搜尋歷史
熱門搜尋
經銷商登入 / 註冊
Autonomous driving that has been burning cash for over a decade is showing an uncommon signal.
A company that still allocates over 70% of its revenue to R&D is approaching operational break-even.
Momenta's first semi-annual report after listing shows that in the first half of this year, the company's revenue was 1.601 billion yuan, a year-on-year increase of 76%; the adjusted net loss was only 14.097 million yuan, narrowing by 96.6% year-on-year.
More notably, this was not achieved by cutting R&D.
Momenta's R&D investment reached 1.163 billion yuan, a year-on-year increase of 18.7%, with revenue growth being four times the R&D growth rate.
Profitability for Robotaxi still needs to be waited for in years, but Momenta seems to be sprinting. At the performance meeting, Momenta Executive Director and Chief of Staff An Ren broke the **myth that 'Autonomous driving is naturally a continuous money burner'**.
If continuous burning of cash is not the fate of autonomous driving, then what is this new path Momenta found?
01. Autonomous Driving Begins to Become a Scalable Business for the First Time
When everyone is pursuing full autonomy in autonomous driving, Momenta is actually solving another more fundamental problem: how can autonomous driving first become a scalable software business.
In 2016, L4 autonomous driving was in the peak of its success, Robotaxi was the hottest story in the capital market, but Momenta simultaneously chose the less sexy L2 mass production route.
And this less sexy path of L2 mass production laid the foundation for three key chips: **an entry point into more vehicle models, real-world data, and engineering skills polished through continuous adaptation**.
Autonomous driving is not pure software. It must enter cars, experience different vehicle models, sensors, computing platforms, and body control, and also face regulations, road conditions, and safety verification standards in different markets.
Every layer of adaptation involves real engineering investment.
The true difficulty lies here: the scaling of autonomous driving requires 'repeated adaptation'. How to turn a technology that relies heavily on engineering adaptation into a product that can be continuously replicated.
At the end of the competition in autonomous driving, it is no longer about the intensity of algorithms, but about **who can replicate the technology to enough cars**.
And Momenta recognized early on that **both mass production auxiliary driving and unmanned driving walk on two legs**:
Engineering standardization allows technical capabilities to be replicated across different vehicle models.
From securing Mercedes-Benz investment in 2017 to the mass production of the first project by the end of 2025, the mass production scale and engineering system buried for ten years begin to change the growth formula of autonomous driving today.
Resources needed to deliver one vehicle project were compressed from 400 people working for two years in the early stage to dozens of people in three months.
Building a set of replicable engineering capabilities lays the foundation for a scaled business.
The strategy of 'Blanket Coverage' is not a tactic of 'signing more clients', revealing Momenta's operational logic: not betting on a few blockbusters, but covering enough vehicle models.
The underlying logic is the law of large numbers: fluctuations in the sales of a single vehicle model will be diluted by the certainty of the overall scale.
OEMs themselves cannot accurately predict the sales of a single vehicle model. Momenta does not gamble on luck; it spreads the fluctuations by coverage.
In the first half of 2026, Momenta added 49 new designations, meaning an average of more than 8 models per month, cumulative 219. New installations reached 321,000 units, cumulatively breaking through 1 million units.
The first 100,000 units took 24 months, while the latest 100,000 units delivery took only 36 days.
'Blanket Coverage' truly hedges not only the uncertainty of sales of a single vehicle model, but also the uncertainty of autonomous driving commercialization.
For traditional Tier 1 suppliers, selling one more component is one more unit of income.
But for autonomous driving, mass production vehicles are not only a source of income, but also the largest real-world data entry point. Mass production scale brings the breadth, density, and continuity of data to another level.
With enough models, single-vehicle fluctuations are flattened by scale. And the larger the scale, the better the data, income, and R&D efficiency will improve simultaneously.
The essence of 'Blanket Coverage' is **actively turning business scale into technical scale**.
The more models are deployed, the more real data there is, the faster model iteration, and the thicker the technical barrier. Momenta's mass production business has rolled into a self-accelerating flywheel.
Scale is the entry ticket to the autonomous driving table. Standing firm in the scale competition, there is a chance to take the initiative in the 'data-model-commerce' closed loop of autonomous driving.
02. True Profit Comes from 'Replication', Not 'Selling More'
Once scale is built up, profits do not automatically follow.
In the 2026 Chinese automotive market, the price war has reached a white-hot state. From complete vehicles to components, from hardware to software, almost all links are being rolled into the vortex of price cuts.
'Blanket Coverage' sounds like a price war approach, but on the contrary, Momenta did not fall into this pit, but rather increased its own technical value.
Entering a new project, re-adapting different sensors, chips, and body control, was once a headache for intelligent driving suppliers. If adaptation costs cannot be reduced, income cannot roll up.
What truly determines whether an intelligent driving supplier can make money is not how many sets are sold, but how much cost is needed to add a set of new systems. Behind this is the fundamental difference between two business models:
The true threshold of scaling lies in engineering standardization, turning 'start every time' into 'accumulate once'.
After technology becomes platformized, the marginal cost brought by each additional customer continues to decrease, and profits will naturally emerge.
Momenta's revenue is divided into two layers:
Digging wells is getting faster, drinking water is getting more, the structure naturally improves.
In the past three years, the proportion of licensing income has continued to increase, pushing gross margins upward.
After vehicle models are mass-produced, for every car sold, Momenta charges a one-time software licensing fee without needing to invest a large amount of engineering manpower.
Compared to the same period in 2025, although the proportion of software licensing income decreased from 39.8% to 37.9%, the income itself still grew by 67.5%.
