As autonomous driving enters 2026, Physical AI has emerged as a new trend. Momenta, crowned as the first Physical AI stock, listed on the Hong Kong Stock Exchange, with its market capitalization exceeding HKD 70 billion on the first day of trading. Yu Beibei, Vice President of ZYT, once stated bluntly that transforming towards Physical AI is a rule of survival concerning the life and death of manufacturers. When almost all intelligent driving algorithm vendors are labeling themselves with Physical AI, we can reflect on a question: what kind of impact will Physical AI actually bring to the landscape of traditional Tier 1 suppliers?
Is the Model of Selling Hardware and Charging Development Fees Still Viable?
The business logic of traditional Tier 1 suppliers has been highly mature over the past few decades, and the entire model can be summarized as a turnkey project. When an OEM needs to develop a new vehicle model, it will present a complete set of intelligent driving system requirements to Tier 1 suppliers. The Tier 1 supplier is then responsible for integrating sensors, chips, underlying software, and core algorithms into a system that can be directly installed in the vehicle.
The starting point of this project is a substantial R&D fee to cover all investments by the Tier 1 supplier in hardware and software adaptation, system debugging, and testing validation. Once the vehicle model officially enters mass production, the Tier 1 supplier will charge a hardware fee per unit for every vehicle sold. This dual revenue structure of R&D fees plus unit prices is the most core source of profit in the traditional Tier 1 model, and has long supported the solid position of giants such as Bosch, Continental, and Aptiv in the global automotive supply chain.
The emergence of Physical AI is breaking this logic. The core of Physical AI is to enable AI to understand the laws of the physical world, allowing it to perceive, reason, and directly generate actions. This means that the profit pool and bargaining power of the entire industry chain are shifting from the hardware end to the model and algorithm end. ZYT has begun to try breaking the single logic of traditional Tier 1 selling hardware and charging development fees, exploring new business forms based on subscriptions, profit sharing, and action tokens. The financial data of Momenta illustrates this even more clearly: its software licensing revenue surged from RMB 23 million in 2023 to RMB 968 million in 2025, a 42-fold increase in three years.
Software is becoming the most core source of revenue for Tier 1 suppliers.
This change is not partial; the basic rules of the entire industry are changing. When intelligent driving capabilities shift from a bonus item to a basic requirement, and when Urban NOA changes from a technical highlight to a standard configuration for automakers, simply selling hardware can no longer support the valuation of a Tier 1 supplier. The narrative of Physical AI is pursued by the capital market essentially because it points to a larger business imagination space: not just providing intelligent driving for cars, but providing a universal brain for all agents that need to act autonomously in the physical world.
New Players Entering, Old Players Transforming?
Physical AI has also blurred the boundaries of competition. Previously, the competitors of Tier 1 suppliers were very clear—other Tier 1 suppliers—but the current situation is completely different.
NVIDIA is one of the most noteworthy variables. Jensen Huang defined Physical AI as the next core growth wave in the AI industry at CES 2026. NVIDIA not only provides chips but has also launched the Cosmos 3 foundation model, the Alpamayo 2 Super vision-language-action model, and the DRIVE Hyperion full-stack platform. It positions itself as a central platform provider for Physical AI and autonomous driving, which means that traditional Tier 1 suppliers need to develop on its ecosystem, becoming its downstream to some extent.
On the other hand, the layout of automakers is also changing. Leading automakers are turning to full-stack in-house development. Huawei, BYD, and Li Auto are all building their own intelligent driving R&D teams. These automakers are shifting from being Tier 1 customers to direct competitors. For those automakers that choose not to develop in-house, they prefer another cooperation method: directly purchasing a pure algorithm solution, handling the hardware part themselves or through third parties, and only entrusting the core software capabilities to suppliers like Momenta. This ensures that their intelligent driving capabilities do not fall behind while maintaining the initiative in hardware selection, making costs and differentiation space more controllable.
