TSLA.US: Auto profits are thin; the multiple is Robotaxi, Semi, and Optimus

Latest snapshot

Q2 revenue

USD 28.2B

Q2 gross margin

0.2

Q2 GAAP NI

USD 1.1B

Q3 deliveries

487K

As of 2026-10-03. The income statement stops at 2026-06-30. Q3 is production, deliveries, and storage only, filed 2026-10-02; the financial webcast is 2026-10-21. Price is the 2026-10-02 close. Not advice.

Thesis

Tesla is two ledgers on one ticker.

The operating ledger grew. Q2 revenue was $28.236B, +26% YoY, and trailing-twelve-month revenue cleared $100B for the first time ($103.6B across the last four quarters). Model 3/Y deliveries were 467,762; total deliveries 480,126. Storage deployments were 13.5 GWh. Services and other revenue was $4.581B (+50%), with a record services gross profit. The core business is not stalling.

The valuation ledger is an option. The Oct 2 close was $370.59, Tencent’s market-cap field $1,309.3B. Trailing GAAP net income is $3.804B, which is about 344x on that capitalization; the vendor PE is 343.1x. Q2 operating income was $398M, a 1.4% operating margin, and free cash flow was -$1.092B. Auto earnings do not pay for the equity. Unsupervised miles, a real Semi fleet, and an Optimus that can be leased as labor are what the price is arguing.

Q3 draws the split more sharply. The 8-K reports production of 464,391, deliveries of 486,532, and 13.7 GWh of storage. Deliveries are 2.1% below Q3 2025’s 497,099 and 1.3% above Q2. “Other models” deliveries were 8,295, about half of 15,933 a year earlier — after S/X production ended, Cybertruck and a just-starting Semi do not fill the line. Tesla’s own release says deliveries and deployments are not a guide to quarterly results. Do not invent a Q3 income statement from them.

Near-term, the question is whether price, opex, and capex let the operating margin leave the 1% area. Further out, the question is whether unsupervised service becomes a mileage statistic rather than a list of six metros, and whether Semi and Optimus get a company-reported unit count. Until Oct 21, deliveries stay an operating print.

Business

Tesla sells vehicles, storage, and a services stack. The robot is still an option, not a segment.

Automotive was $20.516B (+23%), 72.7% of Q2 revenue. The flagship is still Model 3/Y: production 442,936, deliveries 467,762, about 2% under operating leases. Installed annual capacity in the shareholder update — not the current run rate — is >550k for California 3/Y, >950k in Shanghai, >375k Model Y in Berlin, and >250k Model Y in Texas. Model YL launched in the US in July. Cybertruck’s capacity line is >125k, but Q3 “other models” deliveries shrank hard.

Two competitive facts still show up in the numbers, and one is weakening. The network is real: four vehicle plants, in-house cells, lithium refining, and 82,357 Supercharger connectors (+17%). Paid FSD (Supervised) subscriptions were 1.48 million, +56%, and more than 55% of North American deliveries took a subscription. What weakened is hardware margin. Gross margin fell from 21.1% in Q1 to 16.8% in Q2 while operating expenses rose 47% to $4.353B. Attach went up; operating margin did not.

Energy is the second curve and not yet the profit beam. Q2 revenue $3.139B (+13%), deployments 13.5 GWh (+41%), with EMEA and Megafactory Shanghai called out. Installed Megapack capacity is 40 GWh in California and 20 GWh in Shanghai; Texas is still commissioning, with Megapack 3 and Megablock planned this year. Q3 deployments of 13.7 GWh beat last year’s 12.5 and missed the Q4 2025 record of 14.2. Storage dampens auto seasonality. It did not stop group gross margin from falling in Q2.

Services and other is the under-read profit line: revenue $4.581B, gross profit $648M, a 14% margin, +$302M of gross profit sequentially. Maintenance, insurance, charging, and software all contributed. Shifting FSD from an upfront purchase to a subscription at delivery pressures ASP now and turns software into a function of the fleet later. Services are 16% of revenue. That does not justify 340x earnings. It does explain how services can grow 50% in a quarter when deliveries are merely solid.

