Momenta Bets 'Physical AI' for a Good Price

Momenta's listing is not only a critical moment for itself, but also a public defense for the entire autonomous driving and Physical AI industry.

Author | Intelligent Driving Network Ling Jiang Yu Lai

Editor | Langlang Mountain and Zhiming Mountain

After He Xiaopeng declared GX the first Physical AI car, and Li Xiang called the new L9 an Embodied AI flagship, Volkswagen suddenly realized that smart cars and autonomous driving are no longer the hot celebrities that make people excited and flocking in the capital market.

Yesterday was a sweetie, today is an old wife.

Momenta instantly abandoned the outdated concept of autonomous driving and, as the first Physical AI stock, encountered a surge in subscription in the capital market.

On July 8, Momenta officially listed on the HKEX, opening at 301 HKD, higher than the offer price of 295.6 HKD/share. The intraday high touched 314.80 HKD, with market cap breaking 70 billion HKD at one point, and fell back to around 299 HKD at the end of trading.

Due to the public offering of Momenta receiving 414 times excess subscription, its Green Shoe option will likely be fully exercised. This means it will issue about 22.93 million shares, with total raised funds reaching about 6.8 billion HKD, exceeding its initial estimate of 5.89 billion HKD by nearly 1 billion HKD.

One point the Momenta team is proud of is that it has received high favor from international capital. The international offering part received institutional orders exceeding 100 billion HKD, covering sovereign funds and long-term funds from 15 countries and regions.

Including Singapore Government Investment Corporation (GIC), Fidelity International, BlackRock Group, Franklin Templeton, Abu Dhabi Investment Authority (ADIA), Canada Pension Plan Investment Board (CPPIB), Temasek, Wellington, Schroders, and other global top institutions.

Before listing, the 14 cornerstone investors introduced by Momenta had a lineup that can be called luxurious, with subscription amounting to about 3 billion HKD, exceeding half of its total fundraising.

Such craziness, is this the craze triggered by the first Physical AI stock?

Obviously, it is not that simple.

01.

Cooling Intelligent Driving and Heating 'Physical AI'

A particularly dramatic aspect is that on July 8, five companies rang the bell at HKEX simultaneously, one of which was an autonomous driving company Yikong Smart Driving with a positioning close to Momenta, both are enterprises invested by NIO Capital.

The two enterprises played different scripts during the IPO stage and the first day of listing.

Yikong Smart Driving faces mining area autonomous driving. Compared to Momenta, the scale is very small. Compared to Momenta's fixed price model, its offering interval was set at 81.16 to 87.92 HKD, but it ultimately also realized pricing at the upper limit.

Compared to Momenta's huge popularity, in the pre-market trading the night before listing, its average transaction price was 87.07 HKD. Not only was it lower than the offer price, but intraday it once fell nearly 20%, and the closing gain was only 4%, with volatility close to 25%, showing clear divergence between bulls and bears.

But after officially listing on July 8, compared to Momenta opening high but falling in the afternoon, where it once fell below the offer price, Yikong Smart Driving ultimately recorded a 9.99% increase over the offer price, with total market cap reaching 14.299 billion HKD.

Without the support of the Physical AI concept, both Yikong Smart Driving and Momenta failed to achieve profitability. Both are still deeply trapped in losses, with no clear profitability schedule yet. The capital market's enthusiasm and bearishness for the two companies both focus on the future.

Is the concept of autonomous driving really outdated?

In terms of financing amount, Momenta, holding the most mass production clients, did not exceed Pony.ai's 7.7 billion HKD financing when it IPOed on HKEX last year, but it was higher than Horizon Robotics which is布局ing both chips and algorithms.

But under the flood of AI, the stock prices of these listed intelligent driving companies have mostly halved compared to the offer price.

The decline in the heat of the intelligent driving concept in the capital market is also a fact.

But undeniably, what makes Momenta's future more tangible, and what has more imagination space is still its position established in the passenger vehicle market.

02.

