The cars were still in the room, but they had become furniture. On the latest Tesla earnings call, Elon Musk spent more time on the phrase “physical AI” than on the vehicle production numbers that once anchored every conversation. The press release still begins with deliveries, but the voice on the call does not. Crypto Briefing caught the change with a perfectly understated headline: “Elon Musk's Tesla earnings calls are now AI and robotics presentations with a side of cars.” The line is accurate. It is also incomplete, because a side order implies that the main course is obvious. It is not. The main course is a narrative, and narratives require a different kind of auditing.
The audit is not a check; it is a confession. I wrote that sentence after months of reading smart contracts that were technically correct and spiritually empty. It returns to me now because Tesla’s earnings call has become a confession about the future of the company after the car business stops being enough. I have seen this shape before. In 2017, in Zurich, I audited a DAO successor during the ICO boom and found a reentrancy vulnerability that exposed about 500 ETH, roughly $2.1 million at the time. My report was rejected by the frontend team as “too academic.” The code was vulnerable, but the narrative was so strong that nobody wanted to read the audit. That taught me two lessons that frame this analysis: technical correctness cannot save a broken story, and a strong story cannot save broken technicals for long.
Tesla’s AI story is the strongest story in the automotive industry right now. The question is whether the technicals can carry it.
Context: When the Main Course Becomes the Side Dish
Let me first credit what Crypto Briefing got right. The earnings call no longer functions as a quarterly report on car sales; it functions as a roadmap demonstration for artificial intelligence and robotics. This is not a cosmetic change. It is a change in Tesla’s value creation hypothesis. Value will no longer come primarily from selling vehicles; it will come from intelligence that has learned to move through physical space. The vehicle becomes the chassis; the robot becomes the body; the data center becomes the brain.
The financial pressure behind this pivot is real. Tesla initiated a global price war in 2023, and the result was predictable: gross margins fell from the high twenties in 2022 to roughly seventeen or eighteen percent by 2024. Electric-vehicle competition from China, Europe, and even legacy Detroit has compressed the growth premium. A company that was once valued as the only electric car company with scale is now one of many electric car companies with scale. When a company can no longer convince the market that its future belongs to vehicles, it must convince the market that the future belongs to machines that are not vehicles at all.
This is where “the side of cars” becomes the story. From the outside, the earnings call looks like Musk interrupting a car update with robot talk. From the inside, it looks like a capital allocation decision. The management team is telling the market where priority lives. If FSD, Dojo, Optimus, and Robotaxi are the four pillars of the future, the cars are the ground on which those pillars are being built. The ground matters. But the architecture is the point.
A brief history of the narrative makes the shift visible. In 2016, Tesla began selling every new vehicle with the hardware it claimed would one day enable full autonomy. In 2019, Musk promised a robotaxi network “next year.” In 2021, the Tesla Bot appeared on stage as a person in a suit. In 2022, Optimus walked awkwardly across a stage on its own. In 2024, the Cybercab was unveiled without a steering wheel. Each step moved the center of gravity further away from the vehicle and closer to the intelligence that moves the vehicle. The milestone that mattered in 2025 is not a factory opening. It is the moment a neural network takes a decision that a previous version of the software could not.
That is why the source article’s brevity is itself information. Crypto Briefing did not provide detailed technical metrics because its audience does not need them. The crypto market has spent years learning to trade narratives before fundamentals arrive. Tesla has now entered that same market logic. The earnings call is priced not as a car report but as an option on physical AI. The side of cars is the underlying asset; the option is the story.
Core I: Four Projects, Four Tenses
The first thing a forensic reader notices is that the earnings call treats all four AI-adjacent projects as if they were equally mature. They are not. The gap between them is so large that the presentation is not merely optimistic; it is a structural compression of time.
FSD is the present tense, with a supervisor in the passenger seat. Since V12, Tesla’s Full Self-Driving stack has moved decisively toward end-to-end neural networks. Instead of hand-coded rules, the vehicle learns to map visual inputs to driving decisions. This is a real architectural revolution. It is also a system that still requires a human to stay alert. Tesla calls it FSD Supervised, and that single word, supervised, is doing more work than any other word on the earnings call. FSD is the foundation of Tesla’s claim to be an AI company, but the claim is not yet complete. It has passed through the gate from demonstration to deployment; it has not passed through the higher gate from supervised to unsupervised.
