Few products carry a name that promises as much, and delivers as little of it today, as Tesla's Full Self-Driving. The system is officially branded "Full Self-Driving (Supervised)" and classified at Level 2, which is a polite way of saying it still needs an attentive human behind the wheel who is legally on the hook if anything goes wrong. Yet the stock does not trade on what FSD is; it trades on what it might become: a driverless robotaxi network that would turn Tesla from an automaker into a software and mobility platform. That gap between the product and the promise is the whole trade. A software release, a regulatory ruling, or a viral crash clip can swing TSLA well before any of it touches earnings, because the market is constantly re-pricing an AI "second act" that has not yet arrived.
For anyone holding or trading tokenized Tesla on BloFin, this matters directly. TSLAx mirrors the underlying TSLA equity price, so every FSD headline that moves the stock in New York also moves the token on BloFin. This article focuses on how FSD and AI shape Tesla's valuation story. For a primer on the token itself, see what tokenized Tesla (TSLAx) is, and for the full catalyst map including deliveries, margins, and macro factors, see what moves Tesla stock price.
Autopilot to FSD: How Tesla's self driving journey began
Tesla's self-driving ambition did not start with neural networks or billion-mile datasets. It started with a partnership. In 2014 and 2015, Tesla launched Autopilot Hardware 1 (HW1) in collaboration with Mobileye, an Israeli computer vision company. The first wide software release, version 7.0 in October 2015, brought lane keeping and Traffic-Aware Cruise Control to Model S owners. It was exciting but limited: highway-focused, heavily rule-based, and dependent on Mobileye's EyeQ3 chip.
Then came the inflection point. In May 2016, a fatal crash in Florida involving a Tesla on Autopilot made global headlines. The system failed to distinguish a white truck trailer against a bright sky. Tesla ended its partnership with Mobileye shortly after, and the company committed to building its own perception and autonomy stack from scratch.
Key milestones in that early phase:
"Enhanced Autopilot" added Navigate on Autopilot for highway lane changes and interchanges, signaling growing ambition.
"Full Self-Driving Capability" appeared as a purchase option, shifting the brand narrative from driver assist toward eventual autonomy.
Elon Musk began making public timeline promises for coast-to-coast autonomous drives, a feature that would not materialize on schedule.
Tesla's Full Self-Driving initiative became a case study at the intersection of AI and physical mobility. The FSD system continually improves through software updates after the vehicle sale, meaning every car shipped is a potential future revenue stream if the AI gets good enough. That concept, software eating the car, is what turned Tesla from a niche EV maker into a technology stock in the eyes of many traders.
Hardware generations: From Mobileye to Tesla's in-house FSD computers
Understanding the hardware roadmap matters because each generation unlocks new software capability and defines which Tesla vehicles can run the latest AI models.
HW1 (2014 to 2016) used the Mobileye EyeQ3 chip with a single forward-facing camera, radar, and ultrasonic sensors. Compute was minimal, roughly a quarter of a TOPS (trillion operations per second), and the system was limited to highway features. HW1 vehicles cannot run modern FSD software.
HW2 arrived in late 2016 with a dramatic upgrade: eight cameras, forward-facing radar, twelve ultrasonic sensors, and an Nvidia Drive PX compute platform. This enabled "Autopilot 2.0," but the software initially lagged behind what HW1 could do. HW2.5 followed in 2017 with added redundancy and incremental improvements.
HW3, launched in 2019, was the first FSD computer designed entirely in-house by Tesla. A dual-chip design delivering roughly 144 TOPS, it was the backbone for FSD Beta testing and early city-street autonomy attempts.
Then came HW4, which Tesla began shipping in cars starting in January 2023 (source: Teslant). Tesla's HW4 is three to eight times more powerful than HW3, with peak compute around 243 TOPS. Tesla's HW4 has 16 GB of RAM and 256 GB of storage, higher-resolution cameras (roughly 5 megapixel versus 1.2 megapixel), and a new bumper camera. It shipped in refreshed Model S/X, Model 3 Highland, Model Y Juniper, and Cybertruck, all equipped with the latest sensor suite.
Tesla also removed radar entirely, committing to a camera-only "Tesla Vision" perception stack. For bulls, this hardware roadmap is central to the argument that the fleet can be upgraded by software over time: every car with HW4 is a future-capable AI platform, not just a vehicle.