The Terrifying 70% Gross Margin

The 71.6% gross margin announced in the prospectus released by Momenta before listing once triggered a dispute between car companies and Momenta.

In the automotive industry, it is generally believed that Huawei Qiankun is too expensive, so they turn to embrace Intelligent Driving companies known for cost-effectiveness like Zhuoyu and Momenta. They didn't expect their gross margins to actually exceed 50%!

In comparison, the most successful traditional car company now, BYD with Dilyan protection, has a gross margin of only 18.81%, Xiaomi Auto is 20.1%, the most successful in Harmony Intelligent Mobility is Seres with a gross margin of 26.24%. The gross margins of Chinese automotive companies are generally below 15%, some even hovering in single digits. In the industry chain end, CATL, once considered a thorn in the eye, also saw its gross margin at just 24.82% in the first quarter this year.

The logic behind this is actually not complex. Comparing the gross margins of manufacturing with software companies is comparing apples and oranges.

Horizon Robotics, in the same track as Momenta, had a 64.5% gross margin in 2025 (comprehensive gross margin of chips + intelligent driving solutions), WeRide's gross margin is also close to 30%, and Pony.ai is only 15.7%, which is related to its high production operations for Robotaxi and R&D investment.

And Momenta's ultra-high gross margin lies in its about to cross the break-even point.

Momenta's revenue increased from 743 million yuan in 2023 to 2.413 billion yuan in 2025. Gross margin was also gradually improved from 17.5% to 71.6%. R&D investment in 2025 was 1.87 billion yuan, accounting for 77.5% of revenue.

Financial Indicators (Thousand RMB)

2023

2024

2025

Operating Revenue

742,745

1,324,688

2,412,523

Gross Margin

17.5%

49.0%

71.6%

R&D Expenses

(1,281,138)

(1,508,270)

(1,868,906)

Net Loss on Books

(2,570,342)

(3,205,730)

(3,457,912)

Change in Fair Value of Preferred Shares

(1,190,108)

(1,967,501)

(2,842,919)

Adjusted Net Loss (Non-IFRS)

(1,093,037)

(959,204)

(302,804)

Although net losses are still expanding, from 570 million, 3.21 billion to 3.46 billion, but if share-based payments and other factors are excluded, the adjusted loss has narrowed from 1.09 billion yuan in 2023 to 300 million yuan in 2025. By the end of 2025, the company's cash reserves exceeded 10 billion yuan.

The signals transmitted by these numbers are very clear:

Momenta has not yet profited, but it is no longer a pure R&D company living solely on financing. It is a company with fast revenue growth, fast improvement in gross margin, and a business model that has been understood by the market.

On the eve of listing, Momenta announced that its mass production delivery scale broke through 1 million units, successfully delivered more than 100 models, with cumulative designated models exceeding 210 models. Among the top 10 global car companies, 9 have cooperated with it.

In the intelligent driving market, especially high-end intelligent driving, the only enterprise truly comparable to Momenta in scale is Huawei Qiankun.

Regarding high gross margin, Momenta's own explanation is that its revenue composition is roughly divided into two parts:

Technical development services, which means customizing intelligent driving solutions adapted to car manufacturers;

Software licensing, which means collecting intelligent driving system usage fees based on sales volume after model mass production.

The former is related to the number of cooperation models with Momenta, while the latter is linked to model sales volume.

In 2025, Momenta's technical development service revenue was 1.45 billion yuan, and software licensing revenue was 970 million yuan. In 2023, its technical service development revenue share exceeded 90%. Just two years later, its software licensing revenue share has already reached 40.1%.

Software licensing revenue is a business model where marginal cost approaches zero, requiring not too much labor cost, akin to 'lying down to collect money', which significantly raises Momenta's gross margin.

Momenta's gross margin has quadrupled in the past three years. The 71.6% gross margin in 2025 has reached the level of software company SaaS companies. This is exactly the technical compound interest that intelligent driving companies dream of: after a system matures, it can be rapidly replicated in the market.

