Robotaxi Economics – The Multi-Trillion Dollar Transformation of Mobility

Published on September 21, 2025

BE THE FIRST TO ACCESS EVERY NEW REPORT

Key messages

Tesla could start removing human safety monitors in selected areas in Austin before the end of the year. If successful, this would mark a key milestone in Physical AI
Once the human safety monitor can be removed, the number of robotaxis a single support staff operator (remote & ground) can manage is surprisingly high
Robotaxis can be operated at very low costs at scale, making them highly cost-competitive not only with traditional taxi services but also with private car ownership, while still offering very attractive margins
Low-cost robotaxis could unlock a >USD 2 trillion mobility-as-a-service (MaaS) market
Tesla is on track to lead the future MaaS market, potentially scaling its U.S. robotaxi fleet to over 12 million vehicles by 2040 and disrupting traditional automakers and related industries
Conclusion: Based on robotaxi economics, Tesla stock currently represents an attractively priced option on long-term upside, provided FSD progress continues at the current pace and no major hurdles arise

(Source: Laniakea Research, internal analysis)

Note: This valuation represents our current best estimate based on publicly available information as of the date of this report. The outlook may change materially as new information becomes available. Please refer to our Legal Disclaimer for additional details. While we use AI tools to refine wording for clarity and readability, all ideas, concepts, models, calculations and insights in this report are the result of our team’s expertise and manual work.

1. Tesla could start removing human safety monitors in selected areas in Austin before the end of the year. If successful, this would mark a key milestone in Physical AI

Tesla launched its robotaxi service in Austin on 22 June 2025 with a small pilot fleet of 10 – 20 Model Y vehicles. Initial reviews describe the experience as remarkably “smooth” featuring very human-like driving behavior to the extent that other road users do not realize that the cars are operating driverless, a perception reinforced by the vehicles’ appearance as ordinary Model Ys. However, according to the Q2 2025 earnings call, the fleet covered just over 7,000 miles in its first month of operation, equating to an average of only over 11.7 – 23.3 miles driven per vehicle per day. This low utilization was likely influenced by the invite-only influencer participants, many of whom left Austin after only a few days of testing. Even within this limited mileage, several safety-critical incidents emerged, highlighting why the current Full Self-Driving (FSD) [1] version still requires a human safety monitor inside the vehicle:

Anchor 1u

(Source: Laniakea Research, internal analysis)

Lane selection/ complex intersections

Most safety-critical incidents reported appear to be linked to lane selection and routing/ navigation, also the leading causes of disengagements in the publically available FSD version 13.2.X, according to crowdsourced data from teslafsdtracker.com [2].

In this video (at the 7:10 mark), the robotaxi incorrectly selects the left-turn lane one intersection too early. The perception neural network then attempts to execute the left turn, but the navigation module seems to “override” the maneuver to continue straight. This results in erratic and indecisive driving behavior, including subsequently crossing double solid yellow lines to correct its course. In a similar case, the robotaxi turns left from a straight-only lane, interfering with oncoming traffic.

Waymo addresses such issues by using HD maps [3] that provide detailed foresight — effectively 3D blueprints of the road network containing precise lane positions, geometries, and associated traffic rules. Without HD maps, a self-driving system must rely entirely on perception to determine the correct lane (much like a human driving in an unfamiliar area), which is significantly more challenging, albeit cheaper and potentially much more scalable.

The latest FSD software can handle most “ordinary” intersections competently, but complex intersections still require additional training. In our view. the system must be able to recognize when it is in the wrong lane and reroute accordingly, as incorrect lane positioning can never be entirely avoided due to traffic conditions and other drivers’ behavior.

Although these lane-selection issues have persisted for some time present in earlier FSD versions, they can likely be mitigated through expanded local training data and simulation, prioritizing the perception neural network’s planning over the navigation/ routing module, and proactively avoiding high-risk intersections altogether.

Unusual events/ road situations

A test driver reported on X (no video available) that the safety monitor had to intervene to prevent the robotaxi from crossing railroad tracks while red warning lights were flashing and the barrier arms were descending. It remains unclear whether the vehicle would have stopped in time without intervention.

This incident underscores the potentially dangerous edge cases that can challenge Tesla’s current self-driving system. While future AI models with larger parameter counts are expected to handle increasingly rare and complex scenarios, in the near term, rigorous safety testing will be essential before deploying an unsupervised robotaxi service in any specific area.

Road sign detection (speeding)

In this incident (at the 14:40 mark), the Tesla robotaxi did not appear to detect or respond to the posted speed limit sign. While the vehicle correctly reduced speed for both the deer and the speed bump, it subsequently accelerated to a velocity significantly above the limit (27 mph vs. 15 mph). Overall, the vehicle maintained safe operation, but did not comply with traffic laws.

Pick-up & drop-off (objects in blind spots ahead of the car)

The current FSD build for robotaxis does not yet utilize the front bumper camera (introduced in the latest Model Y refresh this year), as demonstrated by YouTuber AI DRIVR in multiple tests. In these scenarios, the human safety monitor must intervene whenever an object is placed directly in front of the car within the blind spot not covered by the windshield cameras. We anticipate that the upcoming FSD version 14 release will integrate the front bumper camera, along with the 10x increase in parameter count compared to FSD v13.2.X. This upgrade should enable safer starts from standstills and improve distance and drivable surface detection.

Pull-over feature

The Tesla robotaxi includes a “pull-over” feature, allowing riders to request an immediate stop while the vehicle searches for a safe location. In practice, this functionality appears underdeveloped, sometimes dropping passengers in the middle of intersections or left-turn lanes in the middle of the road. It seems primarily intended for emergency stops where speed takes precedence over safe positioning. While not strictly safety-critical from the vehicle’s perspective (and not necessarily requiring a human safety monitor in the car), we consider the feature, in its current form, unacceptable for wide public deployment.

Running through red lights (FSD 13.2.X)

To our knowledge, no robotaxi incidents involving running red lights have been reported so far, and the issue may have been resolved by now. However, numerous such incidents were observed in the latest publicly available FSD version 13.2.X (see for example here, or here at the 6:16 mark). As AI models grow larger, the neural networks [4] appear to “predict” traffic light changes based on indirect cues, such as traffic lights in other directions or the behavior of nearby vehicles. Due to the inherent nature of neural networks, there is no straightforward way to explicitly prevent this behavior. While the system may only proceed when it “judges” it be safe, running red lights in these situations is clearly unacceptable from a regulatory perspective. We do not view this as an insurmountable barrier to unsupervised robotaxi deployment, but it highlights the inherent limitations and risks of end-to-end AI systems [5].

Despite its current shortcomings, Tesla appears to be on a clear trajectory towards full autonomy with its camera-only approach. None of the remaining (known) issues seem to require additional sensors such as radar or LiDAR [6], equipment used by competitors like Waymo. Instead, progress hinges on making the AI self-driving system more intelligent. Current HW4 [7] vehicles still have substantial compute and memory bandwidth headroom, sufficient to support a ~10x increase in AI model parameter count (vs FSD v13.2.X). The upcoming AI5 chip — expected by the end of 2026 and set to debut in the Cybercab — should deliver a further 8x boost in performance [8]. Historically, such model scaling has significantly improved self-driving performance reducing intervention rates. Even without major architectural improvements, this level of scaling should translate into meaningful progress toward an intervention-free robotaxi service. The fact that Tesla is approaching Waymo’s performance without relying on costly additional sensors is a major competitive advantage, as we will explore later in the report.

Tesla has reportedly secured a Transportation Network Company (TNC) permit from the Texas Department of Licensing and Regulation (TDLR). This license allows it to operate a ride-hailing service across the state, using "automated motor vehicles". A new Texas law taking effect on 1 September, 2025 mandates that operators of fully autonomous (Level 4+ [9]) vehicles must obtain a separate permit from the Texas Department of Motor Vehicles (DMV) before operating without human oversight. To qualify for this permit operators must satisfy the following criteria:

Vehicles equipped with data recording systems (e.g., event data recorders).
Ability to enter a "minimal risk condition" if the autonomous system fails.
A First Responder Interaction Plan, detailing how law enforcement and emergency services interact with the vehicle.
Vehicles must be registered, titled, and insured under Texas law
Compliance with state traffic laws and federal vehicle safety standards.

The DMV has the power to revoke or suspend permits if the vehicles demonstrate unsafe behavior or if incident reports reveal systemic reliability issues. In other jurisdictions (e.g., California), disengagement metrics are required to be reported annually, and regulators sometimes use them to inform operational approvals. We anticipate Tesla to receive this permit as well following the currently ongoing Austin pilot phase. Comparable approval processes will be necessary in other states, which explains why Tesla is actively hiring vehicle operators in New York, Arizona, Nevada, California, and Florida in order to gather data and start demonstrating the reliability of its self-driving system to regulators.

We expect the upcoming fully retrained FSD version 14, expected to arrive by end of September (likely a limited Tesla employee release first), to be advanced enough to allow Tesla to gradually remove the human safety monitors in select geofenced areas following a 2 – 4-month validation and debugging period (i.e., before the end of the year). This version will still require exclusions for specific complex intersections, unusual road conditions (and probably highways) and will be limited to moderate weather conditions as noted by Elon Musk on X. If successful, we anticipate that the removal of human safety monitors will mark a key milestone, potentially triggering a significant market reaction as investors recognize this as one of the first large-scale, physical applications of advanced AI.

