Author | Janson

Editor | Zhi Hao

Half-year revenue of 342.7 billion yuan yet revenue grew without profit growth, Musk bets on Robotaxi to break the deadlock.

Carwest news on July 23, just now, Tesla released its Q2 2026 financial report, and simultaneously held an earnings call to address investors' concerns.

In the second quarter, Tesla's automotive business recovered significantly, delivering 480,100 vehicles, a 25% year-on-year increase, setting a new historical high for the same period.

▲ Overview of Tesla's financial information for the second quarter of the past three years

Tesla's Q2 revenue reached $28.236 billion (approximately 191.2 billion yuan), up 26% year-on-year; Automotive business revenue reached $20.516 billion (approximately 138.9 billion yuan), up 23% year-on-year.

But sales and revenue growth did not bring proportional profit improvement. In the second quarter, Tesla's operating profit was only $398 million (approximately 2.7 billion yuan), a 57% year-on-year decrease; the operating profit margin dropped from 4.1% in the same period last year to 1.4%. Automotive gross margin also dropped from 21.1% in the first quarter of this year to 16.9%.

Looking at the entire first half of the year, Tesla's cumulative revenue reached $50.623 billion (approximately 342.7 billion yuan), of which Automotive business revenue was $36.750 billion (approximately 248.8 billion yuan), still contributing over 70% of revenue.

▲ Overview of Tesla's financial information for the first half of the past three years

On the profit front, Tesla's first-half gross profit totaled $9.471 billion (approximately 64.1 billion yuan), with an overall gross margin of about 18.7%. Regarding deliveries, Tesla produced approximately 860,100 vehicles and delivered approximately 838,100 vehicles in the first half of the year.

Overall, the delivery rebound in Q2 pushed Tesla's first-half revenue to a new high for the same period in the past three years, but the operating profit margin still dropped to 2.65%, and the problem of revenue growing without profit did not get resolved.

Compared to profit performance, Tesla's progress in FSD (Supervised), Robotaxi, and Cybercab was more prominent.

As of the end of the second quarter, Tesla FSD paid user numbers reached 1.48 million, a 56% increase from 950,000 in the same period last year, and up 200,000 from the end of the first quarter this year. Tesla CEO Musk said that as FSD receives regulatory approval in various countries, market demand will further surge.

In addition, Tesla has started pushing FSD v14 lite to vehicles equipped with AI3 (HW3) hardware, distilling the driving behavior of the v14 series on AI4 vehicles to the previous generation hardware platform.

Regarding Robotaxi, Tesla's service scope has expanded to multiple cities in the US. Meanwhile, Cybercab has started production at the Texas Gigafactory and has entered the public road engineering testing phase.

While the profit pressure on Tesla's Model 3/Model Y vehicle sales business continues, FSD subscriptions and Robotaxi are gradually becoming new growth sources outside of Tesla's automotive business.

I. Revenue Grows But Profits Don't, Automotive Gross Margin Declines Quarter-on-Quarter

Vehicle delivery volume recovered, driving Tesla's Q2 revenue to recover growth.

In the second quarter of this year, Tesla's total revenue reached $28.236 billion (approximately 191.16 billion yuan), up 26% year-on-year; Automotive business revenue reached $20.516 billion (approximately 138.89 billion yuan), up 23% year-on-year.

Specifically, Automotive sales revenue was $20.006 billion (approximately 135.4 billion yuan), Car rental revenue was $364 million (approximately 2.5 billion yuan), and Automotive regulatory credit revenue was $146 million (approximately 1 billion yuan).

▲ Detailed quarterly financial information of Tesla

Tesla stated that Q2 revenue growth was mainly driven by increased vehicle deliveries, FSD subscription growth, expansion of services and other businesses, growth in energy business, and a positive exchange rate impact of about $500 million (approximately 3.4 billion yuan).

However, the decline in average selling price of vehicles offset some of the growth.

In the second quarter, Tesla's gross profit was $4.751 billion (approximately 32.2 billion yuan), up 23% year-on-year; the overall gross margin was 16.8%, down 0.4 percentage points from 17.2% in the same period last year, and down 4.3 percentage points from 21.1% in the first quarter this year.