From code to neural nets: The AI pivot behind FSD
Early Autopilot logic was essentially a stack of if-then rules layered on top of traditional computer vision. If the system detected a lane marking at position X and a car at distance Y, a deterministic algorithm calculated how much to steer, brake, or accelerate. This worked reasonably well on highways with clear markings and predictable traffic, but it broke down in the messy, unpredictable environment of city streets and construction zones.
Tesla's shift was to progressively replace those hand-written rules with deep neural networks trained on video data. Rather than coding every possible scenario, the system learned patterns from millions of real-world driving clips. Deep neural networks are trained to interpret camera images for driving decisions, and Tesla must train them on large video datasets of real driving behavior rather than rely on crafted rules. The transition to end-to-end neural networks replaces rule-based driving aids with AI, a fundamental philosophical change.
FSD Version 12, which rolled out in late 2023 and through 2024, represented a major inflection. Elon Musk said the update removed over 300,000 lines of hand-written C++ and Python logic, replacing many classical modules with a single large "end-to-end" AI model. This model takes raw video frames as inputs and outputs steering, throttle, and braking commands directly.
What does "end-to-end" mean in practice? Instead of separate modules for lane detection, object recognition, path planning, and control, a single neural network handles the entire pipeline. Tesla's AI uses imitation learning to navigate complex driving scenarios based on human behavior, meaning the model learns from how millions of human drivers actually behave in real traffic.
For traders familiar with large language models and AI-driven technology, this is the same paradigm: massive data plus massive compute producing emergent capability. FSD's methodology could serve as a blueprint for general-purpose physical automation across industries, which is partly why Tesla trades with an AI premium.
Data advantage: Billions of miles as Tesla's AI training fuel
Scale matters in machine learning, and Tesla's data advantage is enormous.
FSD vehicles have driven over 10 billion miles as of May 2026, with roughly 3.7 billion of those on city streets. Tesla's fleet has surpassed 13 billion total driven miles, providing data for deep-learning models across all conditions. The FSD software has been trained on 10 billion miles driven, creating one of the largest real-world driving datasets in existence. By late April 2026, the FSD-engaged fleet was logging approximately 29 million miles per day, up from around 14 million miles per day earlier that year.
Why equity markets care:
The Tesla data flywheel collects data from its global fleet to improve machine learning models. Every mile driven, whether FSD is active or not, feeds back into training.
Tesla uses "shadow mode," where the system predicts what it would have done even when not controlling the car. Discrepancies between the AI's prediction and the human driver's action are logged as training examples for rare corner cases.
Neural world simulators accelerate Tesla's training by reconstructing real-world scenes for testing, allowing engineers to replay and modify scenarios without needing physical road time.
Geographic diversity spans many regions, exposing the model to different traffic laws, lighting, weather, and road designs.
Bulls argue that this data scale plus automated labeling (for example, auto-labeling 3D scenes from video) creates a durable AI moat. If true, it supports higher multiples for TSLA and, by extension, tokenized TSLA exposure. Bears counter that raw miles are not the same as high-quality labeled data, and that the long tail of rare but dangerous scenarios requires something beyond scale alone.
The compute story: Dojo's end and the pivot to in-house AI chips
Training end-to-end networks on billions of video miles demands enormous compute. Tesla initially relied on clusters of Nvidia GPUs, and in 2021, at AI Day, it announced Dojo: a custom supercomputer built around proprietary D1 chips and training tiles optimized for video workloads. For several years, Dojo was a centerpiece of the bull case, positioned as a potential multibillion-dollar advantage in training throughput.
That chapter has since closed. In August 2025, Tesla wound down the Dojo program and disbanded the team, with Musk describing the next-generation Dojo design as an "evolutionary dead end." Its lead departed, the follow-on D2 chip was shelved, and Tesla redirected effort toward a new generation of in-house chips, AI5 and AI6, manufactured by external foundries and intended to power both onboard inference in the car and large-scale AI training. In parallel, Tesla continues to run large Nvidia-based training clusters. The strategic point for traders survives the reorganization intact: Tesla is spending heavily on AI compute, and that spending is a direct financial expression of how seriously it is pursuing the FSD and robotaxi future.
High AI capex compresses margins in the short term. In bullish valuation models, though, it is treated as an intangible infrastructure asset that compounds Tesla's competitive position over time. The same infrastructure narrative also lifts other equities, including Nvidia, which is compared to Tesla in a separate article on BloFin Academy.