Tesla’s expansion strategy for its FSD program will likely involve building multiple new service areas in parallel (U.S. first) while continuing to rely on human safety monitors in the near term. This approach allows Tesla to identify problematic intersections, pick-up/ drop-off points, and other challenging road conditions, ensuring safe operations before broader rollouts. Initially, Tesla will likely “deactivate” any “problematic” locations until the software matures and becomes more reliable.

Evidence suggests that FSD is already capable of generalizing well across locations. For example, Tesla has expanded the Austin service area twice, with little indication that performance is tied to location-specific training. Focusing on advanced AI rather than relying on HD maps allows for expansion to other regions more quickly and at scale (Tesla robotaxis can drive anywhere in theory). However, because each major FSD release requires (new) areas to be (re)tested, edge cases to be debugged, and problematic locations to be selectively avoided, geographic expansion is likely to remain gradual. That said, as the FSD model grows larger and more sophisticated, it should increasingly handle rare edge cases without the need for additional debugging.

The company’s long-term goal is to make its robotaxi service available on all roads and to address use cases not currently covered by traditional services due to cost reasons, such as scheduled daily commutes, routine trips to work, or grocery runs. The upcoming public FSD release could accelerate this testing phase significantly using data of millions of private users. Over time, areas deemed safe enough for operation without human safety monitors could be opened-up to privately-owned HW4 vehicles as well, enabling Level 4 unsupervised FSD outside of the robotaxi program. This would represent a major milestone, likely boosting Tesla’s car sales considerably and providing a competitive edge vs other car manufacturers.

Anchor 3u

Anchor 4u

Anchor 5u

Anchor 6u

Anchor 7u

Anchor 8u

Anchor 9u

Anchor 2u

2. Once the human safety monitor can be removed, the number of robotaxis a single support staff operator (remote & ground) can manage is surprisingly high

The economic feasibility of robotaxis depends heavily on their level of autonomy. Specifically, how many miles a vehicle can operate without requiring human or external intervention. A key metric for evaluating this is what we call the “robotaxi-to-support ratio”, which measures how many robotaxis a single support staff member can manage concurrently on average. By definition, this ratio is 1x for traditional taxis (one driver per vehicle). For robotaxis, however, the ratio is determined by two main factors:

Mean Miles Between Interventions (MMBI): the average distance a robotaxi can travel without requiring human assistance.
Support Staff Capacity: the average number of interventions a support staff member (remote & ground operators) can handle within a given timeframe.

Importantly, this ratio focuses only on operational scalability and excludes fixed overhead costs such as R&D, administrative functions or infrastructure costs. As such, it serves as a direct indicator of the marginal profitability of adding an additional robotaxi to the fleet. All else equal, achieving a ratio of ≥1x is the minimum threshold required for robotaxis to be cost-competitive with traditional taxi services.

Alternatively, the profitability of robotaxis can be evaluated using the “robotaxi-to-employee ratio”, which accounts for all employees across the entire organization, not just remote & ground operators. As the fleet grows, fixed overhead costs (e.g., R&D, administration, infrastructure) are spread across more vehicles, making the per-car overhead increasingly negligible. This ratio therefore serves as a broader indicator of organizational efficiency and highlights how close the robotaxi business as a whole is to achieving profitability.

We can make an educated guess where Waymo currently stands in terms of MMBI and robotaxi-to-support/ employee ratio based on publicly available information. For that we derive a simple model to estimate the discussed ratios based on various variables as defined below:

Anchor 10u

In Alphabet’s Q3 2024 earnings call, CEO Sundar Pichai reported that Waymo vehicles were driving over 1 million fully autonomous miles per week and delivering more than 150,000 paid rides per week (see Q3 earnings call: CEO’s remarks). As of May 2025, according to Waymo’s official blog, the company provides over 250,000 paid robotaxi rides per week across Phoenix, San Francisco, Los Angeles, and Austin, operating a fleet of more than 1,500 vehicles (F).

From these data points, we can estimate that Waymo completes approximately 238,000 autonomous miles per day. On a per-vehicle basis, this equates to about 158 miles driven per day and 23.8 paid rides per day. This further implies an average of 6.7 miles per ride (R) when including deadheading [10], and approximately 4.8 passenger miles per ride, assuming 40% of mileage is deadheading. Notably, this aligns closely with the 4.6-mile average ride length derived from the DPU study discussed later in the report.

Because Waymo does not publicly disclose intervention data, we rely on user-reported experiences to approximate its MMBI. For example, on October 27, 2024, a frequent rider posted on X that it took nearly 50 rides before a remote human driver was required to intervene. This claim was later echoed by a Waymo ML engineer, lending credibility to the report. Based on Waymo’s average ride length, this would imply an MMBI of ~238 miles (excluding deadhead miles; the true figure would be higher if those are included).

For comparison, crowdsourced data indicates that Tesla’s FSD v13.2.X achieves an MMBI of around 34 miles as of the date of this report (see teslafsdtracker). This would imply that Waymo is currently about 7x more reliable in terms of overall interventions than Tesla FSD, which we consider very conservative (in reality, Waymo is likely much more reliable in terms of MMBI than the outdated Tesla FSD v13.2.X).

Using the following assumptions, we can get a rough conservative estimate where Waymo’s robotaxi-to-support ratio currently stands:

We estimate that, with a conservative MMBI of 238 miles, Waymo could operate its entire fleet of 1,500 vehicles with just 68 agents dedicated to remote and ground support, resulting in a robotaxi-to-support ratio of 54.2x. While this estimate relies on a simplified model and limited publicly available intervention data, it highlights a key insight: even at modest autonomy levels, a high robotaxi-to-support ratio is achievable, meaning that each additional vehicle added to the fleet should substantially improve profitability (assuming the extra vehicle is fully utilized).

Still, Waymo is not yet profitable, as confirmed by co-CEO Tekedra Mawakana. Based on our derived robotaxi-to-employee ratio of 0.78x, it is evident that the company must scale further to reach profitability, particularly since the average employee salary is likely significantly higher than that of a typical taxi driver. Accounting for the high cost of vehicles (which we will discuss later in the report), we estimate that Waymo would need a robotaxi-to-employee ratio of roughly 2.5x to break even, which corresponds to a fleet of about 5,000 vehicles, assuming all else remains constant. As more data becomes available, we will refine these estimates to improve accuracy and reliability. For now, this model serves more as a useful framework to understand the economics and key value drivers of the robotaxi business.

Tesla currently requires human safety monitors to supervise its robotaxi operations and remain ready to intervene at all times, which means the robotaxi-to-support ratio remains below 1x. That said, Tesla’s intervention rates in the Austin robotaxi pilot appear not far behind those of Waymo. While publicly available data is limited, independent testing by the YouTuber AI DRIVR offers some insight: over four full days of testing, he completed 69 rides without a single safety-critical incident. There were, however, several instances where remote support was needed. For example, when the vehicle struggled with parking lot navigation, flooding, or objects deliberately placed in the robotaxi’s blind spot ahead of the car. Even so, the total number of interventions (as reported in this video) was fewer than a handful across all 69 rides. Assuming an average ride length of 4.6 miles (see average daily usage in the next section), this corresponds to an MMBI of more than 63 miles and an implied robotaxi-to-support ratio of 14.3x as per our intervention model discussed above. Importantly, many of these interventions appear addressable through relatively minor adjustments, such as optimizing pickup and drop-off points. For instance, one problematic drop-off location at the edge of the geofenced area was seemingly fixed within days, suggesting that Tesla can rapidly improve MMBI by acting on intervention data.

Based on this and other early robotaxi test reports, we expect Tesla to achieve a robotaxi-to-support ratio similar to the conservative estimate calculated for Waymo once the human safety monitor can be removed, i.e., around 50x in well tested and optimized areas. This would make the marginal robotaxi highly profitable, as we will explore in the next section.

3. Robotaxis can be operated at very low costs at scale, making them highly cost-competitive not only with traditional taxi services but also with private car ownership, while still offering very attractive margins

(Source: Laniakea Research, internal analysis)

Anchor 11u

We focus our analysis on three vehicles that we believe are the most relevant to current and future robotaxi operations: the mass-produced Tesla Model Y, which is being used in the Austin robotaxi pilot; the upcoming Cybercab, unveiled last October and designed specifically for autonomy; and Waymo’s heavily modified Jaguar I-Pace, its current primary platform (we will also briefly address the upcoming Zeekr RT). Our bottom-up analysis estimates the cost per paid mile (including 40% deadheading) at scale to be approximately 49.7 ct/ mile for the Tesla Model Y, 31.7 ct/ mile for the Cybercab, and 139.6 ct/ mile for the Waymo I-Pace based on the following assumptions (excluding teleoperation and support costs):

Vehicle cost

For the Model Y, we apply the Q2 2025 overall gross margin of 17.2% to the Long Range Rear-Wheel Drive selling price of USD 44,990 at that time [11]. The robotaxi-configured Model Y is believed to include upgrades such as an enhanced telecommunications unit and self-cleaning cameras, which may increase costs; however, we have not modeled these upgrades separately.