Robotaxi and Cybercab are a strategy without a segment. The Q2 update says Cybercab production started at Giga Texas, public-road engineering tests began, and employee rides on the Texas campus started in July. Unsupervised operations expanded in Austin and opened in Miami, Orlando, and Tampa in July. Dallas and Houston were ramping unsupervised. The Bay Area remained FSD (Supervised) with a safety driver under California TCP permit TCP0046782-A. Phoenix and Las Vegas were still in preparation. The “seven metros” line includes that supervised Bay Area city; the consumer page, checked 2026-10-03, lists Austin, Dallas, Houston, Miami, Orlando, and Tampa and does not publish a fleet count. The cumulative-miles chart in the update is not converted here into a point value off the axis.

Semi has crossed from construction into a company claim of high-volume production, and it has not crossed into a disclosed delivery total. At the Q2 update the Nevada plant was commissioning, with production scheduled for 2026. In late September Tesla said, in posts reported by Supply Chain 24/7, that this is its first factory built for Semi, sized for 50,000 trucks a year, about 1.8 million square feet, with Standard, Long Range, and a European spec on one line and in-house 4680 cells, and that customer deliveries were starting. Fifty thousand is capacity. Price, cumulative deliveries, and a date for full rate were not disclosed. Uptime “above 98%” in 2026 and a 30-minute Megacharger top-up of up to 60% of range are company statements without a published definition in that thread. The product was unveiled in 2017 against a 2019 production target; PepsiCo’s hand-built trucks arrived in late 2022. This is a late production line, not proof of fleet demand.

Optimus is earlier still. The company has decommissioned S/X at Fremont, is installing first-generation lines, expects production in 2026, and says the first builds go to Optimus Academy for training data. Texas construction is underway. On July 1 Musk wrote that early production would be extremely slow because the parts are new. That is the company boundary. Electrek, citing The Information on Sept 25, reported a rise from a few dozen units a week in Q2 testing to several hundred in August, a year-end aim above 1,000 a week, and hands that are still largely hand-assembled. A separate Electrek note on an August JPMorgan meeting relayed external sales “as early as” the second half of 2027. None of those rates are 10-Q figures. They stay out of the KPI row until Tesla reports them.

Entity: Tesla, Inc., incorporated in Texas, headquartered in Austin. Gigafactory Shanghai is wholly foreign-owned, not a joint venture. Reuters on July 31, citing the Wall Street Journal, reported that Tesla is considering a separation of the China business ahead of a possible SpaceX merger, to firewall a defense contractor from Chinese assets. That is a report of deliberations, not a signed deal and not an 8-K. The US business, energy, and the FSD data loop remain inside one listed company.

12–24 month pillars

12–24 month pillars

Pillar	What is known
汽车：利用率优先于新工厂	Q2 信：先吃满现有产能。3/Y 仍是交付主体；S/X 弗里蒙特产线已拆除改 Optimus。
Robotaxi / Cybercab	德州 Cybercab 已开工；无人运营在德州与佛州多城爬坡。车队规模公司未在股东信披露。
Semi	Q2 时点内华达工厂为调试；此后公司称高产量产与交付开始。5 万辆/年是产能，不是交付指引。
Optimus	一代线安装中，公司预计 2026 年投产，首批用于 Optimus Academy。周产量未经公司审计披露。
能源与电池	储能回到增长；Megapack 3 / Megablock 计划今年在德州投产。电芯仍是汽车放量的短板。

Value chain

Tesla takes the vehicle spread, the storage project, and service fees on the installed base, and it is trying to extend one vision stack to robotaxis and a humanoid. It is not a contract manufacturer and not a pure software licensor.

Three upstream lines are on the company’s own capacity table and are explicitly early or under construction: Texas 4680 cells at >40 GWh in production, cathode 10 GWh and lithium refining 30 GWh in early ramp, Nevada LFP 7 GWh in early ramp for storage, Berlin 4680 under construction. The Q2 letter says pack capacity is still the main limit on raising vehicle production. On training compute, Cortex 1 is >90 MW and Cortex 2 >115 MW, and onsite Texas compute more than doubled in megawatts in the first half. The Austin semiconductor fab is still in construction and procurement. It does not help this year’s gross margin.

Downstream is retail buyers, fleets, utilities, commercial storage customers, and non-Tesla drivers on the Supercharger network. Auto retail is fragmented. Neither the shareholder update nor the companyfacts extract used here discloses a top-five or a >10% customer, so concentration is undisclosed, not estimated. Storage projects can make a quarter look concentrated, and revenue recognition looks more like contracting than retail.

