Momenta leverages data-driven methods and mass production reuse to overcome the "impossible triangle" of Robotaxi.

By Wu Enen

Edited by Chen Jiying

As Elon Musk transitioned from "Minister Ma" to "President Ma," his primary focus was on Robotaxi.

On June 11, Musk announced that Robotaxi services would be temporarily available to the public on June 22.

He also revealed that on June 28, the first Tesla vehicle capable of full self-driving would be driven directly from the factory to a customer's home.

Musk retweeted a video of a Tesla Robotaxi, a Model Y, driving in Austin, praising its "simple and elegant design".

Tesla has thus sounded the rallying call for fully autonomous driving to transition from testing to scaling up.

Perhaps sensing Musk's ambitious stance, another U.S. Robotaxi company, Waymo, swiftly released its latest news, stating that its weekly service volume had reached 250,000 orders. This was less than two months after the last official announcement of 200,000 orders.

Unlike Tesla, Waymo has always been the "leader" in Robotaxi. In June of this year, the two companies will compete head-to-head on the road for the first time.

To paint a vivid picture, in the current global "live streaming sales room" for Robotaxi, the atmosphere is lively, with drums and firecrackers blaring, setting the stage for the operation to "link up" and accelerate commercialization.

In China, companies similar to Waymo in approach, such as AutoX, Pony.ai, and WeRide, have also accelerated their pace of expanding operations in new cities after several years of operational testing. Momenta, which shares a similar approach with Tesla, also announced that it will operate pre-installed mass-produced Robotaxi models without in-car supervision by the end of this year.

In terms of Robotaxi layout, Momenta is currently the Chinese intelligent driving player most akin to Tesla. Both use data-driven, mapless end-to-end technology and reuse mass-produced vehicles for Robotaxi to complete technological, safety, and commercial evolution.

However, the commercialization of Robotaxi has always faced the challenge of the "impossible triangle" of "safety, cost, and scale," which is also a significant factor restricting Chinese and American Robotaxi players from achieving a closed-loop business model at scale.

Whoever can take the lead in breaking the "impossible triangle" of Robotaxi commercialization will occupy a central position in large-scale commercialization. Among them, Tesla is attempting to surpass from another route.

I. Towards L4: Same Direction, Different Paths

In the 1960s, John McCarthy of Dartmouth College first proposed the concept of "autonomous driving," envisaging the use of computers to control vehicles and achieve the goal of autonomous driving. Since then, autonomous driving has become one of the "crowns" of the automotive industry.

In 2016, Musk officially announced Tesla's Robotaxi plan in "Master Plan, Part Deux," not only proposing to put autonomous driving into practice but also introducing the commercial idea of a "shared fleet of Tesla owners." The vision was beautiful, but it was Waymo and Cruise, among other companies taking leapfrogging routes, that first launched Robotaxi test operations in North America.

This trend quickly spread to China, with many players quickly following suit. Today, leading Chinese and American companies stand side by side in the first tier of the global Robotaxi race.

In the global Robotaxi race, there are three types of players, all moving in the same direction but taking different paths.

The first are the "leapfrogging group" that aim directly for L4, including Waymo, AutoX, Pony.ai, WeRide, etc.

The commonality among the leapfrogging group is that they conduct small-scale tests in defined areas, with hardware and software equipped with safety redundancy, mainly using high-precision maps to steadily collect driving data.

The advantage of this route is that it can quickly realize the on-road operation of Robotaxi, collect feedback data first, and even gain social attention to build the company's brand. AutoX has provided over 11 million rides globally, with a year-on-year increase of 75% in the number of services in the first quarter of this year. The capital market is also optimistic about autonomous driving in China. This year, Cathie Wood, known as the "female Warren Buffett," has increased her stake in Baidu through her fund six consecutive times.

The second route is the "gradualist group" that gradually upgrades from L2 to L4, with Tesla as the typical representative.

Tesla began developing autonomous driving technology as early as 2014 and quickly entered L2 autonomous driving in 2016, gradually adding advanced features such as automatic parking, automatic lane changing, and automatic summoning thereafter.