For the Cybercab, we base our estimate on Elon Musk’s statement during the Q3 2024 earnings call that the vehicle would cost roughly USD 25,000 in volume production. We consider this assumption reasonable given the simplified 2-door design, substantially smaller battery capacity, the new “unboxed” manufacturing process, and autonomy-optimized features such as the removal of pedals and steering wheel, as well as reduced performance requirements in areas like acceleration and handling which should reduce costs.

Waymo does not disclose its vehicle costs. Our estimate is based on analyst consensus of approximately USD 150,000 per vehicle, reflecting the base Jaguar I-Pace price of around USD 72,000 plus significant expenses for sensor and compute upgrades (LiDAR, radar, cameras, wiring, and proprietary AI systems). We expect LiDAR costs to continue declining, which should reduce total vehicle costs to well below USD 100,000 once Waymo transitions to the Zeekr RT (or other more cost-efficient models) as its base vehicle.

Battery capacity

For the Model Y and Waymo I-Pace we use the official OEM specifications of 75 kWh and 90 kWh respectively.

The Cybercab’s battery capacity has not been disclosed. Our estimate is based on remarks from Tesla’s Lars Moravy (March 2025, Munro Live, YouTube):

“At this point we want to make sure the Robotaxi can last the full usage of a day without having to recharge too much. So we got a lot of models going around trying to figure out how a typical Robotaxi would go pick up someone, go to the next job… And we think we can do that with a really quite a small battery pack, it’s under 50 kWh and still have over like close to 300 miles of real world range”.

For our calculations, we assume a 50 kWh battery, noting that the implied efficiency of ~6 miles/ kWh seems optimistic and differs from Tesla’s own published efficiency estimates (see next section).

Efficiency

We use EV powertrain efficiencies as reported in the Extended Tesla Impact Report 2024 (p. 35): 3.8 miles/ kWh for the Model Y, 2.7 miles/ kWh for the Waymo I-Pace, and an estimated 5.5 miles/ kWh for the upcoming Cybercab.

Due to its smaller, lighter, and more efficient design optimized for autonomous driving, we expect the Cybercab to achieve a real-world range comparable to the Model Y Long Range RWD despite having a ~33% smaller battery, and a higher range than the current Waymo I-Pace, which uses a substantially larger battery.

Lifespan, full charging equivalents

For the Cybercab, we anticipate the use of Lithium Iron Phosphate (LFP) chemistry, as already implemented in the Model Y Standard Range RWD. LFP offers higher cycle durability (up to 5,000 full charging equivalents) and lower cell/ pack costs, at the expense of lower energy density (higher weight) and reduced peak performance, which is less critical for robotaxis. LFP batteries can be charged to 100% without adversely affecting cycle life, improving real-world effective range. We conservatively assume 2,000 full charging equivalents for the Cybercab as other factors might limit its lifetime, giving an implied lifespan of approximately 550,000 miles. Over the long term, lifespan could approach one million miles once autonomous vehicles are fully optimized for longevity.

For the Model Y and Waymo I-Pace, we apply industry-standard guidance of 1,500 full charging equivalents before reaching approximately 80% state of health (capacity). Cycling between 20 – 80% and gentle usage can significantly extend battery life, yielding an estimated lifespan of around 400,000 miles.

Maintenance & repairs

We base average maintenance costs on CarEdge’s estimates for the first 10 years of service, assuming 10,000 miles per year. Since robotaxis will accumulate far higher mileage and maintenance costs typically rise with usage, we apply a 1.5x adjustment to align with industry studies (see Argonne National Laboratory TCO study). Using this methodology, we project average maintenance costs of 6.0 ct/ mile for the Model Y and 25.8 ct/ mile for the Waymo I-Pace. EV maintenance costs are significantly lower than for traditional ICE vehicles due to regenerative braking, fewer mechanical parts, simpler drivetrains, and no oil changes or complex engine upkeep.

For the Cybercab, we assume a 40% reduction in average maintenance costs, reflecting its lower vehicle cost and optimized design for reliability and easier reparability.

Insurance

We base our insurance costs (full coverage) on CarEdge’s estimates assuming a 40-year old good driver with good credit, resulting in approximately USD 4,500 per year.

For the Cybercab, we assume significantly lower costs, based on the lowest reported premiums for the Model Y across various platforms. Over time, we expect insurance costs for robotaxis to trend towards zero as accidents become increasingly rare with advances in AI and autonomous technology.

Electricity

Average electricity costs are based on the U.S. Energy Information Administration (EIA) reported transportation average as of May 2025: 13.59 ct/ kWh (EIA, Monthly Electric Power Industry Report). We account for grid-to-battery efficiency losses of 7% for the Model Y and Waymo I-Pace.

For the Cybercab, which reportedly uses inductive wireless charging, we assume 10% efficiency losses, even though Tesla stated the efficiency would be “well above 90%” (see post on X).

Taxes/ fees

Robotaxis are expected to incur additional fees and taxes compared to privately-owned EVs due to their commercial operation and higher annual mileage. Cities and states may impose higher registration fees, per-mile road usage charges, and ride-hailing or congestion surcharges to account for economic activity, infrastructure wear, and urban mobility impacts as robotaxis become more widespread.

We use an estimate of 2.5 ct/ mile to cover taxes, tolls, fees, and road usage. This figure is based on current average fees for ICE vehicle owners (~USD 300/ year mostly from gas taxes), adjusted for the ~6.4x higher mileage typically driven by robotaxis, yielding roughly USD 2,000/ year per vehicle.

Infrastructure

Robotaxis require a dedicated, scalable infrastructure network to operate efficiently and reliably, including charging stations, cleaning facilities (internal and external), maintenance and safety inspection centers, and fleet operation hubs.

We expect these operations to be highly automated, with wireless charging pads enabling robotaxis to self-charge autonomously and automated washing and cleaning stations, as demonstrated in Tesla’s previews (see Tesla Youtube).

Using internal bottom-up modeling, we estimate infrastructure costs of 1.9 ct/ mile for charging and 1.3 ct/ mile for cleaning at scale. These estimates account for CapEx for wireless chargers (including installation and grid upgrades) and automated cleaning systems (including consumables). Our modeling indicates a conservative average cost of roughly USD 20 per cleaning, which, at one cleaning per week, translates to approximately USD 1,000 per robotaxi per year.

Daily usage & deadheading

Operating hours: We assume an average of 14 hours of operation per day for a large-scale fleet, reflecting a mix of peak demand, off-peak periods, and downtime for charging, cleaning, maintenance, and waiting. Actual utilization could be lower if robotaxis primarily serve commuting trips, leaving much of the fleet idle during midday. Higher utilization may be possible if vehicles are repurposed for non-passenger services such as food or goods delivery, though that lies outside the scope of this study.

Average trip duration and distance: Estimates are based on data collected by the Department of Public Utilities (DPU, Rideshare in Massachusetts study), which requires rideshare companies to report trips. The dataset includes 91.1 million rides across urban and suburban areas, providing a representative proxy for initial robotaxi deployment. We consider an average trip distance of 4.6 miles and 15.6 minutes, adding 1 extra minute at the start and end of each trip for passenger entry and exit. Trip distances and durations will likely increase over time as ride-hailing services & use cases expand and costs decrease.

Deadheading: Deadhead is defined as non-revenue driving or repositioning, excluding downtime for charging, cleaning, and maintenance (covered in daily operating hours). We assume deadheading accounts for 40% of total miles driven, based on studies and proprietary Uber and Lyft data (see Springer Nature overview). Optimized fleets in dense urban environments could potentially reduce deadheading to as low as 30%.

At scale, we expect robotaxis to complete approximately 29 paid rides per day, based on 14 hours of operation, 17.6 minutes per ride, and 40% deadheading, covering around 220 miles per day. This suggests that robotaxis can operate a full day on a single charge, minimizing downtime by charging during off-peak hours or overnight.

We project annual mileage at roughly 80,000 miles, implying an operating lifespan of ~5 years for the Tesla Model Y and Waymo I-Pace, and ~7 years for the optimized Cybercab.

Due to high utilization compared to privately-owned vehicles, each robotaxi is expected to replace approximately five regular cars at scale, accounting for deadheading.

(Source: Laniakea Research, internal analysis)

Waymo’s cost structure is primarily driven by the high costs of its vehicles and sensor suite, resulting in significant depreciation and maintenance expenses. Over time, we expect the relative share of taxes, fees, infrastructure, and electricity costs to increase, while the share of insurance, depreciation, and maintenance decreases as the autonomous vehicle industry matures.

Total wage costs for a full-time taxi driver, assumed at USD 50,000 per year, would represent roughly 84% of total costs for the Model Y, 89% for the Cybercab, and 65% for the Waymo I-Pace, highlighting the substantial potential for cost savings through autonomous driving technology.

(Source: Laniakea Research, internal analysis)

At an estimated ~50 ct/ mile, Waymo’s I-Pace robotaxis are roughly 2.8x more expensive to operate than the mass-produced Tesla Model Y. Even the upcoming Waymo Zeekr RT, estimated at USD 80,000 including sensors, is likely 30 – 40% more costly to operate than the Model Y.