Optimus is the future conditional, dressed in prototype optimism. The robot has evolved from the 2022 concept figure to a machine that can fold a shirt, pick up a battery cell, and move through a warehouse in limited ways. Those are meaningful engineering steps. They are not production metrics. The object that appears on the earnings call slide has a unit economics story, a target price near twenty to thirty thousand dollars, and a long-term demand hypothesis measured in billions of units. But the difference between a prototype and a product is not a marketing slide; it is the cost curve, the failure rate, and the safety case. Tesla has manufacturing scale, and that is not nothing. But humanoid robots in unstructured homes and factories require a level of robust generalization that current robotics has not yet achieved. The tense of Optimus is future conditional, and the earnings call tends to pronounce it as present.
Dojo is the present perfect continuous: we have been training, we are training, we will train. Tesla’s decision to build its own D1 chip and Dojo supercomputer was the right long-term move. Renting compute from a competitor is a strategic weakness when the entire business hypothesis depends on AI. But Dojo has not yet replaced NVIDIA in Tesla’s data center. Tesla continues to buy NVIDIA GPUs in massive volumes, which is the clearest evidence that Dojo’s first generation did not close the cost-performance gap. The earnings call tends to say “Dojo is ramping.” The market hears “AI infrastructure.” A forensic reader hears “we have not been able to stop buying NVIDIA.”
Cybercab is the future perfect, with a regulatory asterisk. The robotaxi has no steering wheel and no pedals. That design is the clearest expression of Tesla’s intent. It is also the clearest example of Musk Time Dilation, the phenomenon in which a promised schedule stretches by one to three years as it passes through the gravity well of engineering and regulation. The United States Federal Motor Vehicle Safety Standards do not currently permit a passenger vehicle without a steering wheel to be sold. Congress could legislate an exemption; the National Highway Traffic Safety Administration could grant a waiver; neither has happened. Tesla plans to launch a supervised robotaxi service in Texas and California in 2025 using vehicles that already have controls, before Cybercab volume begins around 2026. This is a coherent path. It is not a short path.
In the code, I found the ghost of the architect. That line has followed me through more protocol audits than I can count. It applies here. The architect behind Tesla’s physical-AI shift is not just Elon Musk; it is the economics of a car company that has to grow into something else. The ghost lives in every slide where FSD borrows credibility from Optimus, where Optimus borrows manufacturing credibility from the car business, and where Dojo borrows relevance from both. The architecture is elegant. The timeline is not.
Core II: The Financial Algorithm Behind the Presentation
If the earnings call is a product roadmap, the underlying algorithm is a series of discounted future narratives. Let me restate the obvious, because it is rarely said directly: Tesla’s current income is still dominated by cars. Perhaps eighty percent or more of revenue comes from selling hardware with wheels. Yet the market’s willingness to assign a technology multiple depends on the company’s ability to make that hardware seem like the least interesting part of the model.
The result is a financial game with three acts. The first act is FSD software. Subscriptions and one-time purchases create a high-margin cash flow layer that has already started to appear. It is not enough to replace car gross profit, but it is enough to show that the hardware can become a software distribution channel. The second act is Robotaxi. If a robotaxi network can operate at a lower marginal cost than human-driven networks, it creates a new revenue curve independent of vehicle sales. The third act is Optimus. That is the “so large it cannot be modeled” act, and it is precisely why the earnings call is willing to be vague about near-term margins. The larger the third act, the less the current quarter matters, until it does.
This is where the phrase “narrative compression” becomes useful. An earnings call can place FSD, Optimus, Dojo, and Cybercab on the same visual plane, and the market will instinctively aggregate their maturity. FSD’s on-road data gives credibility to Optimus’s future dexterity. Optimus’s long-term market size gives permission to spend on Dojo. Dojo gives Tesla the right to call itself an AI company before FSD has reached Level 4. Each project borrows time from the others. None of the loans are visible in the financial statements.
