promise, the NHTSA disposition reads the regulatory gate, and the fleet counts read the only number that cannot be narrated. Signals 1 and 2 land within weeks; 3 and 4 within roughly two quarters; 5 compounds continuously. Uniquely in this series, all five of Tesla's signals are tests of the company's own claims rather than of external conditions — no macro reading, no counterparty audit, no October capex season dependency. Tesla's fate, alone in the Mag 7, is not entangled with the four-sided market's shared wager: it is a single-name referendum on a single organization's ability to deliver enumerated miracles on a contractual schedule. That independence is either diversification or isolation, depending on how the miracles go.

Five signals, in firing order:

  1. Q2 earnings (July 22): the margin behind the delivery beat (was +25% bought with mix collapse?); credit line decay; any robotaxi ride-count disclosure — the absence of numbers is itself the number. Watch also the energy line (13.5 GWh deployed should print record segment revenue) and the capex cadence against the $25 billion guide: the quarter that proves whether negative FCF arrives on schedule.
  2. The Optimus Fremont reveal (this summer): production-intent hardware and a named commercial deployment, or another demonstration — the 19% weighting's first date-stamped test.
  3. FSD v15's actual ship window (late 2026 guidance): the event on which 62% of the sell-side value model converges.
  4. NHTSA EA26002 disposition: engineering analysis → recall, restriction, or closure; the regulatory fork for the entire autonomy thesis.
  5. Robotaxi fleet counts, quarterly (city filings, not press releases): dozens → hundreds → thousands is the only trajectory that validates the 45%; Waymo's million-rides-per-week year-end target is the pace car, and the Texas DMV registry — 42 versus 577 today — is the single most honest scoreboard in the entire Tesla file, updated monthly, immune to keynotes.

Two calibration notes for position thinking (analysis of the instrument, not advice). Correlation: despite the AI framing, TSLA's price is nearly orthogonal to the four-sided market this series maps — it did not participate in the memory complex's moves, the July 1 Meta Compute session, or the tenants' capex repricings; its factor exposure is retail flow, Musk headlines and rate-sensitive long-duration narrative, making it a diversifier inside an AI-heavy book in the statistical sense while being the purest AI promise in the fundamental sense. Skew: at ~196× forward with binary catalysts, the return distribution is bimodal — both the ARK case (+560%) and the Wells Fargo case (-70%) are lognormal-implausible and scenario-plausible, which is why options markets price TSLA volatility at a permanent premium and why the honest summary of the instrument is: a venture-capital position with daily liquidity and an index weighting. Venture positions are sized as venture positions.

The bottom line. Tesla is the greatest execution story in industrial history attached to the largest unaudited claim in the AI economy. It has died three times and been resurrected three times by exactly the mechanism its bulls now rely on — a promise, kept barely in time, repricing everything. The difference in 2026 is that the promises have been enumerated, dated and priced into a compensation contract; the competitive scoreboard publishes weekly; and the collateral business, for the first time, is fighting structural decline in two of its three markets while the multiple assumes it doesn't matter. This series ends each report by naming what the price contains. Tesla's price contains the pattern — Musk delivers, late but decisively — applied to domains where lateness is measured by regulators and rivals scaling at 10× per two years. The five signals above will not settle whether the pattern holds. They will settle something more useful: whether the market's patience or the promises' deadlines expire first. Watch the fleet counts. Everything else is theater until they move.

And place this report's subject beside the series' first, because they bracket the same economy. SK hynix sells the most contracted, most audited, most physically verifiable claim in AI — memory, pre-sold three years forward — at 5.5× forward earnings. Tesla sells the least verifiable claim — autonomous labor, undelivered — at 196×. Between those two prices lies every question this series has asked about what the market believes, whom it audits, and why. The memory company's earnings are priced to die; the promise company's are priced to be born. It would be a fitting irony, and a fitting end to this survey, if the cycle's eventual lesson turned out to be the oldest one in markets: that the future was cheapest where it was already signed for.

Sources

Report generated by the Aya Invest research pipeline. Every claim above traces to a public source; where figures are contested (FSD subscriber counts), estimated (robotaxi economics, Optimus valuations), or derived from leaked/contested documents, the text says so. Trailing/forward multiples corrected to July 2026 source data (~380×/~196×) where screen data was stale. For information and research purposes only. Not investment advice.

FAQ

Why does Tesla trade at ~196x forward earnings?

Because the market is not pricing the car business — BofA's sum-of-the-parts attributes roughly 45% of the value to robotaxi, 19% to Optimus, 17% to FSD licensing and 6% to energy, leaving about 12% for the cars. The report calls this the promise inventory: value booked when promises are announced, never formally written off when they die (Dojo's shutdown erased an estimated $500B of narrated sell-side value with no market reaction). Three structural supports keep the multiple aloft: options flow, passive index buying, and a CEO with 200M followers who vertically integrated the multiple itself.

How does Tesla's robotaxi actually compare to Waymo?

As of mid-2026 Tesla operates roughly 42 driverless vehicles in its Austin-area service while Waymo delivers about 500,000 paid driverless rides per week across multiple cities. Tesla's bet is that a camera-only, fleet-software approach scales exponentially once validated — turning millions of existing cars into robotaxis via software. The gap is the trade: if the bet lands, today's numbers are irrelevant; until it lands, the robotaxi segment that carries ~45% of the SOTP has receipts measured in dozens of vehicles.

What does the $1 trillion comp package mean for investors?

The 2025 package ties Musk's payout to 12 milestone tranches — including 1 million robotaxis in operation, 1 million Optimus robots delivered, and an $8.5T market cap. The report reads it as management's official sum-of-the-parts: the board's own probability-weighted map of which stories must come true. Bulls note every milestone of the 2018 package was hit. The asymmetry the report flags: 2018 paid for execution of a known product roadmap; 2025 pays for invention — robotaxi and humanoid robotics at scale do not yet exist as businesses.

© Aya Research · About · invest.aya-ai.org · For information and research purposes only. Not investment advice.