But the decline in average selling price of vehicles, reduction in regulatory credit revenue, and increased investment in AI projects dragged down overall profit performance.

In the second quarter of this year, Tesla's operating profit was $398 million (approximately 2.7 billion yuan), a 57% year-on-year decrease; the operating profit margin dropped from 4.1% in the same period last year to 1.4%, a year-on-year decline of 2.69 percentage points.

However, Tesla's GAAP net profit was also supported by an investment income. In the second quarter, Tesla recognized $1.005 billion (approximately 6.8 billion yuan) in unrealized gains from its SpaceX equity investment, which was included in other income.

Regarding R&D expenses, Tesla's R&D expenses in the second quarter reached $2.371 billion (approximately 16.1 billion yuan), an increase of about 49% year-on-year.

In terms of automotive manufacturing, Tesla stated that battery pack capacity remains the main factor limiting global automotive production growth.

Tesla is increasing 4680 cell production to support Cybercab, Tesla Semi capacity ramp-up, and Model Y production expansion.

▲ Tesla's battery capacity plans for each factory

Currently, Tesla's Texas 4680 cells have installed annual capacity exceeding 40GWh, the Nevada LFP (Lithium Iron Phosphate) battery project is in the early capacity ramp-up phase, with expected annual capacity of 7GWh, while the Berlin 4680 cell project is still under construction.

In addition, the Tesla Semi new factory is in the equipment debugging phase, planning to start production in 2026.

II. Q2 Deliveries Hit Record High, Cybercab Production Starts, FSD Users Reach 1.48 Million

In the second quarter of this year, Tesla produced 451,800 vehicles, up 10% year-on-year; delivered 480,100 vehicles, up 25% year-on-year, setting a new historical high for the same period.

In the first half of 2026, Tesla cumulatively produced approximately 860,100 vehicles and delivered approximately 838,100 vehicles.

▲ Tesla's quarterly delivery situation

Among them, Model 3/Y remain the core sales driver, producing 442,900 vehicles and delivering 467,800 vehicles in the second quarter, up 12% and 25% respectively year-on-year, contributing approximately 97.4% of the delivery volume.

Other models including Cybertruck produced 8,822 units, down 34% year-on-year; delivered 12,400 units, up 19% year-on-year.

With deliveries exceeding production, Tesla's inventory pressure significantly eased, global vehicle inventory turnover days dropped from 27 days in the first quarter to 15 days, lower than 24 days in the same period last year.

Tesla set delivery records in multiple markets including South Korea, Australia, Japan, Taiwan, Thailand, etc.

Regarding capacity, Shanghai Factory Model 3/Y annual capacity exceeds 950,000 vehicles, still Tesla's largest vehicle production factory; California, Berlin, and Texas factory related model annual capacities exceed 550,000, 375,000, and 250,000 respectively.

Additionally, Cybercab designed specifically for Robotaxi has started production at the Texas Gigafactory, with installed annual capacity exceeding 125,000 vehicles.

▲ Tesla's installed annual capacity and production status for each factory

The mass-produced Cybercab began public road engineering testing in the second quarter, and in July provided ride services for employees at the Texas factory park.

Robotaxi is also accelerating expansion, currently entering 7 major metro areas in the US. Austin continues to expand areas with unsupervised operations, Dallas and Houston are promoting relevant operations, Miami, Orlando, and Tampa launched unsupervised services in July; San Francisco Bay Area still equips safety drivers. Phoenix and Las Vegas are in the pre-launch preparation phase.

▲ Mass-produced Cybercab in testing

FSD has become another growth point. As of the end of the second quarter, Tesla FSD paid users reached 1.48 million, up 56% year-on-year, up 200,000 from the first quarter; increased by approximately 380,000 in the first half of 2026 alone.

In the second quarter, Tesla FSD net new subscription users set a record, over 55% of new vehicles delivered in North America included FSD subscription, and the order rate also reached a historical high. FSD subscription growth has formed a positive contribution to automotive related revenue and operating profit.