Momenta CEO Cao Xudong stated: 'Autonomous driving has very strong scale effects and first-mover advantages. Its effect is stronger than the chip industry. Autonomous driving is software, and marginal cost is zero, so its scale effect is stronger.'

However, this model is not unique to Momenta. Inthe increasingly competitive intelligent driving supplier market today, many intelligent driving companies have combined technical development models and software licensing models to provide or develop their intelligent driving solutions for OEMs at zero cost.

Providing solutions to OEMs at below their own cost has been going on for years.

In comparison, Momenta has to some extent established its own moat. This moat is the dual connection of capital and technology established with many car manufacturers.

In the cooperation between Momenta and 24 global vehicle manufacturers, it covers nine of the top 10 global vehicle groups (such as Mercedes-Benz, Toyota, BMW China, etc.) and all of China's top 10 vehicle manufacturers.

Among these users, 7 car manufacturers with global influence, including SAIC, GM, Mercedes-Benz, Toyota, BYD, Hyundai, Chery, etc., are all investors of Momenta.

Mercedes-Benz and BYD even became cornerstone investors in Momenta's IPO.

In the industry chain end, it covers Bosch, Desay SV, Luxshare. Mobility platforms include Uber, Grab, Stone Venture, etc. Tech giants include Tencent, Alibaba Cloud, Ant Group, JD, etc.

IM Motors has achieved two-way shareholding with Momenta. SAIC Motor is Momenta's largest institutional shareholder, and Momenta also holds shares in IM Motors.

A case often cited by Momenta is that from Mercedes-Benz investing in Momenta in 2017 to mass production at the end of 2025, it took exactly 8 years. But once breakthrough is achieved, it is almost one step to the goal: Momenta obtained all of Mercedes-Benz's electric car and oil car business in 2024.

Momenta is replicating CATL and car companies achieving two-way deep binding in various ways in the intelligent driving field, letting car manufacturers participate in its profit growth.

However, OEMs have always been powerful. BYD not only invested in Momenta but also deeply invested in Zhuoyu. Moreover, at the sensitive moment of Momenta's IPO, Wang Chuanfu and Horizon Robotics founder Yu Kai being classmates triggered speculation from the outside world.

Car manufacturers represented by BYD and Chery, on one hand, have cooperated with almost all intelligent driving suppliers. On the other hand, they are still investing huge funds in self-research.

70% gross margin is indeed an irresistible temptation. After the technical route converges, all teams seem to see that the endgame of intelligent driving is not far away.

'Horizon, Zhuoyu, Huawei, and Baidu' are not only dividing spheres of influence around global OEMs, but also turning around to guard against OEM self-research substitution.

This scene has already played out once in the intelligent driving chip field. This year, self-developed chips from NIO, Xpeng, and Li Auto have been mass-produced and installed on vehicles. As a local star company in this market, Horizon Robotics' stock experienced a halving-style plunge.

'Physical AI' is such a new narrative told to both OEMs and the capital market.

From 'Autonomous Driving Solution Supplier' to 'Physical AI Foundation Model Constructor', this is not a choice of Momenta alone, but a common choice of almost all intelligent driving suppliers.

03.

'Physical AI' Hides the Ambition of a Platform Company

Essentially, autonomous driving is physical AI, and physical AI is the intelligent driving of all things. There is no difference between the two.

It is just that Physical AI opens up the imagination space for capital. It accurately describes a clear vision for an intelligent driving company to move from a one-dimensional supplier relying on OEMs to a higher-dimensional AI platform company.

Physical AI, a concept popularized by Jensen Huang, is relative to Digital AI. In the red sea competition of Digital AI, intelligent driving companies become the pioneers of Physical AI. For some top players, they also have data and scale advantages.

Momenta is undoubtedly an outstanding player in both fields. Momenta stated that it has run through data and business scaling through 1 million mass-produced vehicles.