Three supply risks, echoed in the risk table:

1. Cells. The company names packs as the constraint on vehicle volume. 4680 cells are supposed to feed both Cybercab and Semi, while storage needs LFP. A slow ramp opens a gap between orders and deliveries.
2. China. Reuters, quoting Tesla China’s Grace Tao, says Shanghai accounted for more than half of 2025 global deliveries, with 400-plus local suppliers and 60-plus of them also serving global plants. That localization is the cost advantage and the decoupling cost. The Wall Street Journal has reported a plan to stop using China-based suppliers in US factories by 2027. If destination markets tariff Shanghai exports, the export-hub strategy hits utilization.
3. Chips. Autonomy and the robot both consume inference silicon. The in-house fab is a multi-year project. Until it runs, training and vehicle inference depend on outside supply.

Customers

Customers

Item	Disclosure
汽车零售	面向消费者与车队，客户高度分散。本次引用的 10-Q/股东信未披露单一客户超过 10%。
储能	公用事业与工商项目，单季部署可以很集中，但公司未公布前五大客户占比。
服务与 FSD	装机车辆上的订阅、保险、维保与超充。Q2 付费 FSD 订阅 148 万，同比 +56%。
账期与地域	汽车以交付确认收入为主。路透援引陶琳：2025 年上海占全球交付一半以上，并承担出口。2026 上半年占比不以未经 10-Q 核对的转载为准。

Corporate events

The events that matter for the equity are not another concept reveal. They are S/X ending so Optimus can have the floor, Q2 margins given up to the investment cycle, Robotaxi leaving Austin for Florida, a $5B revolver replaced by $30B of undrawn commitments, and a Q3 delivery print that says auto demand is steady rather than re-accelerating.

Recent events

Recent events

When	Type	Event	So what
2026-05	制造	弗里蒙特停产 Model S/X，产线改 Optimus	旗舰轿车/SUV 退出，机器人占用成熟工厂
2026-07-02	运营	Q2 交付 480,126，储能部署 13.5 GWh	汽车与储能同时放量，财务质量要等利润表
2026-07-22	财报	Q2 营收 $28.236B，经营利润仅 $0.398B，FCF -$1.092B	收入创新高，利润被费用和资本开支吃掉
2026-07	Robotaxi	迈阿密、奥兰多、坦帕无人运营；Cybercab 员工试乘	从奥斯汀试点扩到佛州，仍无车队 KPI
2026-09-25 前后	Semi	公司称内华达工厂进入高产量产、交付开始	产能 5 万辆/年；实际下线数未披露
2026-09-29	融资	新签 $300 亿未提用授信，终止原 $50 亿循环贷	为资本开支预备流动性，不是已举债
2026-10-02	运营	Q3 交付 486,532，储能 13.7 GWh	交付同比约 -2%；利润表 10 月 21 日才公布

The Sept 29 8-K needs its own paragraph. Tesla entered a $20B senior unsecured three-year delayed-draw term loan (at most ten draws over 18 months; undrawn commitments step down to $10B on the first anniversary and $5B at 15 months; loans mature 2029-09-29), an $8B five-year revolver maturing 2031-09-29 with up to $500M of letters of credit, and a $2B 364-day revolver maturing 2027-09-28. Nothing was drawn. Tesla said it does not plan to draw in 2026. The same day it terminated the January 2023 $5B revolver, which had no borrowings and no early-termination fee. The new agreements require at least $5B of consolidated liquidity. This is dry powder for the capex cycle, not new interest-bearing debt on the balance sheet.

Valuation

Valuation snapshot

Valuation snapshot

Metric	Value	Basis
现价 / 市值	$370.59 / $1,309.3B	腾讯财经 gtimg，2026-10-02 16:00 ET 收盘
TTM PE（行情商）	343.1x	gtimg 字段；与下述 GAAP 自算接近
TTM PE（GAAP 自算）	约 344x	市值 $1,309.3B / 四季 GAAP 净利 $3.804B
PB（行情商）	16.85x	gtimg；口径与账面权益不完全相同
P/B（账面）	约 15.1x	现价市值 / 2026-06-30 股东权益 $86.858B
股息率	0	companyfacts 无分红；公司不派息，价格回报即总回报
盈利收益率	约 0.29%	1 / 343，说明倍数定价的不是当期利润

Valuation · ~3 years

Interactive chart available in the reader.