Regarding software, Tesla began pushing FSD v14 lite to vehicles equipped with AI3 (previously referred to by Tesla as HW3) hardware in the US, and pushed it to South Korea in July.

This version distills driving behavior of the v14 series on the AI4 platform to the AI3 platform, adds destination options such as parking lots, roadside, lane entrances, and curbs, and improves performance in scenarios such as navigation, merging/diverging, pedestrian interaction, traffic lights, and vehicle cutting in.

FSD's overseas approvals also made new progress, after the Netherlands, Tesla obtained FSD deployment approval in Lithuania, Estonia, Denmark, and Belgium.

As of July, users in the above opened markets cumulatively used FSD to drive over 50 million kilometers. Tesla stated that FSD implementation is driving local consumers' interest in its vehicles.

III. Cybercab Has No Legal Barriers, Old Owners Need Hardware Upgrade Assessment

During the earnings call, Musk and Tesla management focused on responding to Robotaxi operations, Cybercab regulations, Starlink access, Semi autonomous driving capabilities, and AI3 model hardware upgrades. Carwest organized the following 8 key Q&As:

1. How is the current operation of Robotaxi? How to scale up next?

Tesla AI Head Ashok Elluswamy: Currently, Tesla Robotaxi has cumulatively completed over 380,000 miles (approximately 610,000 km) of unsupervised driving in two US states and six cities, with no safety accidents worth noting. Existing accident reports were all collisions or impacts by other road users when the vehicle was stationary.

Tesla started the Robotaxi project in Austin about a year ago, at that time the passenger seat was still equipped with a safety monitor; at the end of 2025, the first batch of fully unsupervised Robotaxi started operations. Since 2026, the weekly unsupervised driving mileage of the fleet has maintained double-digit quarter-over-quarter growth, expected to continue this pace within the year.

Currently, the Robotaxi fleet has been running early versions of FSD V15. V15 planned about seven major improvement routes, with about 40% of improvement content merged into the version used by the fleet. With subsequent capabilities going live, Tesla hopes the engineering investment and preparation time required to enter new cities will gradually approach zero, finally shifting from city-by-city expansion to state-wide operations.

2. How does Tesla want the US to regulate Robotaxi?

Tesla Automotive Engineering Head Lars Moravy: The US Federal level has made不少 progress, especially Federal Motor Vehicle Safety Standards (FMVSS) are gradually accepting vehicles designed specifically for autonomous driving.

Tesla believes reasonable regulation should be for regulatory agencies to propose safety goals and problems to be solved, then allow enterprises to choose technical solutions, rather than specifying beforehand that a certain technical route such as LiDAR or millimeter wave radar must be adopted.

Tesla ultimately still needs to win regulatory agency and public recognition with actual vehicle safety performance, which will also become the basis for Robotaxi to continue expanding.

3. What factors limit Robotaxi expansion? Will it integrate with third-party ride-hailing platforms?

Musk: Tesla expects Robotaxi will not face insufficient demand problems, its operational economics will be very attractive, and market demand may exceed Tesla's service capability for a long time. Therefore, Robotaxi business will continue to adopt a highly vertically integrated model, at this stage there is no need to rely on third-party ride-hailing platforms for customer flow.

What truly limits Robotaxi expansion is system reliability, which is increasing the "9" in reliability indicators. Musk stated that ideally, Robotaxi needs to achieve reliability close to 99.999999%. Continuously improving reliability is the main constraint for current fleet scaling.

4. Could Tesla and SpaceX merge?

Musk: Tesla and SpaceX have increasing business overlap and cooperation in many fields, but company merger is not suitable for discussion on earnings call, if future matters involve similar items, formal decision and approval process must be undergone.

Tesla General Counsel Brandon Ehrhart: In early 2026, Tesla deepened its relationship with SpaceX through equity investment and framework agreement, both parties are cooperating on projects such as Digital Optimus.

5. Why not expand the Austin fleet first?

Ashok Elluswamy: Tesla did not concentrate all vehicles in Austin, mainly to verify the generalization capability of the FSD tech stack, proving the system can adapt to multiple cities with different road environments without massive additional development.