What FSD actually does today: Feature set vs advanced driver assistance systems vs "Full Self-Driving" name
The gap between the marketing label and the technical reality is one of the most debated aspects of the Tesla FSD story. Here is what the system can and cannot do as of mid-2026.
FSD requires active supervision from a human driver at all times. The driver must remain in the driver's seat with hands on the steering wheel and eyes on the road. Despite the "Full Self-Driving" name, Tesla's system is classified at SAE Level 2, meaning the car assists but the human is always legally responsible.
Core capabilities in North America include:
Lane keeping and adaptive cruise control on highways
Navigate on Autosteer (formerly called Navigate on Autopilot), handling highway lane changes, on-ramps, off-ramps, and interchanges
Traffic light and stop sign recognition and response (FSD can recognize traffic lights and stop signs since 2026 in its latest iterations)
Autosteer on city streets, allowing the system to navigate local roads, intersections, and turns
Collision avoidance with automatic emergency braking when the system detects other cars, pedestrians, or obstacles
Supplementary FSD features include Autopark, Smart Summon (calling the car from a parking lot to a pickup point), and the vision-based parking introduced with Tesla's FSD version 12.3.3, which enabled vision-based Autopark on newer hardware. Arrival options let the car maneuver to a nearby parking spot at the destination. Availability varies by region and regulation.
The critical point for traders: the human driver remains legally responsible for the vehicle at all times. This shapes Tesla's liability profile and directly influences how regulators classify the technology in valuation debates.
Supervised, not autonomous: The April 2024 renaming and what changed
In April 2024, Tesla renamed "FSD Beta" to "Full Self-Driving (Supervised)," coinciding with the wider rollout of the version 12 end-to-end network. Adding "Supervised" was meant to make clear that active human supervision is still required. This was not merely cosmetic. It followed years of regulatory scrutiny and lawsuits alleging Tesla oversold the capability of its Autopilot and FSD features, with regulators in the US and Europe arguing the branding misled customers into thinking the car could drive itself unsupervised.
Investors read the rebrand two ways: as legal risk management, since labeling the system "Supervised" strengthens Tesla's argument that it never claimed full autonomy, and as a confidence signal that Tesla was comfortable enough with real-world performance to encourage daily use. In 2026, Tesla went further, retiring the "Autopilot" name in software and renaming the onboard hardware from "FSD Computer" to "AI Computer," with "Navigate on Autopilot" becoming "Navigate on Autosteer." The changes were framed partly as regulatory compliance, including pressure from California's DMV over self-driving terminology and how advanced driver assistance systems are described and marketed to customers, and partly as a broader effort to recast Tesla's public identity around AI. These seemingly cosmetic decisions can trigger repricing of legal-risk premiums in TSLA, which is why traders watch them closely.
Pricing, subscriptions, and monetization of FSD
The revenue model behind FSD is central to the AI valuation story. If Tesla can convert its installed fleet into recurring software subscribers, margins could look more like a technology company than a car manufacturer.
Historically, Tesla sold FSD as a one-time purchase. The price started at a few thousand dollars in 2016 and 2017, climbed to around $12,000 by 2022, and reached as high as $15,000 in the US before later adjustments. Each price increase signaled Tesla's belief that the software was becoming more valuable as the AI improved.
In 2021, Tesla introduced an FSD subscription option in North America. Tesla's FSD subscription price was reduced to $99 per month, making it accessible to customers who did not want to commit to a large upfront payment. Owners who had previously purchased Enhanced Autopilot qualified for a discounted rate of $49 per month.
By February 2026, Tesla stopped selling the one-time purchase option for new FSD (Supervised) users entirely, moving to subscription-only access. Legacy owners who had already bought lifetime licenses kept their access, but all new customers must subscribe. This shift complicates revenue modeling: analysts now need to forecast monthly churn, take-rates across different hardware generations, and regional pricing variations.
The bull thesis often includes FSD ARPU (average revenue per user) scenarios. Even modest take-rates, say 10 to 20 percent of an installed fleet of millions of Tesla vehicles, at $99 per month would generate billions in annual high-margin software revenue. That kind of recurring income stream supports premium multiples for TSLA and TSLAx, because software revenue is valued differently than one-time car sales. Bears counter that actual take-rates remain unclear and that many customers try FSD briefly before canceling.
Adoption, real-world usage, and why many human drivers still hesitate
The gap between technical capability and customer adoption is one of the most important dynamics for traders to understand.