The Tesla Cybercab is projected to have operating costs of slightly below 20 ct/ mile at scale; accounting for 40% deadheading, this rises to around 32 ct per paid mile.

In the mid- to long-term, we expect further cost reductions, efficiency and reliability improvements, insurance costs trending toward zero as accidents become increasingly rare, and lower deadhead costs from optimized fleet operations. These factors could enable sub-25 ct/ mile operating costs, as suggested by Elon Musk on the Q2 2025 earnings call.

(Source: Laniakea Research, internal analysis)

At full-scale autonomy (mature self-driving systems requiring minimal interventions), operating at a robotaxi-to-support ratio of 256x and support costs of <1 ct/ mile, we expect robotaxi costs to be substantially lower than traditional taxi services and even privately-owned cars, reflecting the fact that the driver accounts for >80% of total costs in traditional taxi services.

For the foreseeable future, robotaxis will still require remote and on-the-ground support to handle rare but inevitable edge cases, ensure regulatory compliance, and assist riders. Remote operators manage unusual situations (e.g., blocked roads, flooding, construction) and customer needs (e.g., locating the car, retrieving lost items), while ground support addresses physical issues such as flat tires, sensor failures, or vehicles stuck in adverse conditions which cannot be solved via teleoperation.

Support costs decline rapidly as the robotaxi-to-support ratio increases with improved autonomy. Based on our model (daily usage, yearly support wage of USD 50k, and 40% deadheading), support costs drop from USD 2.55/ paid mile at 1x, to USD 1.28/paid mile at 2x, eventually reaching <1 ct/ paid mile at 256x, corresponding to eight doublings of MMBI.

We can plot cost per mile versus robotaxi-to-support ratio for the different vehicle models/ scenarios and compare them with traditional taxis and privately-owned cars. This analysis illustrates the cost/ autonomy thresholds at which robotaxis can compete with conventional taxis and surpass private vehicle economics.

We estimate existing urban taxi services to cost roughly USD 3.25/ mile in dense U.S. cities, or about USD 15 per 4.6-mile ride, accounting for base fees and per-mile/ time charges.

For privately-owned cars, we rely on the Bureau of Transportation Statistics (BTS, 2024), estimating USD 0.82/ mile (assuming 15k miles/ year). These estimates do not include parking fees or the time and energy spent finding a free parking spot, which robotaxis eliminate. A 2017 INRIX study found U.S. drivers spent more than USD 3k on parking-related costs. Assuming 15k miles/ year, this corresponds to roughly 20 ct /mile (not adjusted for inflation). A conservative estimate therefore puts the average total cost of owning and operating a car at ~USD 1.02/ mile, higher in dense urban environments and lower in more rural areas. This estimate excludes the opportunity cost of the driver’s time, which we will discuss separately later in the report.

The Tesla Model Y can likely be operated at a slight profit at market rates even with a 1x robotaxi-to-support ratio, whereas Waymo’s Jaguar I-Pace requires a ratio >1x to generate marginal profits to cover fixed costs/ R&D expenses. Applying the Waymo 54.2x robotaxi-to-support ratio from earlier to our cost-per-mile model yields total operational costs of USD 1.44/ mile. This suggests that Waymo could operate the I-Pace at a marginal gross margin of roughly 55% at scale (assuming market-rate fares). Adding an additional vehicle to the fleet would therefore be highly profitable, provided the extra vehicle is fully utilized. At this stage, further cost savings from reduced intervention rates (higher autonomy) are limited, as Waymo has largely already descended the cost curve. Meaningful future reductions would mainly need to come from lower vehicle or sensor costs.

If Tesla were able to eliminate the human safety monitor (which may be possible soon as discussed), the company could likely operate its robotaxi service at marginal costs below USD 0.68/ mile at scale, nearly half the cost of the Waymo I-Pace and below the total cost of ownership (TCO) of a privately-owned car, driven primarily by lower vehicle costs and the absence of expensive sensor equipment.

Robotaxi cost % vs traditional taxi service

(Source: Laniakea Research, internal analysis)

Assuming robotaxi services are priced at current ride-hailing rates (to be discussed later in the report), gross profit margins of up to 93% appear achievable at full autonomy. Even the high-cost, un-optimized Waymo I-Pace could achieve gross margins above 50%, while the currently mass-produced Tesla Model Y has a margin potential exceeding 80%, with the Cybercab potentially surpassing 90%.

(Source: Laniakea Research, internal analysis)

Tesla’s camera-only self-driving approach offers significant profitability advantages. If successful, even the current Model Y would be cheap enough to operate to potentially replace privately-owned cars.

At full autonomy, costs should be low enough to generate substantial gross margins >60% even when offering robotaxi services at USD 1/ mile, likely disrupting private car ownership in many regions. At this price, commuting via robotaxi could be more cost-effective than using a personal vehicle, while also freeing up time for passengers. At that point, we expect new (non-autonomous) car sales to decline drastically as people gradually shift from car ownership to mobility-as-a-service (MaaS, see next section).

Anchor 12u

4. Low-cost robotaxis could unlock a >USD 2 trillion mobility-as-a-service (MaaS) market

To estimate the potential addressable U.S. market for robotaxis, we examine transportation-related Personal Consumption Expenditures (PCE) by product type [12], as reported by the Bureau of Economic Analysis (BEA, revised July 30, 2025). The following expenditure categories represent areas that could plausibly be substituted by robotaxi services (excluding road-based public transit, which is likely to remain cost-competitive against robotaxis):

US mobility personal consumption expenditures

(Source: Laniakea Research, internal analysis)

U.S. consumers currently spend an annualized USD 37.1 billion (Q2 2025) on “taxicabs and ride-hailing services” (including Uber, Lyft, and other companies/ traditional taxis), equivalent to 0.18% of total personal consumption expenditures (PCE). By comparison, total transportation spending, including private vehicle ownership and use, amounts to USD 1.7 trillion, or 8.17% of PCE. While not all of these expenditures can be captured by robotaxis since many consumers will continue to value private car ownership for leisure (e.g., sports cars) or specialized uses (e.g., RVs, motorhomes), a substantial share is likely addressable.

According to the Deloitte 2025 Global Automotive Consumer Study, 44% of U.S. respondents aged 18 - 34 already indicated they would be willing to forgo vehicle ownership in favor of a Mobility-as-a-Service (MaaS) solution. We expect this share to increase substantially once these services are widely available at costs comparable to private car ownership, which as discussed earlier is likely feasible. From an economic perspective, it also makes little sense for private individuals to manage car maintenance, charging/ refueling, cleaning, parking, tolls, and other related tasks, when fleet operators can perform these functions much more cost-effective at scale. For many consumers, mobility is primarily about reaching a destination; eliminating the burdens of car ownership and driving would represent a clear value proposition, even at a premium price point.

A further advantage of robotaxis is the recovery of time currently lost to driving, which could materially increase total spending on mobility services.

According to the American Driving Survey 2023 (AAA Foundation for Traffic Safety, 2024), U.S. private (non-commercial) drivers averaged 60.7 minutes of driving and 29.1 miles per day in 2023. Extrapolated to 237.7 million licensed drivers (U.S. Department of Transportation, Federal Highway Administration), this equates to 88 billion hours behind the wheel and 2.52 trillion miles driven annually.

For context, the average U.S. worker logged 1,805 hours in 2023 (OECD hours worked indicator). Thus, time spent driving corresponds to roughly 48.6 million full-time equivalent (FTE) jobs, or USD 2.4 trillion in imputed labor cost (assuming USD 50k annual wage per FTE). If consumers value their freed-up time at even 20% of their average salary (USD 80k), the implicit value of time savings alone would represent a USD 778 billion annual opportunity, equivalent to roughly USD 0.31 per mile.

While household expenditure shares across categories tend to remain stable over time, we expect transportation and mobility to capture a growing share of consumer budgets as MaaS adoption accelerates and the time-value benefit of robotaxis becomes more widely recognized.

Over time, we expect the theoretically addressable U.S. robotaxi/ MaaS market to exceed USD 2 trillion, representing more than a 50-fold expansion over the current taxi and ride-hailing market. In rural areas, where deadheading costs are higher, it may be more practical for customers to own a robotaxi rather than rely solely on fleet services. Ownership could also appeal to customers seeking premium features, personalization, convenience of leaving belongings in the vehicle, or the ability to transport/ send goods around. We still categorize this model as MaaS, as the vehicle effectively functions as a fully reserved robotaxi, handling its own charging and maintenance during off-hours, while likely incorporating subscription-based recurring revenue streams.

Total adressable market, MaaS, 2 trillion USD

(Source: Laniakea Research, internal analysis)

However, to serve the entire market, robotaxis would need to be exceptionally reliable and capable of operating on all road types and in all weather conditions, including heavy rain, snow, and fog. Achieving this level of robustness is a significant technical challenge that will likely take considerable time to fully resolve.

In practice, many households may only replace their second or third vehicle with MaaS, while retaining at least one car for use cases poorly suited to robotaxis (e.g., vacations, camping, skiing) or for reasons of pleasure, identity, or status. Given that U.S. households currently own an average of 1.83 vehicles, this partial substitution is more realistic than full replacement in the near term. Additionally, cars today often serve as a form of personal storage, a function robotaxis cannot easily replicate.