I saw the same compression in 2020, when I was modeling liquidity protocols for a crypto venture fund in Singapore. A token could be attached to a governance system, a liquidity incentive, and a roadmap in a way that made the protocol feel more complete than it was. The market did not care about the difference between “the code is deployed” and “the code is safe.” It only cared about the intensity of the story. When the pool emptied, only the intent remained. For Tesla, the pool is investor capital, and the intent is physical AI. The compression is not fraud. It is a narrative tax on future timelines, paid in advance.
“To own a piece of art is to inherit its narrative.” I wrote that in an NFT essay in 2021. To own Tesla stock is the same. The equity certificate is not a claim on a factory; it is a claim on a story about a future in which machines learn to move through physical space. The story has been extraordinarily well constructed. The audit is not a check; it is a confession. The earnings call confesses something the balance sheet cannot say: the car business, on its own, is no longer enough. That is why the narrative shift is not merely marketing. It is capital markets survival.
Core III: The Competitive Matrix No One Wants to Read
The other reason the earnings call has become an AI presentation is that Tesla is no longer alone in any of its chosen futures. Every pillar has a specialized competitor with demonstrable advantages.
Waymo is the uncomfortable mirror. Alphabet’s Waymo has already crossed the threshold that Tesla is still approaching: it operates true driverless rides with no safety driver in San Francisco, Phoenix, and Los Angeles, crossing more than one hundred thousand paid trips a week. Waymo chose the map-first, sensor-heavy path; Tesla chose the vision-only, data-economy path. Tesla’s approach may be cheaper in the long run, but cheapness is not the same as certification. The gap between supervised FSD and unsupervised autonomy is exactly the gap between “almost ready” and “operational reality.”
Optimus faces a similarly crowded field. Figure AI has received major funding and shown large-language-model-driven interactions. Boston Dynamics, now under Hyundai, spent decades solving bipedal locomotion. Chinese humanoid-robot companies are moving fast, often with government support. Tesla’s edge is manufacturing. If anyone can mass-produce a humanoid at a price of twenty to thirty thousand dollars, Tesla is a plausible candidate. But a price target is not a product roadmap. The robot’s dexterity, reliability, and safety remain open questions.
Dojo faces the most unforgiving competitor: NVIDIA. The H100 and its successors are not just chips; they are a full software and networking ecosystem. Dojo is a statement of independence. It is also a testimony to how difficult it is to exit the NVIDIA ecosystem once your training framework is embedded in it. Tesla can win this race by building specialized silicon for physical AI, but the unit cost of that victory is enormous, and the earnings call will not reveal it.
China is the quiet fourth competitor. Huawei’s ADS and Xpeng’s XNGP are close enough to FSD in user experience that the phrase “Tesla AI advantage” in China is no longer self-evident. Domestic Chinese systems understand local driving behavior, local road networks, and local regulatory expectations better than a system trained largely in North American conditions. FSD’s entry into China is a real opportunity, but the data-compliance burden is severe, and the competitive lead is narrower than the Western narrative suggests.
This is why the earnings call matters. It repositions Tesla inside the “AI stock” theme rather than the “legacy auto” theme. The moment the market classifies Tesla as an AI company, its competitors become NVIDIA, OpenAI, and Waymo, not Ford, Toyota, and Volkswagen. That classification change is worth more than a single quarter of car sales.
Identity is a protocol; soul is the private key. Tesla is updating its identity protocol. The soul, the ability to execute on a timeline without losing safety or trust, remains the private key.
Core IV: The Governance Blind Spot
The most uncomfortable part of the Tesla AI narrative is not the technology. It is the governance structure around the technology. Elon Musk is simultaneously the chief executive of Tesla and the founder of xAI, which built a massive data center called Colossus with more than one hundred thousand GPUs. There have been public reports of resources, including NVIDIA chips originally intended for Tesla, being redirected to xAI. Whether or not the details are exact, the structural conflict is real.
For a company that claims to be a pure AI enterprise, this is a unique governance discount. OpenAI and Anthropic do not have to ask their shareholders whether they are also paying for the compute of another company. Tesla does. Every time an earnings call mentions AI and robotics, the next question should be about resource allocation between Tesla and xAI. The call usually answers with the word synergies. A forensic reader should hear the word conflict.
In 2017, I learned that the division between code and human intent is where protocols fail. The same is true here. The code may be beautiful. The vision may be coherent. But the allocation of compute, talent, and attention across Musk’s companies is an unresolved variable, and the market is correct to discount it.