FSD Beta had about 360,000 participants by February 2023, a number that grew substantially as Tesla expanded access through free trials, mandatory demos at delivery centers, and in-car prompts encouraging customers to activate the feature through the Tesla app. By mid-2020s, FSD mileage grew into billions of miles driven, but many individual owners still use the system for only a fraction of their total driving.
Reports suggest that a significant portion of owners engage FSD for roughly 10 to 20 percent of their miles. Common reasons include discomfort with the system's behavior in unfamiliar areas, phantom braking incidents, and general unease about trusting AI with safety-critical decisions. Media coverage of disengagements and viral dashcam videos showing the car making unexpected maneuvers, sometimes genuinely concerning, can overshadow incremental technical improvements that long-term users notice between updates.
The safety and effectiveness of partial automated systems depend on clear communication about their limitations. When customers expect the car to be fully autonomous but discover it still requires constant attention, dissatisfaction and cancellations follow. Some reactions can become disproportionately emotional or even seem crazy when expectations outrun what a supervised Level 2 system can actually do. Tesla's challenge is managing expectations while marketing a future that has not fully arrived.
For traders, adoption patterns create a measurable gap between technical capability and monetization. Until a large share of human drivers feel comfortable letting FSD handle most of the drive, the software revenue thesis remains partially unproven. Watching subscription take-rate data, if Tesla ever discloses it granularly, would be a more meaningful signal than raw mile counts.
Safety data, controversies, and the AI risk discount
Safety is the most emotionally charged and analytically complex piece of the Tesla FSD story. Markets do not need to resolve the safety debate definitively; they just need to price its probability distribution.
Tesla's periodic Vehicle Safety Report has claimed lower crash rates when autopilot is engaged (source: Tesla). According to Tesla's Q4 2024 report, vehicles with Autopilot active experienced roughly one crash per 5.94 million miles, compared to one crash per 1.08 million miles without Autopilot or active safety features, and a U.S. national average of about one per 0.7 million miles. Fleet data shows AI-assisted driving reduces accident frequencies compared to human drivers in these aggregate statistics.
However, the picture is more complicated:
Tesla's Autopilot has been involved in 51 fatalities as of recent reporting.
Tesla's Autopilot was involved in 736 crashes since 2019.
NHTSA reported 60 crashes involving FSD beta from August 2022 to August 2023.
Tesla vehicles experienced a crash every 3.2 million miles under FSD.
Autopilot's crash rate increased by 59% after Autosteer installation, according to some analyses.
NHTSA opened 30 investigations into Tesla crashes involving autopilot.
Critics, including NHTSA investigators and independent researchers, have argued that Tesla's safety comparison uses flawed methodology. The comparison between Autopilot-on and Autopilot-off driving does not control for road type (Autopilot is used disproportionately on highways, which are inherently safer), weather, time of day, and other factors. Reporting gaps also exist: Tesla's telemetry depends on cellular connectivity, and some accidents may go unrecorded.
AI systems in autonomous vehicles must perform reliably in rare situations as part of safety verification. The challenge is that rare but severe edge-case failures, such as collisions with stationary emergency vehicles or errors in construction zones, matter disproportionately for regulatory and reputational risk even if they are statistically uncommon.
For traders, persistent legal and safety uncertainty tends to maintain a "risk discount" on the FSD/AI option value in TSLA, even when bulls are optimistic about long-term performance improvements.
Regulators, lawsuits, and how legal overhang shapes the FSD story
Regulatory and legal risks are not abstract for Tesla. They directly change the perceived probability of achieving the AI roadmap profitably, and they create headline-driven volatility that traders need to understand.
Major US regulatory actions include:
NHTSA has conducted multiple investigations into autopilot and FSD, resulting in large recall campaigns that modified driver supervision behavior through over-the-air software updates.
The Department of Justice and SEC have probed whether Tesla's marketing overstated autonomy capabilities, creating potential liability exposure.
California DMV threatened Tesla with a 30-day license suspension over concerns about how self driving capabilities were advertised, leading to naming changes and compliance adjustments.
In Europe, regulators in Germany and the UK have raised concerns about the "Autopilot" branding, arguing it could mislead motor vehicles owners into complacency. UNECE regulations impose additional constraints on what automated features can do on European roads, and individual country approvals add further complexity.