Tesla's gradualist route also has its advantages, with a longer accumulation period and a relatively mature "pure vision + AI neural network technology." Coupled with a large amount of data feeding, it can be quickly replicated once successful.

In addition to the first two routes, there is actually a third "parallel" route, the most typical of which is domestic player Momenta, with L2 and L4 running in parallel. Momenta has proposed a strategy of "one flywheel with two legs," one leg being intelligent assisted driving, providing mass-produced intelligent assisted driving solutions for automakers, and the other leg being scalable Robotaxi, continuously investing in fully autonomous L4 driving solutions.

Interestingly, Momenta was founded in 2016, which happened to be the same year Tesla announced the launch of Robotaxi. After nine years, as Tesla announced that its Robotaxi will soon be operational, Momenta also reached a strategic cooperation with Xiangdao Mobility in the previous month. The two parties will jointly create the world's first Robotaxi fleet based on pre-installed mass production and will first launch operations in Shanghai.

At the same time, Momenta has also reached a strategic cooperation with Uber, planning to start commercial operations in Europe in early 2026, providing a mapless autonomous driving solution based on the mass production vehicle platform, relying on Uber's global network.

From the current perspective, the "concurrent group" and the "gradualist group" are converging towards the same goal, working to solve the impossible triangle of "cost + safety + scale" and open up large-scale commercialization.

II. Pre-installed Mass Production to Lower Robotaxi Prices

What is the biggest factor affecting the commercialization of Robotaxi? Price, price, and price!

Recently, in the latest episode of "The Diary of a CEO," Cathie Wood emphasized her optimism about the future development of Robotaxi again. She pointed out that AI is no longer a technical issue but a cost issue, and "the key to the success or failure of Robotaxi is whether the cost model can work."

When the Robotaxi trend first emerged, there were no related mass-produced models in China, and almost all Robotaxi vehicles were retrofitted. Autonomous driving companies often spared no expense in retrofitting to achieve better results. For example, to ensure perception redundancy, some Robotaxi vehicles even need to carry several lidars.

Therefore, to reduce the cost of Robotaxi to the critical point where ROI can turn positive, the pre-installed mass production solution is the optimal choice.

Mainstream domestic players are all developing pre-installed mass-produced Robotaxi solutions and customizing related models to achieve cost reduction.

Tesla and Momenta, on the other hand, have taken a different approach, reusing mass-produced L2 assisted driving models for L4 Robotaxi.

It is understood that Tesla applies FSD to its existing Model Y and Model 3, achieving a significant breakthrough in the cost threshold of Robotaxi.

In the first domestic Robotaxi solution based on pre-installed mass production, which is a collaboration between Momenta and Xiangdao Mobility, the existing sensors and other hardware of the SAIC IM L7 are reused.

The benefits of pre-installed mass production of mature models are obvious. Since the hardware of mass-produced vehicles has undergone large-scale production and market testing, the cost is relatively low, and there is a stable supply chain, which can significantly reduce the cost per Robotaxi vehicle.

At the same time, pre-installed mass production means that the autonomous driving system is deeply adapted to the vehicle hardware during the vehicle design and production stages and will also follow strict quality control standards. When designing and producing vehicles, automakers will consider the hardware requirements of the autonomous driving system to ensure compatibility and collaborative performance among hardware, improving system stability and reliability.

Moreover, mass-produced vehicles will undergo a large number of road tests and verifications before being launched on the market, and their hardware has been fully tested under various complex conditions. Using these verified hardware for Robotaxi can ensure better stability and durability in actual operations, reducing system failures and safety risks caused by hardware issues.

III. Feeding L4 with L2 Data to Create Learning Robotaxi

The greatest value of hardware reuse lies in the ability to use the massive data obtained from L2 mass-produced models to verify L4 in advance and collect extensive data on extreme situations, thereby enhancing the overall level of autonomous driving model algorithms.

Why does the L4 model need L2 data to feed?

It's actually quite straightforward.

Safety is the bottom line and cornerstone of the Robotaxi model, which has become the underlying consensus among Robotaxi players. After all, there is still a driver to fall back on for assisted driving, but there is no way out or room to shift responsibility for safe driving in Robotaxi.