And the time for Momenta to confidently build the world foundation model has matured. It is NVIDIA's launched world model that is evolving from a simulation tool for intelligent driving companies to a main model that generates data for training or testing, becoming a foundation model that can be applied to 'End-to-End Foundation Model Pre-training' to understand physical laws of the world. By compressing the laws of the physical world into the model, World Model acquires common sense of the physical world.

Momenta's launched R7 World Model claims to achieve a 10 to 100-fold generational leap in the overall system performance and upper limits.

World Models can empower four major business forms of passenger cars, Robotaxi, Robovan, and Robotruck with the same underlying system architecture, and are expected to extend to fields such as Embodied AI.

Momenta's vision for Physical AI is All-in-One Platform, achieving technology cross-scenario implementation, becoming an AI platform company.

This minimalist architecture will thoroughly end the past pattern of isolated development in various scenarios. The cross-scenario reuse of core underlying technologies greatly amortizes R&D costs for parallel multiple business lines and improves the overall iteration efficiency of the system.

Momenta describes it as: 'One Model, Multi-dimensional Coexistence', which can greatly dilute R&D costs, and expand with business boundary expansion, exponentially amplifying first-mover advantages, forming an 'Winner Takes All' industry endgame.

Its gross margin will even exceed 70%.

Cao Xudong believes: 'Any artificial intelligence application, once approaching human levels, will far exceed human levels in a very short period of time.'

This means that it may take ten or twenty years to climb ahead, but surpassing humans may happen within one or two years. Currently, almost all intelligent driving companies have announced transformation towards Physical AI, and its urgency lies in this.

Currently, whether adhering to the World Model route or the similarly hot VLA model intelligent driving companies have basically打通ed L2 to L4.

Momenta adheres to the years-long 'One Flywheel, Two Legs' strategy. Based on unified algorithms and sensor architectures, it has achieved the connection between L2++ data flow and L4 technology flow. Its technology can be reused in passenger cars and Robotaxi.

However, today, this increasingly approaching 'Android' intelligent driving platform's biggest market is still in the passenger vehicle market.

In the CIC Consulting data cited by Momenta itself:

By 2030, the global Robo market scale (Robotaxi, Robovan, Robotruck) will reach 199.8 billion USD, among which Robotaxi global and China market scale will grow to about 81.8 billion USD and 38.1 billion USD respectively. China Robotaxi penetration rate is expected to reach 11.9%, higher than the global 5.7%. Robovan global and China market scale will reach about 85 billion USD and 53.5 billion USD respectively, market penetration rate will reach 6.9% and 14.0%. Robotruck global and China market scale will reach about 33 billion USD and 16.5 billion USD respectively, market penetration rate will reach 0.6% and 1.2%.

But the mass production assisted driving market scale by 2030, global market scale will reach 305.9 billion USD, still far exceeding the Robo market:

In 2025, global and China L2-level - Highway NOA / L2-level - City NOA assisted driving and higher-level intelligent driving market scale reached about 20.4 billion USD and 13 billion USD respectively. It is expected to expand to 305.9 billion USD and 166.6 billion USD by 2030, with compound annual growth rates of 71.9% and 66.5% respectively.

And the Robotaxi market faces huge uncertainty due to city access policies, different regulatory regulations in various places, and operating costs.

Momenta's IPO fundraising this time will invest about 20% in Robotaxi commercialization and scaling. On one hand, L4 has huge imagination space, but it is also the most uncertain part. When Robotaxi will truly contribute clear revenue still has no clear schedule.

Meanwhile, in the passenger vehicle market, although Momenta has obvious advantages, it has not killed the game.

In the traditional so-called Horizon, Huawei Qiankun Intelligent Driving, Zhuoyu, there are still WeRide, Pony.ai. Also preparing for IPO are Qingzhou Smart Driving, DeepRoute.ai.

Skyway Technology, Baidu Apollo, etc. have also established certain advantages in the L4 market.

04.

Intelligent Driving Market Has Not Formed Winner Takes All

An intelligent driving platform similar to Android will eventually appear, but ultimately who wins remains, and no company has established an absolute advantage.