The transition will also be slowed by asset amortization (households wanting to extract value from existing cars) and reluctance to behavioral shifts, which may only unfold across generations. Giving up car ownership would represent a large change in lifestyle for many people living outside of city centers. While we expect younger generations to adopt this lifestyle quickly (not even learning to drive anymore), older generations may take much longer to adopt.

Considering these factors, robotaxis/ MaaS may only realistically capture a fraction of the total addressable market over the next decade. It should be noted however that MaaS does not necessarily exclude private ownership: manufacturers may offer customizable fully autonomous vehicles to individual buyers, charging ongoing fees for autonomous operation. This hybrid model could help mitigate some adoption barriers.

With the U.S. economy representing roughly one-quarter of global GDP, scaling to a worldwide market implies a total addressable market (TAM) of >USD 8 trillion. However, profitability per vehicle mile will likely be lower in regions with lower per-capita incomes, constraining margins relative to the U.S. and other advanced economies.

Many consumers would likely adopt MaaS even at costs per mile well above USD 1.02 (TCO of a personal car), depending on how much they value the free time gained by not having to drive. Using U.S. household income distribution data from DQYDJ, we can estimate the number of people for whom switching to MaaS would be beneficial, based on their income and the fraction of their wage (f) they assign to the value of in-car free time.

Share of owners that would benefit from switch to MaaS

(Source: Laniakea Research, internal analysis)

Assuming f = 20%, it would be financially advantageous for over 30% of people to switch to MaaS at USD 1.50 per mile. Millions of consumers likely want to stop driving or use chauffeur-like services but currently cannot afford to do so. Robotaxis could gradually capture this untapped demand by steadily lowering costs per mile as the service scales.

Based on an average cost per paid mile of USD 3.25 and a USD 37 billion market, we can estimate total annual paid taxi miles driven at approximately 11.4 billion (excluding deadheading miles, which would make the total significantly higher). Using our model, where a robotaxi can complete 132 paid miles per day (~48,000 per year), we estimate that a minimum of 238,000 robotaxis would be required to serve the entire U.S. taxi and ride-hailing market. Given that Tesla aims to eventually produce at least 2 million units of Cybercab per year as stated by Elon Musk during the Q3 2024 earnings call, it is clear that the company is targeting the entire transportation and mobility market, potentially replacing much of the non-autonomous private car usage.

To cover all 2.52 trillion miles driven annually in the U.S., a minimum of 53 million robotaxis would be needed (likely more than 100 million incl. the share of privately-owned robotaxis with lower utilization). This figure is substantially lower than the 278.9 million registered cars in 2022 (AutoInsurance.com), reflecting the much higher utilization rates achievable with robotaxis.

5. Tesla is on track to lead the future MaaS market, potentially scaling its U.S. robotaxi fleet to over 12 million vehicles by 2040 and disrupting traditional automakers and related industries

According to a study published by Obi in June 2025, Waymo rides in San Francisco were found to be 30 – 40% more expensive than comparable trips with Uber or Lyft while 70% of respondents said they preferred a driverless vehicle over a traditional rideshare or taxi with a driver. We speculate that this preference is driven primarily by privacy, enhanced safety (eliminating risks such as reckless or fatigued driving, poorly maintained vehicles, harassment, or uncomfortable interactions), and the appeal of a customizable in-car experience, including personalized climate control, music, and entertainment.

However, these results may be influenced by tech-savvy early adopters who are willing to pay a premium simply to experience the latest technology. Over time, we expect this premium to diminish as driverless robotaxis become mainstream. For our analysis, we therefore assume an average fare of USD 15 per ride (average of 4.6 miles), consistent with Uber and Lyft pricing in the referenced study. This translates to an average cost of USD 3.25 per mile, as discussed before (in practice, shorter trips tend to be significantly more expensive on a per-mile basis, while longer trips are cheaper).

Applying these pricing assumptions to our model allows us to estimate the payback period [13] for the different vehicle models, that is, the time required to recover each car’s production cost. For this calculation, we use Waymo’s derived robotaxi-to-support ratio of 54.2x and assume the same ratio for Tesla, as it should achieve similar intervention rates once the human safety monitor is removed.

Anchor 13u

Robotaxi models payback period at USD 3.25 per mile operating costs

(Source: Laniakea Research, internal analysis)

We estimate that the Waymo I-Pace would need to operate for more than a year at scale (around 29 paid rides per day) to recover its production costs and turn profitable. By contrast, the far less expensive Tesla Model Y could achieve payback in just over three months, while the Cybercab could reach break-even in under two months. Although building a large-scale robotaxi fleet is highly capital intensive, we expect these funding needs to be readily supported by the substantial cash flows generated from early deployments.

In our view, having a vehicle that can roll out of the factory and pay for itself within just a few months of autonomously transporting customers is truly groundbreaking.

Looking further ahead, we expect prices to decline materially as operators benefit from exceptionally high margins and rising profitability. If robotaxi services were priced at the current cost level of privately owned cars, around USD 1.02 per mile (USD 4.69 per ride), and fleets operated at a robotaxi-to-support ratio of 256x, the economics would still remain very attractive for the Model Y and Cybercab (and the Waymo Zeekr RT to a lesser extent).

Robotaxi models payback period at USD 1.02 per mile operating costs

(Source: Laniakea Research, internal analysis)

MOIC at USD 1.02 per mile operating costs

(Source: Laniakea Research, internal analysis)

Anchor 14u

Our calculations indicate that, under this scenario, the Waymo Zeekr RT would achieve a 2x marginal multiple on invested capital (MOIC) [14] (while the Waymo I-Pace would remain loss-making). By comparison, the Tesla Model Y and Cybercab would generate 4.5x and 10.1x their vehicle costs over their lifespans, with the Cybercab showing a long-term potential exceeding 20x. These economics strongly suggest that robotaxis could ultimately undercut the TCO of private cars and fundamentally reshape the transportation and mobility market, once AI capabilities reach sufficient maturity.

On Tesla’s Q1 2025 earnings call, Elon Musk stated that he expects autonomous driving technology to begin “affecting the bottom line of the company, and start being fundamental” by the second half of 2026. He further predicted that “millions of Teslas will be operating fully autonomously in the second half of next year.” In our view, this impact could materialize through two primary channels, though likely with some delay:

Robotaxi service: For robotaxis to contribute materially to Tesla’s earnings (defined here as >5% of current net income), we estimate that >3,000 Model Ys would need to operate at near-full utilization with minimal human intervention and an ~80% operating margin. This scale would represent roughly double the size of Waymo’s current fleet and about 1% of the total U.S. taxi/ rideshare market. While Tesla has the vehicle base to support such deployment, scaling from the 10 – 20 pilot vehicles in Austin and the cars in San Francisco to >3,000 fully driverless robotaxis within a year appears ambitious. A more realistic interim step is the planned expansion in Austin to >100 vehicles by the public launch targeted for September 2025. To achieve a material earnings impact by the end of 2026, however, Tesla would need to rapidly scale beyond Austin into multiple additional cities soon thereafter.

FSD sales/ subscriptions: Tesla began shipping HW4-equipped cars capable of running the company’s latest FSD software in January 2023 (HW3 vehicles no longer support current versions and would need to be retrofitted). We estimate that there are now 2.5 – 3 million HW4 cars in operation, a figure that could rise to ~4 million by mid-2026, assuming current production trends. At the present subscription price of USD 99 per month, achieving a material earnings impact (>5% of net income) would require adoption by roughly 6.5% of HW4 owners. We believe this level of uptake is unlikely until FSD transitions to unsupervised operation, where drivers are no longer required to pay attention. Given the regulatory and technical challenges discussed earlier in this report, this will likely only be possible in restricted, geofenced areas in certain states. While Tesla could command higher pricing once unsupervised autonomy is available, the addressable market of owners who could fully utilize such a service would likely remain very limited in the near term.

In the short term (next 5 years), Tesla is likely to target the taxi and rideshare market, offering fares at levels comparable to or slightly below current providers. Based on our model, we outline several short-term scenarios for the U.S. market (excluding overhead costs, which should be negligible at scale):

Tesla robotaxi valuation cases, bull, base, bear

(Source: Laniakea Research, internal analysis)

Anchor 15u

Anchor 16u

Tesla Base Case: Tesla’s growth will likely be constrained more by FSD development and regulatory approvals than by robotaxi production capacity. If Tesla can begin operating without human safety monitors by year-end in select geofenced areas, it could realistically scale from the current 10 – 20 vehicles covering roughly 173 square miles in Austin to approximately 100,000 vehicles across around 5,000 square miles over the next five years. Achieving this level of scale would heavily disrupt the current taxi and rideshare market, effectively capturing 38.8% of the market. While this may seem ambitious at first glance, Bond Capital’s research (page 302), based on YipitData booking data, indicates that Waymo already captured an estimated 27% market share in its San Francisco operating zone within just 20 months of operation, surpassing Lyft. Robotaxis could therefore take over the taxi and rideshare market more quickly than many expect (at least in certain regions), with Waymo and Tesla potentially dominating the space in the coming years. If priced at USD 3 per mile (USD 13.80 per ride), the U.S. robotaxi service would generate revenues of USD 14.4 billion, with an attractive marginal gross margin of 87.9% and USD 12.6 billion in gross profit. At a 50x price-to-earnings (P/E) multiple [15], Tesla’s U.S. robotaxi business would be worth an estimated USD 708 billion in 2030 (assuming 2% inflation p.a.). Discounted to today, using a 19.5% discount rate derived using the capital asset pricing model (CAPM) framework [16] (current (19 September 2025) 10-year U.S. treasury risk-free rate of 4.13%, beta of 2.07, assumed market risk premium of 5% and additional idiosyncratic risk premium of 5%), this equates to a valuation of USD 290.6 billion. If Tesla can simultaneously scale globally, the total robotaxi business could be worth up to twice as much, assuming slower scaling and lower margins outside the U.S.

