Tesla FSD: Can the starry sea withstand the test of reality?

In the previous article “Tesla's 'Secret Maneuver', is the Robotaxi story just a 'smokescreen'?”, Dolphin Jun suggested that Tesla's "Robotaxi banner" is likely "not really about the wine." Therefore, the core questions that follow are, after rapidly expanding to a market value of $1.5 trillion,

What is the commercial path for Tesla's FSD?

In comparison internally and externally, has anyone found the broad road to intelligent driving?

Due to the high uncertainty of the current path and technological iteration, the final outcome may be extremely uncertain, so this article represents only one perspective and focuses on inspiring thoughts.

1. What is the commercial path for Tesla's FSD?

Looking back at Tesla's stock price over the past year, due to zero growth in car sales in 2024 and the lack of a significant increase in FSD adoption rates after the V12 version was pushed in March 2023, Tesla's stock price has actually been fluctuating within the range of 150-260.

After the third-quarter financial report announced the new car rhythm and sales targets for 2025, the stock price only rose back to the upper point of the range around 260.

What truly gave Tesla's stock price wings of imagination and caused it to soar was the dual resonance of "Trump's election + V13 push," which directly pulled Tesla's stock price to $1.5 trillion, with a single share exceeding $450.

From the current standpoint, Dolphin Jun believes it is necessary to re-recognize the new "FSD" and the current competition in the intelligent driving industry.

First, let's talk about Tesla's FSD commercial planning. Based on various information, Dolphin Jun categorizes Tesla's monetization strategy for FSD into three types:

1) To C, targeting consumers:

This involves selling FSD intelligent driving services to Tesla owners. However, this service is based on hardware shipments to accumulate Tesla's ownership, using free trials to form the number of active FSD users, and converting them into active paid FSD users through natural subscriptions, promotional subscriptions (such as direct price reductions, subscription discounts on car payments, insurance, offering points, etc.).

Under the V12 version, the commercialization of FSD (mainly in North America) is quite awkward for users:

a. There are approximately 1.8 million Tesla vehicles in North America;

b. The number of active FSD users is 850,000, while the number of paid users among them is only 550,000c. When V12 was first launched, the conversion rate for paid FSD reached 20%, but it has since only hovered around 10%. In June, the one-time buyout price for FSD was reduced to 8,000 yuan, and after promotions such as continuous subscription discounts and cheaper car insurance, the conversion rate rose to just under 15%.

Under version V13, Tesla's internal goal is to raise the FSD conversion rate to 30% and the number of paid users to 800,000 to 1 million. The path to achieve this is:

a. The self-operated Robotaxi network composed of Model 3/Y will be trial-operated in specific cities in Texas and California, using the ride experience of these taxi users as a promotional point for positive dissemination on social media. V13 aims to minimize the mixed driving feedback from private car owners in various scenarios and road environments by controlling the demographics and scenarios.

b. Various marketing measures, including indirect price reductions;

c. Improving product strength, as the ride-hailing project can collect driving data at specific points to enhance the V13 model.

Looking briefly at the short-term commercialization goals, even if there are 1 million paid FSD users by 2025, under the most ideal monetization path assumption (SaaS model, all users adopting the $99 monthly subscription model, and subscribing every month), the guaranteed FSD SaaS revenue in 2025 would be $1.2 billion.

Because the implied FSD valuation in the current stock price is too high, Dolphin directly compares it with Office 365, which is currently the most universal, essential SaaS software with a valuation in the first tier.

The latest total paid users of Office 365 has exceeded 400 million seats, generating approximately $50 billion for Microsoft annually. Moreover, $50 billion itself is a very good reference—$50 billion in revenue, at a 15X PS, would lead to an FSD valuation of $750 billion, supporting half of Tesla's current total market value of $1.5 trillion.

If Tesla's FSD charges $99 per month under the full SaaS model, to reach the $50 billion level, the number of paying driver accounts that need to be penetrated would be over 40 million.

Based on the estimate that every car sold is currently in service, Tesla's total global fleet is only around 7 million vehicles.

Using only Tesla's fleet as the valuation basis, it is clear that it cannot support the sky-high valuation of FSD.

2) Corporate Sales—FSD Alliance

Thus, in Tesla's commercialization path, planning to connect a smart driving alliance of different car brands centered around FSD has been almost continuously brewingMoreover, in order for FSD to be installed on a sufficient number of vehicles, the hardware threshold requirements for the cars must be low enough, and even new energy vehicles with weak hardware should be able to use this FSD system.

As a software service provider, Dolphin has observed that domestic companies need to make moderate adjustments and adaptations when delivering to customers. It is very likely that there will also be a "corporate version" of FSD in public sales.

Tesla's ideal pricing model is to charge based on the actual number of vehicles used by the customer terminal, the number of paid devices using FSD in the customer's fleet multiplied by the annual fee charged per account.