Civil lawsuits alleging deceptive FSD marketing have had mixed outcomes. Some cases have been dismissed, while others are heading toward trial or settlement, creating a rolling legal headline risk. Each outcome slightly shifts market expectations about future fines, required design changes, and the speed at which FSD and any robotaxi service can be scaled, as well as how quickly Tesla can secure regulatory approval for broader FSD or robotaxi deployment in new markets.
For traders, every regulatory win or loss recalibrates the perceived timeline and risk premium. A favorable ruling accelerates the bull case; an adverse decision compresses the AI option value. This is why informed traders track regulatory filings and court dates alongside software release notes.
Regional rollout
The FSD story is not uniform globally, and regional differences affect how quickly Tesla can monetize its AI investment. The most advanced FSD (Supervised) releases have historically rolled out first in the US and Canada, where the fleet is largest, the regulatory environment for Level 2 systems is relatively permissive, and Tesla's data infrastructure is most mature.
Europe has moved more slowly, with UNECE rules and country approvals constraining features like automated lane changes and speed control. China follows a distinct path: Tesla relies on local mapping providers and must comply with data-localization rules that keep China-collected data in China, limiting Tesla's ability to pool it with the global dataset. South Korea, Australia, New Zealand, and other Asia-Pacific markets have seen features roll out with region-specific limits. These markets add valuable data diversity but also regulatory complexity that slows the launch cadence. For traders, regional timelines matter because they determine how large the addressable subscription market is at any given time.
Tesla's vision-only bet vs lidar-heavy competitors
Tesla's decision to rely almost entirely on cameras and neural networks sets it apart from virtually every other company pursuing autonomous driving. Understanding this technical bet is important because it directly shapes the investment risk profile.
Tesla Vision uses eight cameras (more on HW4-equipped vehicles) and processes their outputs through a single neural network. Tesla phased out radar and never adopted lidar, based on Elon Musk's argument that human drivers navigate with vision alone, and that a sufficiently capable AI system trained on video can do the same. The car does not carry a lidar sensor, which can cost thousands of dollars per unit and adds hardware complexity.
Competitors like Waymo and the former Cruise operation historically used lidar, radar, HD maps, and cameras together, operating in tightly geofenced zones where every curb and lane marking is pre-mapped. This approach is arguably safer in controlled environments but far more expensive and harder to scale to new cities.
Bulls argue that Tesla's camera-only approach is cheaper per vehicle and more scalable worldwide. If the AI is capable enough, software margins are protected because the hardware cost per car stays low. The fleet can expand to any road the car can physically drive on, rather than being limited to pre-mapped areas.
Critics see it as an unnecessary constraint. Lidar provides direct depth measurement that cameras must infer, and some edge cases, such as distinguishing a white truck from a bright sky, are inherently harder with vision alone. Self driving cars from other companies have achieved driverless operation in limited areas, something Tesla's camera-only system has not yet matched at scale.
Equity markets effectively treat Tesla's "no lidar" policy as a leveraged bet on AI progress: if it works, the upside to margins and software multiples is large, but if vision-only proves insufficient for true autonomy, the FSD option value could shrink sharply. This is a binary risk that keeps TSLA and TSLAx volatile around every major FSD capability update.
Robotaxis and Cybercab: The endpoint of the FSD/AI thesis
The robotaxi is the endgame of the FSD and AI story. If Tesla can deploy autonomous vehicles at scale as a ride-hailing service, the revenue model transforms from selling cars into selling rides and software. Musk first promised coast-to-coast autonomous drives in the late 2010s and, in 2019, claimed Tesla would have a million robotaxis on the road "next year." Neither happened on schedule. More recently, Tesla revealed the "Cybercab," a dedicated, low-cost vehicle designed for robotaxi duty without a steering wheel or pedals, and launched an early, supervised robotaxi service in Austin, Texas, in June 2025, operating in geofenced zones with human safety monitors.
A successful rollout, even in a handful of cities, could turn Tesla from an auto manufacturer into a hybrid of carmaker and mobility platform, with service margins closer to technology firms than to automakers. But much of that value remains speculative, with operations still limited in geography, hours, and regulatory permission. For the routes, permits, fleet size, and service economics, see the dedicated robotaxi and Cybercab article. The key point here is that FSD and AI progress is the technological prerequisite for the robotaxi scenario to play out at all.