This makes the sustainability of Momenta's high gross margins also uncertain.

First, the continuous decline of single car software licensing unit price has become an industry trend. With more players entering the track and acceleration of penetration in low-end models, city NOA software licensing fees have dropped from several thousand yuan in the early days to the level of thousands, with some lightweight solutions having even lower unit prices.

To seize the mid-to-low-end model market, Momenta actively sinks pricing. While expanding on a large scale, it will inevitably face the pressure of shrinkage in single model profit.

Second, customer concentration is extremely high, bargaining power is fragile. In the past three years, the top five customers contributed 62.6% to 86.7% of revenue. In 2025, the revenue share of the largest customer (presumably BYD) reached 21.6%. If a core customer has sales lower than expected or turns to self-research/competitors, licensing fee revenue will suddenly stop. BYD as Momenta's largest customer, annual installation is about 300,000 vehicles. Its strategic choice will directly affect Momenta's revenue stability.

Third, the rebellion of self-research by car manufacturers is accelerating. Practices by top car manufacturers such as Geely, BYD, and NIO show that only by breaking external binding with independent technology can pricing rights be reclaimed.

Geely insists on self-research of Three Electric Systems, Intelligent Driving, Cockpit. Zeekr 9X is equipped with self-developed Qianli Haohan G-ASD Intelligent Driving System, gross margin stabilized above 20%. NIO announced full fleet replacement with self-developed Shenji NX9031 chip, single car cost directly reduced by 10,000 yuan.

It can be said that for Momenta, 71.6% gross margin is both a moat and a gunpowder keg.

Cao Xudong predicts 'Long-term third-party supply will be the market mainstream', but ultimately forming a 2-3 'Winner Takes All' industry endgame is obviously not what OEMs hope to see.

Momenta has the opportunity to be the one among Chinese intelligent driving companies that is first regarded by the market as a platform-type AI enterprise. But if it fails to achieve monopoly advantage in time, it may also be quickly pulled back to the traditional autonomous driving valuation framework of 'High R&D, Heavy Investment, Unprofitable'.

From this angle, Momenta's listing is not only a critical moment for itself, but also a public defense for the entire autonomous driving and Physical AI industry.

Written at the end

After Embodied AI, Physical AI concepts enter public discourse, autonomous driving seems to have lost its charm. An intuitive manifestation is the influx of a large number of autonomous driving talents into Embodied AI startups. Musk dismantling Model S/X production lines to produce robots is a crude and jaw-dropping move.

So, can putting a Physical AI crown on autonomous driving attract the capital frenzy triggered by Digital AI?

Anthropic, founded only five years ago, realized a 965 billion USD stunning valuation of course did not repeat on Momenta.

What can be explained is that the Physical AI concept has a certain deception. It is not completely isolated and relative to Digital AI. For OpenAI, Gemini these digital large models, they can also enter any business track focused by Physical AI.

On May 16 this year, Zhuoyu Technology CEO Shen Shaojie issued a warning to his peers:

'I am very worried about those doing large models entering intelligent driving, which would be a dimensional strike against us. What I am really afraid of is that digital AI large model companies reverse and wipe us out.'

'Except for Tesla, no intelligent driving company has truly implemented Scaling Law (model quality changes as computing power / data / parameters expand); but OpenAI, ByteDance have already run through Scaling Law.'

And large model companies represented by OpenAI follow 'Multimodal + World Model + Autoregressive Reasoning', which naturally suits 'Visual Prediction + Physical Reasoning + Long-horizon Decision Making', which is exactly the technical core of intelligent driving systems.

Large Modelsdon't need to learn to drive from scratch, just fine-tuning, to quickly reach even exceed L2+/L4 capabilities.

In other words, although the company positioning most similar to Momenta is Huawei Infront and Horizon, the long-term competitors for these intelligent driving companies are neither each other nor OEM self-research, but may come from barbarians outside.