Robotaxi is still in the early stage of exponential growth, the absolute number of vehicles seen by outsiders is not large, but the fleet operates basically continuously, single vehicle driving time is far higher than private cars. Therefore, Tesla focuses more on unsupervised driving mileage, rather than simply focusing on vehicle quantity.

Musk: Cybercab is a brand new model, not having millions of vehicles providing road data like Model 3 and Model Y. Tesla needs to use temporarily added steering wheel, acceleration and braking pedals Cybercab to accumulate mileage, complete calibration for chassis and vehicle dynamics. After relevant data meets requirements, Cybercab quantity in each city will increase significantly.

Additionally, ride-hailing and traffic regulatory requirements vary by city and state in the US. City-by-city expansion also helps Tesla solve software, operations, and local regulatory issues one by one, then expand single market fleet size.

6. What federal regulatory barriers remain for removing steering wheels and pedals from Cybercab?

Lars Moravy: Besides regulatory agencies promoting relevant rule adjustments, Cybercab currently has no other major federal regulatory barriers.

In the past few years, Tesla has explained Cybercab's product planning and technical progress to the National Highway Traffic Safety Administration and other departments. Although both sides cannot be completely consistent on all issues, communication is smooth, regulatory agencies are pushing to cancel traditional requirements such as vehicles must be equipped with steering wheels and pedals according to autonomous driving technology development.

7. Can Cybercab become a Starlink mobile hotspot? When will Tesla Semi add autonomous driving capabilities?

Musk: Starlink terminal mounted on Cybercab may become ground network relay in the future, providing Wi-Fi service for nearby mobile phones or other devices, but relevant solution has not been officially announced.

Regarding Semi, autonomous driving capabilities can alleviate US truck driver shortage, and improve safety and comfort of long-haul transport. However, Semi currently occupies a low proportion in Tesla fleet, Tesla will prioritize investment of R&D resources in larger models such as Model 3, Model Y, and Cybercab.

It is expected that Semi's autonomous driving capabilities will make substantive progress at the end of 2026 or early 2027, and will be put into use in 2027, catching up with the timing of Semi entering large-scale mass production.

8. Will old cars equipped with AI3 hardware be upgraded to new hardware?

Musk: In the long term, upgrading hardware for vehicles equipped with AI3 or earlier computing platforms, and already adopting camera-based solution, may be economically feasible. But if vehicles need massive structural modifications, upgrading may not be economically viable.

Tesla may not simply upgrade AI3 vehicles to existing AI4, but wait for the next generation computing platform. The company is developing improved AI4, performance has a moderate increase compared to existing AI4, expected to launch production around mid-2027.

AI5 is also expected to enter mass production around mid-2027, and will be applied to Optimus first. Musk also revealed that AI6 chip has entered the design phase, with the goal of becoming a globally leading edge computing chip.

Conclusion: Tesla Aggressively Pursues Growth

Overall, Tesla's first-half automotive sales walked out of the valley, but low pricing and increased R&D investment continue to squeeze profits, delivery growth has not yet translated into stronger profitability.

At the same time, FSD paid users growing rapidly, Cybercab put into production, Robotaxi accelerating city expansion, showing Tesla is shifting growth focus from simply selling cars to software subscriptions and mobility services.

Next, determining the success or failure of this transformation is not only whether FSD capabilities can continue to improve, but more importantly whether Robotaxi can cross the reliability and regulatory thresholds to achieve true large-scale commercial operations.

July 16, inside the GAC Trumpchi factory in Panyu, Guangzhou, a right-hand drive Trumpchi M8 PHEV slowly drove off the production line. This car had no exaggerated livery, nor the complex ceremony of a red carpet, but its appearance brought a certain weight to the air of the entire scene. This is the 30 millionth vehicle produced by GAC Group.

When GAC Group Chairman Feng Xingya handed the keys to Thai owner Tony Jaa, the moment captured in the camera was not merely a warm interaction between GAC representatives and users. In 2026, a year when industry price wars raged fiercely and competition evolved from 'fighting over prices' to 'fighting over leadership', GAC offered a different narrative through its 29-year milestone.