How to cross the safety threshold? To achieve large-scale commercial Robotaxi, it is necessary to deploy at least tens of thousands of Robotaxi in one city. To ensure the smooth operation of such a large-scale unmanned fleet, the autonomous driving system needs to achieve at least ten or even a hundred times the safety level of human drivers to achieve safe operation at scale. To meet such safety standards, the autonomous driving system must be capable of solving millions of long-tail problems on the road.

And only by feeding enough massive data to the algorithm and iterating the algorithm in a data-driven manner can these long-tail problems that Robotaxi is bound to encounter be addressed to the greatest extent.

How much data is needed?

Currently, the industry consensus is that approximately 100 billion kilometers of data are required, equivalent to the data accumulated by 10 million passenger cars running for one year.

The more important question is, where does the data come from?

Players like Tesla, which belong to the "gradualist group," have millions of L2 model sales, continuously providing data streams from all over the world. After being processed and learned by algorithms, these data can be fed back to L4.

According to Tesla, Robotaxi services will use data to "feed back" technology, helping Tesla continue to develop and validate the FSD network, mobile applications, vehicle allocation, task control, and remote assistance operations, among others.

Momenta, which follows a similar path to Tesla, is also the same. With L2 and L4 running in parallel, Momenta's automaker friend circle is growing wider and wider.

During this year's Shanghai Auto Show, Momenta successively signed strategic cooperation agreements with well-known brands such as Audi, Cadillac, Honda, Toyota, and Volkswagen, with over 130 models in cooperation for mass production.

Currently, over 15 automakers worldwide have established cooperation with Momenta, covering Japanese giants such as Toyota, Nissan, and Honda, German giants such as Volkswagen, Audi, and Mercedes-Benz, American giant General Motors, as well as local automakers such as BYD, FAW, SAIC, GAC, and Chery.

According to a report by Zooth Research, from January 2023 to October 2024, Momenta's market share in the third-party intelligent driving market for urban NOA reached 60.1%.

Momenta's vehicle loading volume is growing at a rapid pace, with the first 100,000 vehicles taking two years to load, the second 100,000 vehicles taking only half a year, and the third 100,000 vehicles taking less than three months.

Today, 300,000 mass-produced vehicles equipped with Momenta's L2 assisted driving system are driving on the road, capturing real "road fragments." Given that the complexity of Chinese roads far exceeds that of Europe, the United States, Japan, and South Korea, the value of the "data gold mine" accumulated by Momenta is helping it enter a new stage of scalable L4.

After the L2 product line can output a steady stream of data, Momenta relies on a data-driven automated closed loop, which is then fed back to the Robotaxi mass production product technology stream, forming an efficient and interconnected mechanism based on a unified sensor platform.

At the same time, like Tesla, Momenta relies on mapless end-to-end technology, which can break the geographical constraints of high-precision maps and quickly replicate learning results to different urban road environments through deep learning and real-time perception, enabling rapid adaptation.

Under the premise of safety, using pre-installed mass production to break the "cost" barrier and using mapless end-to-end technology to achieve rapid city expansion, the "impossible triangle" of Robotaxi is being shattered.

IV. Conclusion

Today, the distinct value of various Robotaxi routes is increasingly apparent. Tesla, representing the gradualist approach, and Momenta, embodying the concurrent method, must persist in validating the practicality of their respective routes. This endeavor not only paves the way for a new paradigm in the trillion-dollar future mobility market but also lays the groundwork for explosive growth.

Moving forward, these two entities will be both allies, walking hand in hand, and rivals, competing fiercely on the same stage.

Tesla, akin to Apple in fostering a closed-loop ecosystem, as the leading electric vehicle manufacturer globally, is poised to integrate Robotaxi technology into the millions of vehicles it manufactures annually. Conversely, Momenta resembles the Android system, distinguished by its openness. Its network of automaker partners is continually expanding, and with the concerted efforts of its ecological alliance, it is anticipated to emerge as a formidable contender in the Robotaxi sector, alongside Tesla, in the years to come.