What the price is discounting: about $1.31T of equity for about $3.8B of trailing GAAP earnings, a ~0.3% earnings yield. The same method at the end of Q1 2024 — price $175.79, TTM earnings still $13.9B — was roughly 44x. The multiple expanded because the denominator collapsed, from $15.0B of 2023 net income to $3.79B in 2025, while the stock reached $444.72 at the end of Q3 2025. The Q3 2026 quarter-end close was $354.81, 20.2% below that prior-year quarter. A drawdown from the high has already happened. The absolute multiple is still extreme.

PE and PB are shown only in quarters that have a diluted-share frame and four reported earnings quarters. There is no Q4 share frame, and Q3 2026 earnings are not filed, so those points are null, not zero. The dividend yield is 0 for the whole window: companyfacts has no dividend series, so price return is total return. Book PB uses quarter-end stockholders’ equity. The vendor’s 16.85x and the ~15.1x computed on book equity are different objects.

Price · quarter-end

Interactive chart available in the reader.

Financial trends

Revenue · 8 quarters

Interactive chart available in the reader.

Gross margin · 8 quarters

Interactive chart available in the reader.

Net income · 8 quarters

Interactive chart available in the reader.

Revenue bottomed in Q1 2025 and has since posted year-on-year growth in three of the last four quarters. Q2 2026 at $28.236B is the high of the eight-quarter window, +26.1% sequentially and +25.5% year on year. Q4 2025 revenue of $24.901B matches the shareholder update; it is FY $94.827B minus the first three quarters, not an interpolation.

Gross margin is a different shape. Q1 2026 at 21.1% is the high of the window; Q2 fell back to 16.8%, a relative decline of about 20% versus Q1 and only about two relative points versus the prior year. Operating margin fell further, from 4.1% in Q2 2025 to 1.4%, because opex grew nearly twice as fast as revenue. A gross-margin chart without the opex line overstates the health of group profit.

Net income has a base-effect trap. The Q4 2023 residual is $7.928B, far above neighboring quarters, so Q4 2024’s −73% year on year is not an operating disappearance. 2024 quarters use the later companyfacts frames; the Q2 2026 update notes that 2024 was recast for the crypto-asset standard. Q2 2026 GAAP net income was $1.114B (−5%); non-GAAP net income $1.153B (−17%); diluted EPS $0.32. Adjusted EBITDA was $3.273B, an 11.6% margin, down 353 bp. EBITDA still stands; operating income is thin.

Financial health

Financial health

Financial health

Item	Reading	Note
判断	流动性厚，自由现金流转负	投资周期，不是偿债危机
经营现金流	Q2 $4.697B；FY2025 $14.747B	单季来自 Q2 股东信；年度来自 SEC
自由现金流	Q2 -$1.092B	资本开支 $5.789B，同比 +142%
现金 / 现金+短投	$15.219B / $43.524B	2026-06-30。短投是流动性的大头
有息债	已提用余额：N/A	LongTermDebt 无 CY frame，不估算。9 月 29 日 $300 亿授信均未提用
授信契约	合并流动性 ≥ $5.0B	8-K；公司称 2026 年不计划提款
审计 / 持续经营	未见持续经营强调	意见类型以 2025 Form 10-K 审计报告原文为准，此处不转述未逐页核对的措辞

Operating cash flow does not diverge from earnings in a way that needs a channel-stuffing story: Q2 OCF was $4.697B against $1.114B of net income. The divergence is free cash flow. Capex was $5.789B, +142%, and FCF was -$1.092B. Cash was $15.219B; cash plus short-term investments $43.524B, still +18% year on year but about $1.2B below the Q1 level of $44.743B.

companyfacts has no usable CY frame for long-term debt, so this note does not invent a debt balance. What is known: the Sept 29 facilities were undrawn, and the old $5B revolver was undrawn when it was cancelled. Liquidity covers near-term capex. If the Texas robot building and the fab hit an equipment peak together, the $30B is a buffer that becomes interest-bearing debt only when drawn. The liquidity covenant is $5B. Cash plus short-term investments is not close to that floor.

The audit-report wording belongs to the FY2025 Form 10-K. The 8-Ks and the Q2 update cited here do not contain a going-concern emphasis.

Operations

Q2 operating table from the shareholder update, unaudited:

Days of supply at 15, with deliveries above production (451,758), means Q2 worked down the Q1 inventory build. The new-vehicle operating-lease count fell from 167,163 in Q3 2025 to 141,876 in Q2 2026. The company’s footnote says that from Q4 2025 the count is new vehicles only and that Q3 2025 was adjusted to drop used vehicles, so the Q2 2025 figure of 172,882 is not a clean year-on-year base. The direction is still toward outright sales rather than leases. That is separate from FSD subscriptions rising: the metal is getting cleaner, the software is becoming an annuity.