At a time when many shout "long-termism" while engaging in short-term games, GAC's 30 million might be the most substantial interpretation of these four characters.

30 Million Units Produced, GAC Turns the Home Field into a "Global Showcase"

This production line ceremony was not turned into a dull marathon of leadership speeches, but instead had a hard-core tech-style romance. The most intuitive feeling on site was: GAC is no longer just Guangzhou's GAC.

At the event site, GAC connected multiple global production bases simultaneously. As vehicles numbered from 29,999,995 to 29,999,999 appeared sequentially in a "cloud reveal", the GAC Honda P7, GAC Toyota bZ7, Qijing GT7, Aion N60, and Hyper S600 all took the stage. Finally, the right-hand drive Trumpchi M8 PHEV in the center of the stage made its grand finale appearance as the 30 millionth vehicle.

Choosing the right-hand drive M8 PHEV as a milestone model is itself a strong signal that GAC's ambition long ago was no longer limited to the domestic market. In the first half of this year, GAC's independent brand export volume reached 121,500 vehicles, up 132% year-on-year, almost matching last year's full-year level. Thailand's June sales surged 207% month-on-month, and the European market is also building a full-chain operational system from production to sales.

30 million users are not just the result of Chinese car owners voting with their feet, but also the recognition of the "Made in China Intelligence" by markets in 110 countries and regions worldwide.

Building cars is not a 100-meter sprint, but a never-ending marathon

Of course, marketing can tell stories, but a volume of 30 million cannot be sustained by marketing alone. In GAC's narrative, the most critical underlying logic supporting this number is quality.

At the event site, a sentence from Feng Xingya left a deep impression: "Quality is the bottom line GAC will never yield." When spoken by GAC, this sentence is indeed persuasive. In the wave of the industry pursuing "cost reduction and efficiency improvement", many car companies do subtraction where it cannot be seen to compress costs, but GAC continues to do addition in quality verification.

Before each new car launches, GAC requires at least two winters and one summer of on-site road testing; the regular steady cycle lasts up to 21 months; from Turpan's 60°C high temperature to Hailar's -40°C extreme cold, from high humidity, high salt fog to mountains and sandstorms, the "Five Highs, One Mountain, One Dust" full environment assessment covers all extreme scenarios users might encounter.

This "hard work" is also reflected in data: The Magazine Battery has been installed in a cumulative 1.5 million vehicles, with safe driving mileage exceeding 160 billion kilometers; the Star Spirit Safety Protection System has served nearly 2 million users, cumulatively preventing potential accidents 6.28 million times. By the end of 2025, GAC also took the lead in the industry to launch the independent brand "Three Responsibilities" policy, actively guaranteeing the Three-Electric systems, battery degradation, and intelligent parking accidents.

Behind these data lies a simple truth: Quality is not spoken, it is verified.

Turning "User-Centric" from a Slogan into a Closed Loop

At present, "user-centric" has almost become the standard slogan for every car company, but few can truly turn user feedback into a closed loop for product improvement.

GAC's approach includes some details worth noting. The "All Microphones On" activity allows R&D heads to face users directly, converting real car usage pain points into product optimization priorities; the establishment of an exclusive user insight department guarantees user requests are "responded to in every case" through organizational mechanisms.

Hard investment in the service system is also progressing synchronously: By the end of June, 600 county-level outlets have been certified, planned to increase to 1,000 within the year; self-operated charging piles exceed 27,000, core urban areas achieve "a station within 1 kilometer in a straight line"; Online "Super Butler" promises 5-second response, 2-hour problem solving.

These actions cannot be called disruptive innovations, but they point to a more pragmatic logic: Users do not need flashy concepts, but rather someone to repair broken cars, no long queues for charging, and someone to manage problems.

Panyu Action One Year, Reform Is Not a Banquet

Last November, GAC launched the "Panyu Action", moving the headquarters to Panyu, and starting independent brand integrated operations. At that time, there was waiting and also questioning outside. The organizational reform of traditional car companies has always been known to be easy to say but difficult to do.