Tesla Bull Case: In the bull case, scaling is still limited by FSD progress and regulatory approval, but we assume more significant advancements in FSD technology. Faster version iterations, fewer debugging requirements, and reduced exclusions for edge cases would allow for quicker expansion of service areas, potentially covering 15,000 square miles by 2030. In this scenario, Tesla could fully disrupt the taxi and rideshare market while gradually lowering prices, thereby expanding the overall market by tapping latent demand from consumers who value the comfort of a car but cannot afford a private chauffeur. Margins would be slightly compressed relative to the base case, but would remain highly attractive, resulting in a U.S. robotaxi valuation of USD 708.7 billion at a 50x P/E multiple.

Tesla Bear Case: For the bear case, we assume slower FSD advancement, longer version iterations, and persistent challenges with edge cases, leading to a much slower expansion of service areas, approximately 1,500 square miles by 2030. Tesla would still have a noticeable impact on the taxi and rideshare market, but at a much smaller scale. Using the same 50x P/E multiple, this scenario results in a U.S. robotaxi valuation of USD 87.2 billion.

Tesla could generate significant additional revenue through FSD subscriptions. Based on current production and sales trends, there could be roughly 5 million HW4+ Teslas on the road in the U.S. by 2030 (Tesla currently sells about 40% of its cars in the U.S., or roughly 650,000 new HW4+ vehicles per year). Tesla could offer unsupervised FSD in the same geofenced areas where its robotaxi service operates. Given that the driver would hand over full responsibility to Tesla (though they might still need to intervene if remote support cannot resolve an incident), demand for this option is likely to command a premium. Once Tesla is operating in major U.S. cities, this option could become attractive for many owners. We expect unsupervised FSD could be priced at USD 299 per month, or alternatively at a per-mile rate for infrequent users, and could capture a significant portion of Tesla owners depending on geographic availability. If 10% of the projected 5 million Tesla owners in 2030 subscribe, this would generate additional revenues of USD 1.8 billion per year, which translates to an increase in valuation of USD 37 billion (2025), assuming the same 50x P/E multiple.

We can run similar cases for Waymo based on the Zeekr RT economics discussed earlier in the report, although Waymo plans to add an additional 2,000 Jaguar I-Pace vehicles to its fleet first by 2026 [17]. Waymo could potentially achieve scaling similar to Tesla, as the main limiting factors are likely to be autonomous technology and regulatory approvals rather than production capacity, areas where Waymo currently holds an advantage. If Waymo continues to grow at its planned pace of 133% per year (from 1,500 to 3,500 vehicles within a year), it could reach 100,000 vehicles by 2030. Although margins are somewhat lower for Waymo due to its higher-cost robotaxis, valuations are very similar to Tesla at current taxi and rideshare pricing. Vehicle and system costs only become critical once prices drop further, particularly if the goal is to target the entire transportation and mobility market.

Anchor 17u

Waymo robotaxi valuation cases, bull, base, bear

(Source: Laniakea Research, internal analysis)

However, while the bull, base, and bear case approach yields results broadly consistent with current market valuations, it has several limitations. First, it is inherently difficult to predict how quickly robotaxi services will scale, and investors are notoriously poor at forecasting timelines, making nearly any valuation defensible. Second, the P/E multiple used in these scenarios, admittedly chosen somewhat arbitrarily, is highly sensitive to future growth expectations at that point in time. For these reasons, it is useful to “zoom out” and consider the longer-term steady-state that the transportation and mobility market is likely to reach.

Given the highly attractive gross margins, robotaxi operators will likely target the entire mobility market once the current taxi and rideshare markets have been captured, gradually reducing fares to reach a broader customer base. As previously discussed, a vehicle like the Cybercab could operate at roughly USD 0.82 per mile, comparable to the cost of private car ownership while effectively providing free self-driving (and not cause parking fees), yet still generate strong gross margins of ~60%. At this price point, it should be possible to convert at least half of the population to robotaxi usage, driven by the free time gained relative to manually-driven cars and the added comfort and privacy compared to public transport.

We expect the future U.S. MaaS market to be dominated by a few large players, driven by strong network effects. As more people use a service, wait times decrease and operational costs drop due to reduced deadheading, creating high barriers to entry for new competitors. Additionally, the largest operators will have access to the most data, enabling faster AI advancement and improved safety. In markets characterized by strong network effects, the market leader could capture up to 70% of market share, while the second-largest player might secure around 20%, similar to the current Uber/Lyft dynamic.

Based on the robotaxi economics and the AI advancements already demonstrated by Tesla and Waymo, it is in our view reasonable to assume that the transportation and mobility market will eventually be fully transformed; the question is not if, but when. For this reason, we propose an alternative valuation approach that focuses on the long-term market share the “winning” provider of MaaS could capture. To operationalize this, we generate a sensitivity table that calculates the present value of the business depending on the projected “end-year” and the share of the mobility market captured, using the following assumptions:

Robotaxi fare: USD 0.82/ mile (USD 3.80 per average ride)
Vehicle economics: 33 ct/ mile (Cybercab at full scale incl. 40% deadheading, not incl. further long-term cost saving potential)
Production limitation: Cybercab production ramp-up similar to Model 3/ Y (top-left trajectories fall away in the table below)
U.S. mobility (MaaS) market: USD 2 trillion as discussed before
Addressable market share: 2/3 of total U.S. population, representing the ~300 biggest urban areas with ~79,000 square miles total and >1,000 people per square mile density
Terminal P/E multiple: 50x, representative of the high margin MaaS revenue model and further growth potential ahead (international and in adjacent opportunities)
Annual inflation (average): 2%
Cost of equity/ discount rate: 19.5% (same as above).

Long-term Tesla robotaxi valuation sensitivity table (#m robotaxis vs timeline):

(Source: Laniakea Research, internal analysis)

We believe Tesla is currently on a trajectory to capture 20% of the entire U.S. mobility market by 2040 (30% of the addressable market), implying a future valuation of approximately USD 19.7 trillion in 2040 (incl. inflation, in 2040 USD) and a current valuation of approximately USD 1.4 trillion. Reaching this scale would require more than 12 million robotaxis operating across all major U.S. urban areas, serving over 50 million customers.

This could also be achieved through a hybrid model, where robotaxis are both operated directly and sold to end customers with a subscription-based self-driving service, similar to today’s Model 3/Y offering. Such a model would extend coverage into rural areas where centrally operated fleets may be less economical. While this approach would require significantly more vehicles to be produced, it could generate comparable or slightly higher profits per customer, combining the upfront margin from vehicle sales with recurring FSD subscription revenue.

At an estimated USD 0.82 per mile, a robotaxi would generate roughly 5x the annual profits of a privately owned vehicle, thanks to much higher daily utilization. Depending on how many consumers prefer to own a personal robotaxi, the total number of vehicles required to achieve a 20% market share could increase substantially. Even so, our valuation estimate remains valid, though achieving it would demand substantially higher production volumes.

The sensitivity table indicates that even if the robotaxi rollout is delayed by several years or Tesla captures a more modest 15% market share, equating to just 6.2 million operational robotaxis in the U.S. by 2040, the company’s current market valuation would still be justified (note that we are only looking at the U.S. market and not globally).

The mobility transformation could be akin to the smartphone revolution, which began with the iPhone in 2007 and unfolded over a similar timeframe, although affecting a much larger market and a far greater share of annual consumer expenditures (~50x larger). We speculate that many traditional car manufacturers will have shrunk significantly or disappeared entirely at this point, as the transformed market leaves little room for the current 800+ auto manufacturers.

(Source: Laniakea Research, internal analysis)

Given the significant growth potential and valuation implications, assessing potential robotaxi adoption trajectories is critical. Tesla has already expanded its Austin service area to 173 square miles, and at the current pace, it should be possible to cover the 300 largest U.S. urban markets (roughly 79,000 square miles) by the mid-2030s. However, additional data is needed to accurately gauge the speed and scalability of this rollout and Tesla is yet to remove the human safety monitors from its robotaxis. The main limiting factor for Tesla will likely be the pace of FSD technology development and the ease with which new geographies can be mastered. Waymo has already proven that autonomous vehicles can operate safely and reliably. In our view, the transformation of the mobility market by 2040 is inevitable, whether driven by Waymo improving cost efficiency and scaling its network, or Tesla advancing its FSD platform. We will continue to monitor the pace of progress over the coming months and years in order to better evaluate which growth trajectories are most feasible and likely to materialize.