The Q3 8-K is units only. Model 3/Y production 457,387 / deliveries 478,237; other models 7,004 / 8,295; total deliveries 486,532; about 1% of 3/Y and 4% of other models under operating leases. Deliveries again exceeded production by about 22,000. Storage 13.7 GWh. ASP, regulatory credits, and auto gross margin wait until Oct 21.

Geographically, the Q2 letter lists record deliveries in Korea, Australia, Japan, Taiwan, Thailand, and several European and Latin American markets, and it does not split China retail. Reuters on July 31, quoting Grace Tao, says Shanghai accounted for more than half of 2025 global deliveries, with 400-plus local suppliers, and that the plant is the export hub for Europe, Canada, and Asia-Pacific. This page does not adopt a first-half 2026 China sales-mix percentage that is not in the 10-Q. A retail-versus-export split needs CPCA or the next geographic note.

Q2 2026 mix

Q2 2026 mix

Segment	Revenue	Mix	Note
汽车	$20.516B	72.7%	同比 +23%。Q2 交付 480,126（+25%）
能源生成与储能	$3.139B	11.1%	同比 +13%。部署 13.5 GWh（+41%）
服务及其他	$4.581B	16.2%	同比 +50%。毛利 $648M，毛利率 14%
公司毛利	$4.751B	16.8%	环比从 Q1 的 21.1% 回落；经营利润率仅 1.4%

Competition

Tesla is no longer the only company that can build electric cars at scale, and it is not the leader in paid driverless rides. Both statements are why the share charts are null. There is no quarterly global share series here that can be checked back to a filing, and a smooth curve would be fiction.

Market share

Interactive chart available in the reader.

Peer share

Interactive chart available in the reader.

Peers

Peers

Name	Lane	Edge	Gap
特斯拉	电动车 + 储能 + 自研智驾	垂直整合、数据闭环、超充、美国监管试验场	中国价格带、欧洲本土品牌、Robotaxi 车队仍小于 Waymo
比亚迪	电动车成本与中国/出口量	电池垂直整合、全价位、插混缓冲	软件与美国市场准入
大众集团	欧洲规模燃油/电动切换	渠道、品牌矩阵、欧洲本地生产	软件组织与单车盈利
通用汽车	北美电动与 Cruise 遗产	本土政治、卡车利润池	智驾叙事弱于特斯拉
Waymo（Alphabet）	付费 Robotaxi 规模锚	已运营城市的无人里程与监管记录	不单独上市，硬件成本与扩城速度

Against BYD, the China price ladder, plug-in hybrids, and battery cost are Tesla’s home-market disadvantage. Turning Shanghai into an export hub protects utilization at overseas prices; it does not take the domestic share back. BYD’s gap is US market access and a fleet-data subscription loop, which is the part Tesla still holds.

Against Volkswagen and GM, EU countervailing duties are squeezing “made in China, Western brand” imports. Transport & Environment’s comparison of 2024 with Q1 2026 says Tesla’s share of China-built BEV imports into the EU fell from 26% to 19% as Berlin replaced Shanghai. That is trade policy, not a sudden product failure. GM’s North American truck profit is thicker and its autonomy story is thinner.

Against Waymo, the Robotaxi note still treats Waymo as the paid-scale anchor and Tesla as the high-beta option. Since Q2, Tesla has added unsupervised Florida service, and Cybercab has gone from employee rides to reported public rides in Austin. Tesla has not published a weekly paid-ride figure that can sit next to Waymo’s. Until it does, “vision is cheaper, therefore Tesla wins” is a cost hypothesis, not a share fact.

Humanoids stay out of the share table. Optimus competes with factory automation budgets and other unprofitable robot programs, not with BYD’s car sales. External sales timing, and lease-versus-sell, are still press relays.

In one line: autos are a war of price and utilization, software is a subscription base Tesla actually has, robotaxis are regulatory time Tesla is buying, and the robot is using Tesla’s own factory as customer number one. The equity multiple folds four clocks into one number.

Management

The three seats that have to be named did not change in the last 24 months.