The situation of traditional Tier 1 suppliers has thus become very subtle. Giants such as Bosch and Aptiv are also heavily investing in software capability building, but they face a fundamental challenge: while hardware manufacturing capabilities remain indispensable, it is no longer the hardware itself that determines the ultimate profit attribution of this business.
Consumers can hardly perceive the difference in whose chips or whose sensors are equipped in a car, but whether Urban NOA is easy to use and whether high-speed lane changes are smooth, users can judge as soon as they get their hands on it. This shows that in the profit distribution of the entire supply chain, the right to speak is shifting from hardware integrators to software suppliers that master core algorithms and data capabilities. If traditional Tier 1 suppliers cannot build their own barriers in this link, they will easily be squeezed into pure hardware contract manufacturers, and their profit margins will become thinner and thinner.
Two Routes, Who Has a Better Chance of Winning?
The current autonomous driving supplier market has divided into two major camps. One is the software-hardware integration route led by Huawei and Horizon Robotics, relying on self-developed automotive chips to build a complete ecological barrier; the other is the third-party pure algorithm supplier represented by Momenta, which only outputs software solutions without underlying hardware support.
These two routes have their own advantages and disadvantages. The advantage of the software-hardware integration route is its strong closed-loop capability, allowing complete control from chips to algorithms to data, without worrying about supply chain bottlenecks. However, the disadvantage is that it is asset-heavy and requires large investments, and the iteration cycle of chips is much slower than that of algorithms.
The advantage of the pure algorithm route is being asset-light and fast-iterating. Momenta's mass production solutions have been implemented in more than 10 countries and regions globally, and the number of mass-produced vehicles equipped with its system has exceeded 1 million. According to data released by CIC Consulting in June 2026, in China's third-party Urban NOA supplier market, Momenta's market share has reached 65%. However, the risks of this route are also very obvious. Without underlying hardware, the bargaining power in the industry chain is weak. Moreover, with the rise of the in-house development wave among automakers, pure algorithm suppliers face the long-term risk of customer loss.
ZYT, on the other hand, takes a different path. It uses binocular vision combined with a medium-to-low computing power platform to compress NOA capabilities into the 100,000 to 150,000 RMB vehicle price range. As of March 2026, it has bound 20 automakers and covered 32 brands. This high cost-performance strategy is highly competitive in a market where price wars are intensifying, but whether it can establish real technical barriers in the Physical AI era remains to be seen.
Who Will Be the Real Winner in the Physical AI Era?
So, under the dividend of this Physical AI wave, who is the biggest winner?
From the current situation, those most likely to reap the biggest dividends are the new Tier 1 suppliers that have mastered the closed loop of data, models, and scenarios. Momenta's "one flywheel, two legs" strategy (mutual reinforcement between mass-produced vehicles and Robotaxi) is essentially building this closed loop. Mass-produced vehicles provide massive real-world data, the data trains stronger world models, stronger models support higher-level autonomous driving capabilities, and then monetize through software subscriptions and other models. Once this flywheel starts spinning, the first-mover advantage will be very obvious. As Momenta CEO Cao Xudong said, the marginal cost of software is zero; scale is not only about costs but can also bring exponential improvements in experience.
Companies like NVIDIA that provide underlying computing power and foundation model platforms will also be solid beneficiaries. No matter how algorithms evolve, the demand for computing power is rigid. Strategy& estimates that the global Physical AI market size will reach 430 billion euros by 2030, and NVIDIA occupies a key position in the infrastructure layer of this market.
Under this wave, those at the greatest risk are the traditional Tier 1 suppliers with slow transformation. If they cannot quickly build software capabilities and data closed-loop capabilities, they are likely to be downgraded to pure hardware contract manufacturers, with their profit margins severely squeezed. In the Physical AI era, the profits of the automotive industry are shifting from the hardware body to the intelligent brain. Whoever can seize this brain will master the future pricing power.
#AutonomousDriving #PhysicalAI #Tier1