Currently, from the comprehensive information available, some American car companies, such as General Motors and Ford, may have a certain purchasing intention, but this is likely based on observing the purchasing conversion rate of Tesla's fleet during the version iteration process.

In other words, in an optimistic scenario, if V13 is relatively successful, external authorization for FSD may not happen until after 2026, and revenue from it may not materialize until 2027 or 2028.

This cooperation model has already set a small example for Tesla to some extent, as XPeng has done, but in essence, it reflects the bargaining power in commercial pricing. Whether it can achieve per-vehicle charging actually depends on Tesla's product strength and the monopolistic nature of its products.

3) Smart Ride-Hailing - Robotaxi

Dolphin has already mentioned in the previous article “Tesla's 'Dark Maneuver', Is the Robotaxi Story Just a 'Facade'?” that ride-hailing is a tough business involving people, cars, drivers, platforms, and government regulation, with an unsexy track and business model.

For Tesla, smart ride-hailing Robotaxi requires providing a) FSD as the virtual driver software service, which must take full responsibility in case of issues, so the level of FSD intelligence must be at least L4; b) hardware - the vehicle. Ideally, it should be delivered as an integrated software and hardware solution to avoid insufficient hardware compatibility, which would prevent FSD from achieving its best performance.

Combining the interview and research information collected by Dolphin, let's take a look at Tesla's actual planning route:

1) Testing

Period: Start in Q2 2025, lasting for one and a half years, ending around the end of 2026;

Scenes: Specific cities in California and Texas;

Hardware Services: Model 3/Y, with the possibility of new cars delivered in the first half of 2025, likely fewer than 10,000 units; self-operated fleet;

Smart Driving Services: FSD V13

2) Mass Production

Period: Start in the second half of 2026, scaling up to 500,000 units, with possibly 30,000 to 50,000 Cybercabs;

Hardware: Mainly existing Tesla models; the fleet will still be self-operated;

Scenes: Expanding to the entire United States.

3) Vision:From self-operation to expansion: Sell this integrated taxi service to car rental companies, ride-hailing operators, and individual consumers (individuals can also buy cars, and while sitting in the office, their own cars can earn money by carrying passengers in a shared ride-hailing manner).

Increase sales volume: Based on this, achieve the annual sales target of 4 million for Cybercab.

In fact, the overall statement above can be summarized into several core messages:

a) The deployment of Robotaxi in the first phase is not just about rolling up sleeves to engage in the ride-hailing business; one significant purpose is to improve the paid conversion rate of FSD in the to-C consumer business, with a strong marketing objective;

b) In the second phase target, the deployment of 500,000 self-operated ride-hailing vehicles, compared to the 15 million active drivers under Didi, even if two drivers share one ride-hailing vehicle, that would still require 7 million cars. Tesla's plan for 500,000 ride-hailing vehicles is not at the scale expected for a ride-hailing business.

c) After the demonstration effect of ride-hailing, the focus will shift to expanding the circle, delivering smart car solutions to ride-hailing drivers and car rental companies, thereby further achieving the annual sales target of 2 million for Cybercab (without steering wheel, brake pedal, etc.) with a steady production capacity of 4 million.

How difficult is this goal? The best-selling car globally, the Toyota Corolla, peaked at less than 1.5 million annual sales worldwide. Dolphin Jun estimates that Tesla may need two variants of Cybercab (for example, a sedan version and an SUV version) to achieve Musk's annual sales target of 4 million.

From the initial testing to the expansion where ride-hailing platforms are also partners, the ultimate goal of Tesla's smart ride-hailing project seems to be to foster Tesla's next truly annual production of 4 million cars, capable of integrated hardware and software delivery (owners willing to pay for software), using FSD software to promote hardware car sales, and then charging separately for FSD, achieving a closed-loop ecosystem similar to Apple's hardware-software integration.

This plan also aligns with Dolphin Jun's core judgment in the previous analysis _: “No matter how big the pie is for Tesla's Robotaxi, the ultimate effect is to create a mobility platform that allows more users to experience autonomous driving technology, complete user mindset cultivation, and in turn promote the real purpose—selling cars + integrated FSD software business.”

PS: From this perspective, the possibility of Uber being disrupted by FSD in the short term is not significant; when pessimistic expectations are fully priced in, one should consider the potential opportunities for Uber.

How feasible are these business route planning diagrams? Dolphin Jun will attempt to use the paths of domestic peers as a reference for comparison to understand the feasibility of achieving these business pathsII. Comparison of Domestic and International Players: Has Anyone Really Touched the Broad Road of Intelligent Driving?

According to Musk, in Europe and North America, Tesla's FSD cannot find competitors even with a telescope. Only in the Chinese market are there attempts to carve out their own paths, but there are still a few years of gap.