Key seats

Key seats

Role	Name	Since	Last 24 months
董事会主席	Robyn Denholm	2018-11	席位未变。2025 10-K/A：董事任期至 2026
CEO	Elon Musk	2008	仍任 CEO。2025 CEO Performance Award 把交付与无人 Robotaxi 放进考核
CFO	Vaibhav Taneja	2023-08	未更换。2026-09-29 授信 8-K 仍由其签署
汽车制造（历史）	Tom Zhu 等	—	委托书曾列全球汽车制造角色；下层销售/项目负责人变动以媒体报道为主，不升格为三席变更

Denholm has been chair since November 2018 and is still chair in the FY2025 10-K/A, with a term expiring in 2026. Musk remains CEO and a director. Taneja has been CFO since August 2023, after serving as chief accounting officer, and he signed the Sept 29 credit-facility 8-K. The governance and finance seats are stable.

What is not stable is the CEO’s time and the layer under those seats. The 2025 CEO Performance Award, referenced in the shareholder-update footnotes, counts new-vehicle deliveries and unsupervised robotaxis placed into commercial operation. That incentive pays for opening cities quickly; unit economics can lag. Electrek reported in March 2026 that VP of Finance Sendil Palani left, and it recited earlier departures across powertrain, software, and sales in 2024–2025. Those are not 8-K changes to the three seats. They do mean execution continuity is weaker than the continuity of chair, CEO, and CFO. Call: governance seats are steady; the bench is in transition. Strategy will not pivot because the CFO chair is empty. It can slip because of CEO attention and program-lead gaps.

Outlook

For the next two quarters the base case is not “robots rewrite the P&L.” It is auto revenue staying near $20–21B a quarter, group gross margin oscillating between the mid-teens and low-20s, opex still elevated, and free cash flow decided by capex timing. Q3 deliveries were roughly flat with Q2 and slightly down year on year. If ASP holds, revenue should stay high. If China and Europe price the car down, revenue can fall while units do not. Storage has now printed 13.5 and 13.7 GWh without breaking the 14.2 record. Services should slow as the subscription base grows, and the sign should stay positive.

Over 12–24 months the company’s own roadmap is to fill existing vehicle plants before announcing another one; to start Megapack 3 in Texas; to move Nevada Semi from “deliveries have started” to a number; to use Fremont Optimus output for training, with a higher-volume Texas robot building discussed around summer 2027; and to keep the Austin fab early. Musk’s public line on the robot ramp is that it will be slow. The next Robotaxi test is not another city announcement. It is paid miles, the safety-driver ratio, and whether a vehicle without a steering wheel can be sold or operated under federal rules. Whether NHTSA has issued a formal order is a question for the agency’s docket. This note does not promote a fine quoted in secondary coverage into an assessed penalty.

Geopolitics is an operating assumption. Shanghai is both the lowest-cost export base and the hardest supply chain to replace. If the reported China separation becomes an 8-K, the market will have to price that asset on its own, and US plants will pay more for cells and parts. EU duties have already changed the Shanghai-to-Europe mix; Berlin has to absorb that volume or the export hub simply relocates the China price war. US tariffs and subsidies push the supply chain home, which is the same project as the 4680, lithium, and LFP ramps: cost is higher until they work, and Semi and Cybercab are short of cells if they do not.

Scenarios

No price target. Three evidence sets.

Base. Quarterly deliveries stay in a 450–500k band. Gross margin stays between the mid-teens and low-20s. Operating margin stays in the low single digits. Free cash flow flips sign with Semi, Optimus, and AI capex. Robotaxi adds cities without paid miles large enough to be their own line. Semi has customers and an annual rate far below the 50k nameplate. Optimus stays an internal training fleet. The multiple stays high and twitchy. That is the world the current price partially assumes and has not falsified.

Upside. All of the following, not one of them: company-reported paid unsupervised miles that keep rising; a clear Cybercab regulatory path and a falling safety-driver ratio; services gross profit still making records, so group operating margin gets back above the mid-2025 area around 5%; storage deployments that hold above 14 GWh with Texas Megapack actually in production. A disclosed positive Semi margin would be extra. One of these items does not turn 340x into a fundamental.

Downside. Auto gross margin breaks below Q2 without an opex decline; a serious crash or a permit freeze stops Robotaxi; the China business is separated at a bad price, or destination tariffs stall Shanghai and utilization falls; the new facilities are actually drawn and free cash flow stays negative. Downside does not require Optimus to fail. The auto margin is already thin.