Excursus: Tesla has recently announced a new compensation plan, the “2025 CEO Performance Award”, which will be put to a shareholder vote at the annual meeting on 6 November 2025. The plan ties Elon Musk’s compensation to a series of operational and market capitalization milestones, including the deployment of 1 million robotaxis in commercial operation. To unlock the full award (12% of Tesla’s shares), Tesla must achieve by 2035:

A market capitalization of USD 8.5 trillion (6-month/ 30-day trailing average)
USD 400 billion in Adjusted EBITDA (3-year trailing average)
Additional operational milestones: 20 million vehicles delivered, 10 million active FSD subscriptions, and 1 million humanoid bots deployed

The U.S. robotaxi adoption trajectories discussed above would play a critical role in reaching these targets by 2035, contributing as follows (assuming 2% inflation p.a., % in brackets depicting level of end-milestones achieved):

Tesla CEO Performance Award scenarios and level of achievement

(Source: Laniakea Research, internal analysis)

We believe Tesla is on track to have 5.9 million robotaxis in commercial operation in the U.S. alone by 2035, generating approximately USD 195 billion in adjusted EBITDA and supporting a market capitalization of around USD 8.4 trillion (at a 50x P/E multiple). This would make a substantial contribution toward meeting the targets outlined in the 2025 CEO Performance Award.

FSD subscriptions could provide additional upside, though we view the goal of 10 million active subscriptions as overly ambitious (unless the technology is licensed and integrated into major 3rd party manufacturers). At 20 million vehicles delivered by 2035 (not all equipped with HW4+), half would need to carry an FSD subscription to meet this target. In our view, mass adoption will only occur once FSD reaches Level 4+ capability, where driver attention is no longer required. Even then, deployment will likely be restricted to specific geofenced areas, limiting its relevance for most Tesla owners. On more conservative assumptions, 10% adoption (2 million subscriptions) at USD 299 per month, FSD subscriptions would generate about USD 7.2 billion in annual revenue/ profits, a relatively small contribution compared to the robotaxi business. Unless development and commercialization of humanoid robots accelerates dramatically, we believe the robotaxi business will be Tesla’s single largest driver of valuation by 2035.

In the U.S., Tesla is currently ahead in terms of having the necessary capabilities to capitalize on this immense market opportunity and is therefore likely to be the dominant player in the future, at least as it currently stands:

(Source: Laniakea Research, internal analysis)

We note that Tesla’s approach is inherently riskier than Waymo’s, as it relies heavily on the cheap camera-only autonomous system being sufficiently reliable. However, if successful, scaling and disrupting the transportation and mobility markets could be much faster and easier. Additionally, Tesla’s vertical integration could prove to be a significant advantage in this space.

Nevertheless, there are also substantial potential roadblocks ahead. Accidents will inevitably occur as the robotaxi service scales, and such incidents could halt expansion efforts for months or even years. Securing regulatory approvals may take far longer than anticipated, with extended validation and testing periods required to prove safety and build public trust. Progress on FSD could also stall, prove fundamentally unreliable, or encounter difficulties in scaling to new areas, particularly when dealing with edge cases and ensuring consistency across different geographies. In addition, Tesla may face labor pressures and political pushback, including potential bans, restrictions, and resistance from legacy automakers seeking to protect their interests. Infrastructure and logistics present further challenges, as a large-scale rollout of robotaxis will require substantial investments in maintenance depots, cleaning operations, and charging hubs. We believe these potential roadblocks will ultimately be overcome, as the benefits of robotaxis far outweigh the associated challenges and risks. In our view, pressure on regulators to approve robotaxis will likely continue to build, to the point where not approving them would effectively mean forgoing the opportunity to save lives (be unethical), given that robotaxis will likely prove to be much safer than human drivers.

Tesla has also floated the idea of allowing private owners to place their vehicles into the robotaxi fleet and share the revenues generated. While appealing in theory, we believe this model is unlikely to scale meaningfully (if at all) for several reasons:

Safety and liability risks: Tesla would lose control over vehicle condition (e.g., tire wear, maintenance), making it difficult to guarantee safe operation and exposing the company to significant liability.
Service quality concerns: Tesla could not ensure consistent standards (e.g., interior and exterior cleanliness). Owners would also be responsible for dealing with vandalism or damage, along with the associated costs.
Inferior economics: Privately owned Model 3/Y vehicles would be far less profitable than purpose-built Cybercabs, which are designed to operate more efficiently and can be charged and cleaned autonomously. A decentralized fleet of owner-supplied vehicles would also be harder to coordinate and optimize compared to a centralized, Tesla-managed fleet, reducing utilization and margins.
Owner incentives misaligned: Owners may be reluctant to subject their cars to the wear and tear of high-mileage commercial use, especially if the economics are less attractive compared to Tesla operating its own fleet.
Bottlenecks elsewhere: As discussed earlier in the report, Tesla’s growth will likely be constrained by FSD progress and regulatory approvals, not by vehicle availability, reducing the need to convert millions of privately owned cars into robotaxis.

Beyond passenger mobility, there are significant adjacent opportunities, such as food delivery and industrial trucking, that could strongly benefit from autonomous driving technology and its supporting infrastructure. As a result, the market leader in autonomy is well positioned to extend its dominance into these sectors as well. Moreover, advancements in AI (perception, real-world decision-making) as well as scaled manufacturing capabilities, may translate effectively into the field of autonomous robotics. We plan to explore these industries in detail in a future report.

6. Conclusion: Based on robotaxi economics, Tesla stock currently represents an attractively priced option on long-term upside, provided FSD progress continues at the current pace and no major hurdles arise

Inevitable mobility transformation driven by attractive robotaxi economics

(Source: Laniakea Research, internal analysis)

We believe that advancements in autonomous driving technology, combined with the attractive economics of robotaxis, will inevitably lead to a fundamental transformation of the transportation and mobility markets, with the leading player capturing most of the market. Many automakers do not yet seem to recognize that they are in a race against the clock. If a vehicle is not capable of autonomous operation within the next decade, it will likely face significant challenges in reaching a mass market. Similarly, manufacturers that have not transitioned to cost-effective EVs will struggle to supply vehicles to robotaxi fleet operators. As people stop driving themselves, features like speed and handling will matter far less. At that point, the industry will no longer require 1,000+ car brands, putting many automakers at existential risk. This transformation will also have many knock-on effects, including reduced gasoline consumption, fewer parking lots, and a reshaping of car finance and insurance industries (something we plan to explore deeper in another report). Autonomous vehicles may represent the largest near-term real-world AI opportunity, and one that does not depend on artificial superintelligence (ASI). Simply matching and “imitating” good human driving behavior is sufficient, making it achievable with current methods rather than requiring any major new AI breakthroughs.

We are somewhat puzzled that financial markets have not reacted more strongly to advancements (or setbacks) in autonomous driving technology, given the enormous revenue potential at stake. For Tesla in particular, with its low-cost, highly scalable approach, the pace of FSD progress could have profound implications across industries. Perhaps skepticism stems from Elon Musk’s repeated, since-2018, assurances that “full autonomy” was just a year away – promises that have yet to materialize and have eroded market confidence in its near-term viability.

We recommend closely monitoring the upcoming release of FSD version 14, with a focus on the following criteria to better assess both the timeline for removing human safety monitors and the potential pace of robotaxi scale-up:

Disengagement rates and edge-case handling – measurable improvements in reliability and reduced intervention rates (higher MMBI), particularly in complex scenarios
Safety-critical performance – progress in areas such as red-light compliance, lane selection at complex intersections, navigation of unusual road situations, and operation under heavy weather conditions
Geographic generalization – performance outside the current Austin and San Francisco service areas, demonstrating the AI’s ability to adapt to new environments and avoiding overfitting to existing test regions
Validation and debugging efficiency – the time required to identify, address, and deploy fixes for issues, which indicates the degree of reliance on manual fine-tuning and ability to scale quickly
Sensor integration enhancements – including the incorporation of the front bumper camera to improve drivable surface recognition and object detection, as well as the use of onboard microphones to detect and appropriately yield to emergency vehicles

Assessing these criteria should offer valuable insights into the potential growth trajectories Tesla may take. We will refine our model over the coming months based on FSD developments and plan to provide an updated outlook accordingly.

At present, Tesla’s market capitalization appears to already reflect some degree of expected disruption in the ride-hailing market, implying a robotaxi fleet of roughly 100,000 vehicles by the end of the decade. To justify this valuation and the elevated P/E ratio, substantial progress in FSD technology will be required.

On the upside, the market appears to underappreciate the broader potential of the mobility transformation. Sub USD 1/ mile robotaxi services could unlock attractive MaaS margins, while additional revenue streams from adjacent opportunities, such as food delivery and industrial trucking, also appear not to be priced in. Historically, cost reductions of this scale have consistently driven transformative shifts that are often underestimated at the time.