The decline in proportion is not a stall in the licensing business, but technical development services grew faster, increasing the proportion by 1.9 percentage points. This means **Momenta's expansion efficiency is improving, locking more vehicle model entry points with lighter investment**.
Supporting efficiency improvement is the R7 World Model. Compared to the previous generation R6, R7 achieved performance improvements of 3 to 5 times in multiple scenarios.
More noteworthy is: it does not require customers to switch hardware. Sensor and chip configurations continue R6, upgrades mainly rely on software and algorithms.
SAIC Volkswagen Flagship SUV ID. ERA 9X, Global Premiere Equipped with Momenta R7 World Model
For an intelligent driving supplier, the most important significance of software upgrades is on the one hand **improving user experience**. Customers who have already used R6 do not need to spend extra hardware money, one OTA exchange can replace several times the experience upgrade.
On the other hand, it is **allowing past R&D investment to continuously generate income and value**. R7 allows hardware to continue releasing value through OTA, and R&D investment is called repeatedly.
R&D results continue to release, users continue to stay on the platform, **the lifetime value of a single set of technology is thus improved, which is the true technical moat**.
Domestic mass production verification tests scaling capabilities, overseas verification tests the generalization capability of technology.
Daily operations can form efficiency advantages through long-term cooperation, supply chain, regulations, and road environments.
But overseas, different countries' roads, regulations, driving habits, and vehicle platforms all change. Whether original technology can be migrated at low cost is the real test.
Overseas business expansion has become Momenta's core growth main line. Vehicle models equipped with Momenta solutions have entered more than 10 countries and regions, including Norway, UK, UAE, Singapore, Australia, New Zealand, Thailand.
Robotaxi business is testing or operating in 3 countries and 6 cities.
Momenta approaching operational break-even is not because it won in a price war.
Under higher intensity R&D investment, delivery efficiency is up, platform reuse is running, revenue growth 76%, R&D growth 18.7%, making every penny of R&D investment leverage more income.
The self-generation logic of autonomous driving is **maximizing the efficiency of R&D investment and making revenue growth continuously outpace investment growth**.
03. When the Scale Flywheel Spins, Autonomous Driving is Not Just Autonomous Driving
If Momenta just wanted to become a more profitable intelligent driving supplier, stopping here is actually enough.
What is truly worth noting is why Momenta continues to put a lot of funds into the world model, Robotaxi, Robovan, and even robots.
The answer may be hidden in the three words 'Scale'.
When mass production vehicles accumulate data, computing power, models, and engineering capabilities to a certain scale, the value of this capability no longer belongs only to cars.
Cao Xudong drew an analogy: realizing autonomous driving is like landing on the moon. Some choose to climb Mount Everest which is closest to the moon, Momenta chooses to build a rocket.
Mass production business is the process of building the rocket; R7 World Model is the engine of this rocket.
The operation of the R7 World Model is based on the closed loop of 'four Scaling':
First layer: **Data Scaling**, more data, model understands the world better.
As of June 2026, over 1 million mass production vehicles are equipped with Momenta solutions, over 13 billion kilometers of real driving mileage, over 100 million segments of 'Golden Data', helping the model understand the operational laws of the physical world more accurately.
Second layer: **Compute Scaling**, more computing power, higher training efficiency.
Momenta's daily over 10 million kilometers of simulation, computing power scale is increasing year by year, models can repeatedly deduce, make mistakes, and optimize in the virtual environment.
Third layer: **Model Scaling**, stronger models, better product experience.
With the growth of data and computing power, the parameter size of the R7 World Model is expanding rapidly, stronger scenario understanding capabilities, more human-like driving behavior, and safer decision-making performance are converted into user-perceivable product experiences.
Fourth layer: **Business Scaling**, more cars, more data.
Data flows to training computing power, computing power feeds models, models get stronger, experience gets better, more vehicle models are willing to place orders, commercial income comes in.
And these new orders of cars on the road become the source of the next round of data. R7 is not an ordinary model, it is a machine that accelerates itself by scale.
If only mass production business is done, R&D investment is expenses, every penny spent, profits get thinner.
But in this model, R&D investment is the fuel of the flywheel, exchanging for the ability to continue rotating and self-accelerating.
Passenger cars, Robotaxi, Robovan, Robotruck, and even future robots can share the same world model foundation.
In the past, R&D investment in autonomous driving was like project costs one by one.
At this stage, **the world model is more like an infrastructure that can be reused across vehicle models and scenarios**, it should not be limited by 'auxiliary driving'.
The break-even path, the core lies in the technical foundation being called repeatedly, the vehicle mounting volume is just a natural result on this road.
Momenta Robovan for scenarios such as express delivery and night delivery
Momenta bets 70% of revenue into R&D, looking not just at immediate profits, but also the larger value space beyond mass production business, promoting from 'ADAS' to a broader physical world.
Robovan is already in trial operation in Suzhou, targeting urban last-mile logistics. Momenta's timeline is also very clear:
Momenta approaching operational break-even is not because it suddenly found the secret to cost reduction and efficiency improvement.
Precisely speaking, **autonomous driving has changed from an expensive technology to an industrial system capable of self-generation and continuous expansion**, the profit inflection point is just a signal that this system starts to turn.
Automotive software may just be the starting point for Momenta to enter physical AI.
If the world model eventually becomes a universal foundation for understanding the physical world, then cars may just be one of its scenarios that are scaled first.
Once real-world data, model capabilities, and mass production scale are ready, the boundaries of business are no longer auxiliary driving, but entering any field that requires understanding the physical world.