Here, we will briefly look at the domestic players, whose technical paths and business collaborations are quite diverse, but very few have truly made significant progress.

Domestic players following Tesla's approach of gradually upgrading from L2 to L4 include manufacturers like XPeng, Li Auto, Nio, Huawei, and Xiaomi. In contrast, those aggressively pushing for L4 and L5 with laser radar are mainly Baidu, Pony.ai, and WeRide. When these service providers attempt to implement L2, they generally find that it is not the same technical path, making it difficult to excel at L2. Additionally, Momenta is pursuing a dual-path strategy, as it started with L2 early and has accumulated certain mass production experience, allowing for a quicker switch to end-to-end paths.

Source: Founder Securities Research Institute

1) Manufacturers — The Value of Intelligent Driving is Mainly Reflected in Sales, Without True Independent Sales

For the commercial monetization of these two types of players, let’s first summarize the progressive camp dominated by manufacturers. Simply put, in the domestic market, when manufacturers target end consumers, FSD generally lacks independent charging capability and is mostly bundled in the overall vehicle sales.

For example, the price difference between the Li Auto Pro and Max versions is 30,000 yuan, while the actual additional cost due to differences in intelligence may only be over 10,000 yuan. The remaining 15,000 to 20,000 yuan can be understood as brand premium, higher training costs for advanced intelligent driving, and pricing premiums for intelligent driving services.

2) B2B Sales

In Tesla's corresponding commercial monetization plan for the B2B version of FSD, there are already different types of companies operating in China. However, the operational models are not entirely satisfactory.

In this direction, the main players are independent third-party service providers, primarily following the L4 technical path.

① Transportation Scenarios

1) Pony.ai, Its revenue does not come from ride-hailing but mainly from providing autonomous truck transportation services to B2B clients (scenario downscaling) and technical downscaling (from L4 to L2, providing mass production solutions for OEMs) to achieve self-sustainability.

2) WenYuan ZhiXing, it has more government/traffic management-related businesses than XPeng, such as Robobus and Robotruck. The operation involves purchasing regular buses, then retrofitting them into smart buses, and subsequently selling the vehicles to government clients and others.

The unit price is over 2 million, and the government purchases mainly for demonstration purposes, making it difficult to achieve a large-scale order explosion.

② Traditional OEM Scenario

When L4+ Robotaxi has not yet achieved commercial monetization, independent third-party intelligent driving service providers, in the absence of L4 Robotaxi and toB logistics scenarios, are developing auxiliary driving scenarios for car manufacturers, which essentially still target C-end private car owners.

However, on the L2 technology path, the current high-level intelligent driving solution route has switched from traditional rule-based algorithms to end-to-end solutions, with non-visualization + city NOA being the benchmark for L2+ auxiliary driving providers.

But the technical path for L4 manufacturers' Robotaxi implementations still mainly adopts modular algorithms + high-precision map solutions, usually limited to operations in fixed areas, and there are not many L4 manufacturers that have successfully transitioned to this solution.

From the current competitive landscape, DJI and Horizon, which pursue extreme cost-performance routes for L2+, as well as Huawei and Momenta, which are at the forefront of high-level intelligent driving technology, have formed the first tier of intelligent driving suppliers based on their technical strength and mass production experience. The market structure is increasingly moving towards centralization, leaving little room for other L4 manufacturers to overtake in the curves.

Of course, another aspect is XPeng's cooperation with Volkswagen, part of which involves licensing the XNGP intelligent driving system to Volkswagen, but initially, it will not be linked to sales, and only after 2026 might it be tied to sales.

Overall, while sorting through this area, a very obvious feeling for HaiTun Jun is that the common characteristics of L4 technology deployment, even for trucks primarily running on highways, still face issues such as excessively high hardware costs for L4 technology deployment; the need for safety personnel, and the actual human replacement capability is not strong, leading to significant commercialization difficulties.

Moreover, L2 players primarily from car manufacturers have not yet developed an independent commercial model for sales, and what ultimately materializes is merely using intelligent driving to leverage more car salesSoul Searching Question: Can Tesla Become an "Alternative"?

From the perspective of the domestic commercialization landing mentioned above, Tesla's proposed FSD commercialization path has actually been implemented to varying degrees in China. However, whether through direct sales to consumers or service sales to businesses, the performance has been relatively poor, relying either on blood transfusions from the parent company or financing support.

Can Tesla's planned FSD commercialization path succeed? The key issue should still be focused on the technical path and product strength. In the next article, Dolphin Jun will compare the commercialization of Baidu Robotaxi and the technical situation of FSD with domestic players to see which players in the intelligent driving track have the potential to emerge successfully.

Related research:

“Tesla's 'Secret Maneuver': Is the Robotaxi Story Just a 'Facade'?”

“FSD: Unable to Support Tesla's Next Valuation Miracle”