After more than a year of promotion, results are being released. In the first half of the year, GAC Group cumulative sales were 773,100 vehicles, up 2.35% year-on-year; independent brand cumulative sales were 346,000 vehicles, up 35.69% year-on-year. More importantly, the proportion of energy-saving and new energy vehicles has reached 62.82%. GAC's electrification transformation is no longer a slogan, but solid sales data.

From starting amidst controversy in 1997, to the birth of Trumpchi in 2010 realizing the independent dream, to now Qijing GT7 carrying Huawei Qiankun Intelligent Driving ADS 5 to rush the high-end market, GAC has experienced four leapfrog developments. The trajectory of these 29 years illustrates a truth: In the automotive industry, there are no shortcuts. All "speed" is backed by "slow" accumulation.

30 Million Is a Milestone, Not a Finish Line

Speaking of this gratitude event itself, from July 16 to August 31, GAC's six major brands launched multiple favors around new purchase, trade-in, and additional purchase, covering fuel, hybrid, pure electric, and extended-range power types. This is a return on the support of 30 million users, and also an active layout for the additional and replacement purchase market in the era of stock competition.

But looking from a longer cycle, the significance of 30 million is not just this. In the transformation period of the Chinese automotive industry from large to strong, GAC's path provides a reference: Not harvesting the market through short-term price wars, but accumulating reputation through quality; Not relying on a single technical route, but planning new energy power across the board; Not only staring at domestic share, but actively going overseas to participate in global competition.

This might not be the most clever approach, but it might be the way to go further. 30 million users are not the end of GAC's story, but the starting point of the next 29 years. For GAC, which is currently leaping from "Manufacturing" to "Intelligent Manufacturing", from "Local" to "Global", and from "Car Enterprise" to "Technology Ecosystem Enterprise", the real test has just begun.

When many brands were still relying on price cuts to drive sales, Geely Automobile (including the three brands Geely, Lynk & Co, Zeekr) instead relied on consecutive 3 months of Year-over-Year and Month-over-Month dual growth (237,000 vehicles in May), quietly walking out another path. Behind this is not luck, but more like a designed "Methodology to Resist Involution".

For any manufacturer, product is always first. The reason Geely can achieve today's results is because they almost avoided all the price war disaster zones. In the interval called "Blood Sea Market" of 100,000 to 150,000 RMB, Geely did not simply pile up features for price reductions, but used Galaxy M7, Star Shine 7, and Starship 7 to form a "Technical Combo Punch" — for example, the Star Shine 7 MAX series comes standard with four-wheel drive, turning configurations that usually cost extra in the same class into basic models, using "Value Equity" to replace "Price Involution", which allowed the new car to break 5,000 units in the first month of launch.

Going deeper, its growth quality is hidden in the structure. The new energy penetration rate is 56%, meaning for every two cars sold, one is a new energy vehicle. However, looking closely at the composition, in Zeekr's 34,000 deliveries, 9 Series, 8 Series, and other high-ticket models accounted for nearly 50%. The average price per vehicle rose 52% year-over-year. This "rising volume and price" is more persuasive than simple sales figures. This shows its new energy growth is not relying on low-price electric vehicles to boost volume, but truly standing firm in the high-end market. Zeekr 009 continuously winning the championship in the MPV market above 400,000 RMB is the proof.

More importantly, the technical layout balances both short and long terms. In the short term, it relies on i-HEV hybrid technology to extend the life of the China Star series, allowing fuel vehicle users to complete upgrades without changing usage habits, which allowed the China Star to break 100,000 units in a single month. In the long term, it bets on 900V high-voltage architecture (Lynk & Co 10 Series), Qianli Haohan Smart Solutions, positioning itself early in the 200,000 to 250,000 RMB market. Even the acquisition of Radar Automobile seems calculated — avoiding price involution in passenger cars, targeting the pickup truck segment where users have low price sensitivity and high scenario premium. Now Radar is not only the domestic sales champion but has also captured the Thailand market, becoming a surprise weapon for profit optimization.

Geely's sales numbers prove one thing: When everyone is fighting with bayonets in the existing market, the true growth potential may be hidden in the analysis of user needs — not that users only buy cheap ones, but that users do not want to pay for "unnecessary premiums".