In our view, Tesla stock currently represents an attractively priced option on long-term upside, with the potential for outsized returns as the mobility transformation takes shape. We believe Tesla is on a credible path to capture this future market, more than justifying its valuation. That said, this trajectory remains highly contingent on continued FSD progress. If development were to stagnate, or if evidence emerged that unsupervised FSD is unattainable with the current hardware and sensor suite (which we consider unlikely), the investment case would weaken materially. It is therefore essential to monitor FSD progress continuously and adjust positions swiftly as new information emerges.

Disclosure: We are long shares of Tesla, Inc. (NYSE: TSLA)

Legal Disclaimer

1. No Investment Advice or Solicitation

All information and materials on this website are for informational purposes only. Our content, including research, reports, and opinions, reflects our views on investment ideas and disruptive technologies. Nothing on this website should be construed as an offer, solicitation, or recommendation to buy or sell any security, financial instrument, or service.

Laniakea Research does not provide investment advice or services to the public. You should consult with qualified financial, legal, and tax professionals before making any investment decisions. Any decisions you make based on our research are done at your own risk.

2. Accuracy and Liability

While we strive to ensure the accuracy and reliability of all information, all content is provided on an "as is" basis. Laniakea Research makes no express or implied warranties regarding the information's correctness, merchantability, or fitness for any particular purpose. We are not responsible or liable for any damages or lost opportunities resulting from the use of this data.

All opinions expressed are subject to change without notice, and Laniakea Research assumes no obligation to update or revise any report or information published.

3. Disclosure of Interests

You should assume that Laniakea Research, including its members, partners, affiliates, employees, and/ or consultants, may hold a position in the securities or companies discussed in our research at the time of publication. This means we may realize gains if the price of these securities moves in a favorable direction.

4. Proprietary Information

The Laniakea Research name, logo, and all content on this website, including research reports, text, graphics, and other materials, are the proprietary property of Laniakea Research and are protected by copyright and trademark laws. You may not reproduce, distribute, or otherwise use this content without our express written permission.

5. Third-Party Links

This website may contain links to third-party websites for your convenience. Laniakea Research is not affiliated with these sites and does not endorse or take responsibility for their content. The information on linked sites is prepared by independent providers, and any use of these sites is at your own risk.

6. User Conduct and Severability

You agree not to use this website for any unlawful or prohibited purpose. You will comply with all applicable local, state, national, and international laws and regulations. If any provision of these terms is found to be invalid or unenforceable, the remaining provisions will remain in full force and effect.

Anchor 1

Footnotes

[1] Tesla FSD (Full Self-Driving) is Tesla’s AI-powered self-driving system, designed with the ambition of achieving fully autonomous driving in the future.

[2] The Tesla FSD Tracker is a community-driven project where users log miles driven with FSD enabled and record how often they need to intervene, helping estimate the system’s reliability and progress across software versions. It provides useful insight into trends like “miles between disengagements”, but the data is limited by self-reporting, small and biased samples, and inconsistent definitions of what counts as a failure. As a result, it shows broad improvement over time but cannot fully represent FSD’s safety or performance in all real-world conditions.

[3] HD maps (high-definition maps) are highly detailed digital maps that provide centimeter-level information about roads, including lane markings, traffic signs, and road geometry. In autonomous driving, they help vehicles localize precisely, predict upcoming road features, and plan safe routes. Unlike standard GPS maps, HD maps act as an extra layer of environmental awareness to complement real-time sensor data.

[4] In self-driving systems, neural networks are AI models that process sensor data to recognize lanes, vehicles, pedestrians, and traffic signs. They learn from large amounts of driving data to predict how objects will move and help the car make safe driving decisions. Essentially, they act as the “brain” that interprets the environment for autonomous navigation.

[5] End-to-end AI systems use a single neural network to directly map raw sensor inputs (like camera images) to driving actions, without breaking the process into separate perception, planning, and control modules. This approach can simplify the system and potentially learn complex behaviors from data, but it is limited by a lack of transparency, difficulty in diagnosing errors, and heavy reliance on massive, high-quality training datasets to handle rare or dangerous scenarios safely.

[6] In self-driving cars, LiDAR (Light Detection and Ranging) is a sensor technology that uses laser pulses to create precise, 3D maps of the vehicle’s surroundings. It helps the car detect objects, measure distances, and understand the shape and position of obstacles in real time. LiDAR is often used alongside cameras and radar to improve perception and navigation in autonomous driving systems.

[7] HW4 refers to “Hardware 4,” Tesla’s fourth-generation onboard computing system for full self-driving.

[8] Elon Musk stated on a recent interview at the All-In Summit (see All-In Podcast on Youtube) that AI5, Tesla’s fifth-generation AI chip, would have 8x more compute, 10x more memory, and 5x better memory bandwidth and be up to 40x faster than HW4 on specific metrics.

[9] Level 4 refers to the self-driving classification system (SAE Levels 0 – 5) which is a standardized way to describe how much control a vehicle’s automation system has versus the human driver. It ranges from Level 0, where the driver performs all tasks, to Level 5, where the car can operate fully autonomously in any condition without human input. Intermediate levels show a gradual shift of responsibility: Level 1 assists with one task, Level 2 can handle steering and speed simultaneously but needs driver attention, Level 3 manages driving in certain conditions with the driver ready to intervene, and Level 4 can drive independently in restricted areas. This classification helps regulators, manufacturers, and consumers understand the capabilities and limitations of automated driving systems.

[10] Deadheading refers to the operation of a taxi without a passenger onboard, typically when repositioning to collect a fare, returning to a high-demand area, or traveling back to a depot. These non-revenue-generating trips contribute to additional energy consumption, operational costs, and vehicle wear.

[11] Tesla does not disclose gross margins by business segment. The reported 17.2% gross margin therefore reflects a combination of automotive, energy generation, services, and other operations. However, it serves as a reasonable proxy for automotive performance, as automotive revenues accounted for approximately 74% of total revenues in Q2 2025, with Model 3/ Y representing over 96% of vehicles produced.

[12] Personal Consumption Expenditures (PCE) measure the value of goods and services purchased by U.S. households and nonprofit institutions, and are widely used as a key indicator of consumer spending and inflation trends.

[13] Payback refers to the period required for an investment to generate sufficient cash inflows to recover its initial cost. It provides a simple measure of risk and liquidity, but the payback method does not account for the time value of money or cash flows that occur after the payback period.

[14] Multiple on Invested Capital (MOIC) is a performance metric that expresses the total value generated by an investment relative to the initial capital invested. It offers a straightforward measure of overall return, but MOIC does not reflect the timing of cash flows and therefore does not account for the time value of money.

[15] The Price-to-Earnings (P/E) multiple is a valuation metric that compares a company’s share price to its earnings per share, providing an indication of how much investors are willing to pay for each unit of earnings. While widely used for relative valuation, it can be distorted by accounting practices, cyclical earnings, or differences in growth expectations.

[16] An appropriate discount rate can be estimated using the Capital Asset Pricing Model (CAPM) framework, adjusted for project-specific risk. The risk-free rate is typically proxied by the 10-year Treasury yield, representing the return on a riskless investment. This is then combined with the equity beta, which measures the sensitivity of the investment to overall market movements, multiplied by the market risk premium (the expected excess return of the market over the risk-free rate). To account for unique, non-systematic risks specific to the project, such as technological or regulatory uncertainties, an idiosyncratic risk premium can be added, yielding a total discount rate suitable for valuation.

[17] For simplicity and comparability, we apply the same discount rate used for Tesla to Waymo. The project-specific risks associated with robotaxi deployment should be broadly similar for both companies.

Anchor 2

Anchor 4

Anchor 3

Anchor 5

Anchor 6

Anchor 7

Anchor 8

Anchor 9

Anchor 10

Anchor 11

Anchor 12

Anchor 13

Anchor 14

Anchor 15

Anchor 16

Anchor 17

BE THE FIRST TO ACCESS EVERY NEW REPORT

Key messages

1. Tesla could start removing human safety monitors in selected areas in Austin before the end of the year. If successful, this would mark a key milestone in Physical AI

Lane selection/ complex intersections​

​

Unusual events/ road situations

2. Once the human safety monitor can be removed, the number of robotaxis a single support staff operator (remote & ground) can manage is surprisingly high

​​​

3. Robotaxis can be operated at very low costs at scale, making them highly cost-competitive not only with traditional taxi services but also with private car ownership, while still offering very attractive margins​​

​

Vehicle cost

​​​​

4. Low-cost robotaxis could unlock a >USD 2 trillion mobility-as-a-service (MaaS) market

​

5. Tesla is on track to lead the future MaaS market, potentially scaling its U.S. robotaxi fleet to over 12 million vehicles by 2040 and disrupting traditional automakers and related industries

​

6. Conclusion: Based on robotaxi economics, Tesla stock currently represents an attractively priced option on long-term upside, provided FSD progress continues at the current pace and no major hurdles arise​​​

Disclosure: We are long shares of Tesla, Inc. (NYSE: TSLA)

​​

Legal Disclaimer

​Footnotes

Lane selection/ complex intersections

3. Robotaxis can be operated at very low costs at scale, making them highly cost-competitive not only with traditional taxi services but also with private car ownership, while still offering very attractive margins

6. Conclusion: Based on robotaxi economics, Tesla stock currently represents an attractively priced option on long-term upside, provided FSD progress continues at the current pace and no major hurdles arise

Footnotes