How AI progress feeds into Tesla valuation models
Most sophisticated valuation frameworks split Tesla into a "base" EV manufacturer business plus one or more AI options. The base business covers car sales, energy storage, and services. The AI options include FSD software revenue, a potential robotaxi network, and possible licensing of the autonomy stack to other OEMs.
How these scenarios differ is where the debate lives:
Conservative models assign modest probability and low revenue per vehicle from FSD. They might assume a 5 to 10 percent take-rate on subscriptions and no robotaxi revenue in the next five years, resulting in a valuation closer to traditional automaker multiples.
Aggressive AI-centric models assume multi-thousand-dollar annual software ARPU, high take-rates by the early 2030s, and meaningful robotaxi contribution. In these models, Tesla's AI business alone could justify a significant portion of the current market cap.
Market reactions to FSD-related events reveal which model the marginal buyer is pricing. When Tesla announces a major software version launch (for example, V12), a widespread free trial, or an FSD price cut, the stock often moves sharply. These reactions reflect traders re-marking the "AI option value" rather than reassessing the underlying car business.
Conversely, regulatory setbacks, high-profile accidents, or Musk timeline slippage tend to compress the AI option. The insurance risk and legal liability associated with FSD also enter valuation models as potential future costs that could offset software margins.
Tokenized Tesla on BloFin mirrors the underlying TSLA price but does not convey voting or dividend rights. Traders using TSLAx are purely exposed to how equity markets reinterpret the FSD and AI story over time. This makes understanding the valuation mechanism, specifically what drives the AI option up or down, essential for anyone trading the token. For more on the pros and cons of tokenized stock exposure, BloFin Academy has a separate guide.
Why crypto and AI traders watch Tesla's FSD story
Tesla sits at the intersection of multiple themes that are deeply familiar to crypto traders: disruptive technology, AI narratives, high volatility, cult-like retail interest, and sentiment that can shift on a single social media post.
FSD updates, Dojo scaling news, and robotaxi hints often trade like "AI catalysts," producing intraday moves in TSLA that resemble how altcoins respond to mainnet launches or big partnership announcements. The pattern is similar: a narrative-driven asset with a passionate community, limited near-term cash flow proof, and enormous potential upside if the technology delivers. This makes Tesla a natural fit for traders who already navigate the crypto markets on BloFin.
Traders on BloFin can use tokenized Tesla (TSLAx) products, spot or derivatives, to express short-term views on AI-driven catalysts without needing a traditional brokerage account. You can create a new account on BloFin, fund it, and access TSLAx trading alongside your crypto positions, all within one platform. The 24/7 trading availability means you can react to after-hours FSD news or weekend Musk tweets in real time, rather than waiting for NASDAQ to open.
But risk management is critical. The same AI hype that drives upside can magnify drawdowns when timelines slip, safety issues surface, or regulators push back. Sophisticated traders treat Tesla's AI story as a volatile trade, not a guaranteed outcome. Position sizing, stop losses, and awareness of upcoming catalysts (earnings, delivery numbers, software releases, regulatory hearings) are all part of the toolkit.
The bottom line: if you are already comfortable trading high-beta digital assets, Tesla's FSD and AI narrative operates on similar dynamics, just in the equity wrapper.
Timeline recap: Key AI and FSD milestones for context
A quick chronological anchor helps frame how far the technology has come and how many promises remain unfulfilled. In 2014 and 2015, Tesla launched Autopilot HW1 with Mobileye. By 2016, the Mobileye split and the first fatal-crash headlines pushed Tesla to build its own stack. The in-house FSD computer (HW3) shipped in 2019, and FSD Beta reached a select group of North American drivers in 2020. AI Day in 2021 introduced Dojo, and updates followed in 2022. Through 2023 and 2024, FSD V11 and V12 unified highway and city-street driving and completed the shift to end-to-end neural networks, and in April 2024 the system became "FSD (Supervised)." In 2025 and 2026, FSD expansion reached parts of Europe and Asia, Tesla launched supervised robotaxi pilots in Austin, wound down the Dojo program in favor of new in-house AI chips, and retired the Autopilot name while renaming the onboard hardware the "AI Computer." The consistent pattern is substantial technical progress paired with timelines that were almost always more optimistic than reality, and that is essential context for anyone trading on future AI promises.
Open questions: What the market still doesn't know about Tesla's AI future
Several unresolved questions will drive future repricing of TSLA and TSLAx. Traders do not need answers today, but awareness of these uncertainties helps distinguish thesis-changing news from incremental noise.