Risks

Risks

Risks

Risk	Level	Note
估值与利润脱节	高	约 340x TTM GAAP 净利。Q3 利润或 Robotaxi 节奏不及预期时，倍数压缩不必等基本面崩塌
智驾监管与事故	高	无方向盘 Cybercab、各州许可、安全员政策。一次重大事故可以冻结扩城
中国与供应链脱钩	高	上海贡献过半产量的历史；媒体报道公司评估剥离中国业务。美国工厂 2027 年前退出中国供应商是公司既定方向
费用与资本开支	中	Q2 营业费用 +47%，FCF 为负。Optimus/芯片/Semi 若只烧钱不贡献收入，经营利润率会停在低个位数
中国需求与欧洲关税	中	国内零售承压、出口依赖目的地政策。欧盟反补贴改变中国产电动车的入境结构
关键人与执行层	中	CEO 同时运营 SpaceX 等主体。CFO 稳定，但媒体报道 2024–2026 年多层业务负责人离职
Optimus / Semi 延期	中	两项目都有多年跳票记录。产能口号不等于交付

Tracking

1. Q3 financials on 2026-10-21. Does auto revenue follow 486,532 deliveries, which way does gross margin move versus 16.8%, what happens to opex growth, and is free cash flow still negative? This falsifies “flat deliveries, stable profit.”
2. Company-reported Robotaxi KPIs. Paid miles, the unsupervised city list, safety-driver policy, and whether Cybercab is in the paid fleet. Another city name without miles is narrative continuity. Compare with Waymo in the industry note.
3. Semi deliveries versus the 50k nameplate. A customer total or a quarterly build rate is the evidence. A start-of-production post is not.
4. Optimus units in a shareholder update or 10-Q. Press reports of several hundred a week are a lead, not a KPI, until Tesla confirms them. External sales before 2H 2027 would be upside. No useful internal units by year-end 2026 would be downside.
5. China and the credit line. An 8-K on separation, China’s share of revenue, and whether the $30B is still undrawn. A draw, or liquidity moving toward the $5B covenant, is the financial-health switch.

References

• Q2 2026 update (revenue, margins, cash flow, deliveries, capacity, Robotaxi cities, FSD subscriptions): https://assets-ir.tesla.com/tesla-contents/IR/TSLA-Q2-2026-Update.pdf
• Same exhibit on EDGAR: https://www.sec.gov/Archives/edgar/data/1318605/000162828026049213/exhibit991.htm
• Q2 production and deliveries (480,126 / 13.5 GWh): https://www.sec.gov/Archives/edgar/data/1318605/000162828026046717/exhibit99111111.htm
• Q3 production and deliveries (486,532 / 13.7 GWh; financials 2026-10-21): https://www.sec.gov/Archives/edgar/data/1318605/000162828026064366/exhibit991111111.htm
• Sept 29, 2026 credit facilities 8-K: https://www.sec.gov/Archives/edgar/data/1318605/000162828026063820/tsla-20260929.htm
• SEC companyfacts, CIK 0001318605: https://data.sec.gov/api/xbrl/companyfacts/CIK0001318605.json
• FY2025 10-K/A, directors and CFO: https://www.sec.gov/Archives/edgar/data/1318605/000110465926053166/tm2611837d1_10ka.htm
• Robotaxi page (six cities as checked 2026-10-03): https://www.tesla.com/robotaxi
• Prices: Sina US daily closes on or before quarter-end; the 2026-10-02 close matches Tencent gtimg
• Market cap and vendor PE/PB: Tencent gtimg usTSLA, 2026-10-02 16:00:01
• China operations and a reported possible separation (WSJ via Reuters, not company-confirmed): https://www.reuters.com/business/autos-transportation/teslas-china-operations-ev-makers-global-production-powerhouse-2026-07-31/
• Semi high-volume claim as reported (capacity is not output): https://www.supplychain247.com/article/tesla-semi-high-volume-production-nevada-factory
• Optimus ramp as press relay, not a company KPI: https://electrek.co/2026/09/25/tesla-optimus-production-ramp-hands-ai-generalization-problems/
• EU tariffs and the mix of China-built BEV imports (T&E): https://www.transportenvironment.org/articles/presence_chinese_automakers_eu_car_market
• Press report on the VP of Finance departure (not an 8-K): https://electrek.co/2026/03/10/tesla-vp-finance-sendil-palani-leaves-17-years/
• Related notes: Robotaxi, BYD