Core V: The Sentiment Reading
There is also a softer signal buried in the transcript. When a chief executive spends more time describing a robot than a car, he is telling investors which emotions he wants them to feel. Cars are becoming commoditized; they provoke comparisons of price, range, and charging speed. Robots are not yet commoditized; they provoke wonder, fear, and the hope of a different future. The earnings call is deliberately shifting the emotional register of the company.
This is why retail sentiment around Tesla has become so sensitive. The stock is one of the most heavily traded equities in the United States, and its price moves are not always proportionally connected to quarterly financials. The market is trading the intensity of the narrative, not the discounted value of near-term cash flows. When the narrative is strong, FSD misses are forgiven. When the narrative weakens, even a good car quarter is not enough.
Crypto Briefing’s decision to cover Tesla from a crypto-native angle reinforces this point. The publication is not primarily staffed by automotive journalists. It is staffed by people who have watched tokens rise and fall on the strength of a roadmap. Tesla is now being read through that same lens. A car company that presents itself as an intelligence company is no longer judged by miles per gallon. It is judged by whether the story can be repriced before the next deadline.
Contrarian: The Risk Is Not Overpromising. It Is Overcoherence.
The standard criticism of Tesla’s AI roadshow is that Musk overpromises. The timeline slips. The unsupervised version of FSD has been around one year away for almost a decade. Therefore, the story is unreliable.
I want to push against that. The more interesting risk is not that the story is false; it is that the story is too coherent. The market is not stupid. It knows the difference between a car company and an AI company. The price action suggests that investors are not being deceived; they are buying an option on physical AI. The earnings call is the option prospectus.
If that is true, then the danger is not narrative hype. The danger is narrative correlation. By linking every project to the same physical-AI thesis, Tesla has removed the firewall between its separate businesses. A delay in FSD does not just hurt FSD. It weakens Optimus, because Optimus borrows perceptual intelligence from FSD. A delay in Dojo does not just hurt training efficiency. It weakens the entire claim to AI independence. When the pool empties, only the intent remains. And when the earnings call is the pool, an empty promise can drain the whole company at once.
The other contrarian layer is about the market’s relationship with time. Conventional analysts complain that Musk is always late. The market already knows that. The real risk is that he will be late in the wrong order. If FSD reaches unsupervised autonomy but Optimus remains a museum piece, the physical-AI thesis survives. If Optimus works but Cybercab is blocked by regulation, the thesis survives. But if the deadlines begin to fail in sequence, the market will stop distinguishing between the pillars. It will price Tesla as a company that cannot ship any of its futures. That is the deeper danger of a narrative this coherent: it leaves no room for gradual disappointment.
In crypto, I watched communities maintain a narrative long after the technical foundation shifted. They did not fail because the story was weak. They failed because the story was too strong to allow course correction. Tesla is not a crypto token, but it has adopted the same architecture of belief. The earnings call is the ritual that renews that belief. When the ritual stops working, the repricing will be violent.
Takeaway: Buy the Milestones, Not the Slides
The question for the next twelve to twenty-four months is not whether Tesla is an AI company. It is whether the pillars of that identity can be independently verified. The earnings call will not provide that verification. It will continue to compress timelines and aggregate maturity. The data that matters lives outside the presentation.
Watch three things. First, the administrative process for the Cybercab: a change in the Federal Motor Vehicle Safety Standards is a necessary condition, and its absence will be a definitive negative. Second, the safety case for unsupervised FSD: Tesla must eventually publish accident rates with credible confidence intervals, not anecdotal videos. Third, Optimus’s hours of continuous operation and manufacturing cost curve: a prototype that can fold a shirt is a museum piece. A robot that can work a full shift in a factory is a business.
The car business is not dead. It is the foundation. But a foundation is not a building. Tesla’s earnings call has become a confession, and in that confession, the cars are the part of the story that no longer needs to be said. The side of cars is still on the plate. Yet the more time Musk spends on AI and robotics, the more the market will demand evidence that the timeline is not a narrative collage. When the pool empties, only the intent remains. That is the lesson I carried out of Zurich, through DeFi Summer, through the NFT crash, and into the present. The intent is beautiful. The timeline is where the audit begins.