Regulatory unknowns: When, if ever, will regulators classify any Tesla system above SAE Level 2? What will the liability framework look like when a car operating under FSD supervision causes an accident? Will the human driver remain solely responsible, or will automaker liability increase? These questions affect insurance costs, risk premiums, and deployment speed.
Economic unknowns: What is the long-term FSD subscription pricing power? Will customers pay $99 per month indefinitely, or will competitive pressure from other self driving cars and ride-hailing alternatives compress prices? What are typical robotaxi utilization hours and maintenance costs? How will capex for AI compute and data center infrastructure balance against software-like gross margins?
Technical unknowns: At what rate can end-to-end neural networks reduce long-tail driving errors compared with human drivers across varied road networks in different countries? Can a camera-only system handle severe weather, poorly marked roads, and rare edge cases as well as lidar-equipped competitors? Will HW3 vehicles be capable of running future software versions, or will millions of older cars be left behind?
Licensing and strategy: Will Tesla license FSD to other OEMs, creating a platform revenue stream, or keep it exclusive to Tesla vehicles?
Traders who can accurately assess when new information genuinely shifts probabilities on these questions, rather than just generating noise, will have an edge in trading TSLA and TSLAx around AI catalysts.
The bottom line
Tesla's FSD and AI story is one of the most debated narratives in global markets, and that is exactly what makes it tradeable. The base business sells cars; the premium in the stock prices a future that is real in direction but genuinely uncertain in pace. Progress since the Mobileye days has been substantial, yet almost every timeline has arrived later than promised, and the hard questions about safety, regulation, and camera-only autonomy remain open. Informed traders do not treat the outcome as guaranteed in either direction. They focus on how each new milestone, whether technical, regulatory, or financial, shifts the probabilities in their own models, and they size TSLAx positions for volatility rather than certainty. The AI story is not settled, and an unsettled story is one you can trade.
Frequently asked questions
Is Tesla's Full Self-Driving actually fully self-driving?
No. Despite the name, FSD is branded "Full Self-Driving (Supervised)" and is classified at SAE Level 2, meaning it assists the driver but requires an attentive human with hands ready on the wheel and eyes on the road. The driver remains legally responsible for the vehicle at all times.
How does FSD news affect tokenized Tesla (TSLAx) on BloFin?
TSLAx mirrors the underlying TSLA equity price, so an FSD headline that moves the stock, such as a software release, a regulatory ruling, or a safety incident, also moves the token. Because BloFin trades around the clock, you can react to after-hours or weekend FSD news rather than waiting for the Nasdaq open.
Why does Tesla trade with an AI premium instead of an automaker multiple?
Markets price two businesses in one stock: a base EV maker plus an AI option covering FSD software revenue, a potential robotaxi network, and possible autonomy licensing. When traders believe in the AI second act, they apply software-like multiples; when they doubt it, the stock drifts toward traditional carmaker valuations. That tension is why the stock moves sharply on AI news.
What happened to Tesla's Dojo supercomputer?
Tesla wound down the Dojo program in 2025, with Musk calling the next-generation design an evolutionary dead end, and disbanded the team. Tesla redirected its effort toward new in-house AI chips, AI5 and AI6, made by external foundries, while continuing to run large Nvidia-based training clusters. The heavy AI-compute spending, and its impact on margins and competitive position, remains a live part of the valuation debate.
Why is Tesla's vision-only approach considered a risk?
Tesla relies on cameras and neural networks rather than lidar, betting that a capable-enough AI trained on video can match or beat human vision. If it works, hardware costs stay low and the system scales to any road; if vision-only proves insufficient for true autonomy, the FSD option value could shrink sharply. Equity markets treat this as a leveraged bet on AI progress, which keeps TSLA and TSLAx volatile around major capability updates.
Researched and written by the BloFin Academy editorial team with AI-assisted drafting, reviewed for accuracy against public sources at the time of publication. Figures such as FSD pricing, fleet mileage, regulatory status, and robotaxi availability change frequently; verify current details against a live source before trading.
Disclaimer: This article is for educational purposes only and is not financial, investment, or trading advice. Tokenized Tesla (TSLAx) mirrors the price of TSLA but does not grant shareholder rights, voting power, or dividend entitlements. Trading digital assets and derivatives involves substantial risk, including the possible loss of your entire investment. Always do your own research and consider your risk tolerance and time horizon before trading.