Telsa isn't just a car company - it's an undervalued AI empire
Elon Musk controls an AI and robotics powerhouse - and its growth isn't fully valued into Tesla's stock price. John D. Rockefeller controlled oil - from wellhead to gas pump. Elon Musk controls AI - from training cluster to robot hand.
Elon Musk is the richest person in history. He got there the same way John D. Rockefeller did.
Rockefeller didn't build his fortune by finding more oil than anyone else. Oil was everywhere. Any idiot with a shovel and a dream could find oil. Pennsylvania was lousy with wildcatters, all of them convinced they'd struck it rich. Most of them had. The problem was that "rich" didn't stay rich for long when everyone else was rich too.
Rockefeller understood something his competitors didn't. The money wasn't in the oil. The money was in everything between the oil and the customer. So he bought the pipelines. He bought the refineries. He bought the railroad cars and negotiated rebates that made his shipping costs a fraction of everyone else's.
By the time his competitors got their oil out of the ground, Rockefeller owned every inch of infrastructure between the wellhead and the kerosene lamp.
Rockefeller's Standard Oil didn't compete. It collected tolls.
Musk seems to be running the same playbook for artificial intelligence. And Wall Street remains divided on what his Tesla $(TSLA)$ actually is. Bulls like Wedbush Securities analyst Dan Ives call Tesla "the most undervalued AI play in the market today."
Bears like GLJ Research's Gordon Johnson argue that Tesla is "just a car company" with an inflated multiple. For most of the past decade, the bears set the consensus. They're wrong.
I've spent years watching Tesla from the sidelines, convinced the valuation was insane. It was insane - for a car company. But for the only publicly traded entry point into a vertically integrated AI empire? Maybe not.
I own the stock now.
The receipts
When people tell you something is impossible, what they usually mean is impossible for them.
Here's what Musk has actually built.
Jensen Huang, Nvidia's CEO, said something remarkable about Musk on the "Bg2 Pod" podcast in October 2024. Nvidia makes the chips that power virtually every AI system on the planet. The company is worth about $4.5 trillion. Huang has been running Nvidia since co-founding it 32 years ago. When he talks about who's building what in AI, he's not speculating. He's reading the purchase orders.
Huang's assessment of Musk: "As far as I know, there's only one person in the world who could do that. Elon is singular in his understanding of engineering and construction and large systems and marshaling resources. It's just unbelievable."
What prompted that praise? Musk's xAI's Colossus data center went from empty dirt in Memphis, Tenn., to 100,000 Nvidia GPUs operational in 19 days. Huang said this process typically takes several years.
Several years is normal. Musk did it in 19 days. That's not a rounding error. That's not even the same sport. That's a different species of execution.
Likewise, Tesla's Shanghai Gigafactory went from breaking ground on Jan. 7, 2019 to producing cars in under 11 months. The automotive industry said it couldn't be done.
Industry standard for a new factory is four to five years, minimum. By December 2019, Tesla Model 3s were rolling off the line. When people tell you something is impossible, what they usually mean is impossible for them.
Gavin Baker, a hedge-fund manager who has covered Nvidia for more than 20 years and is considered one of the most technically informed investors in AI, explained the challenge on Patrick O'Shaughnessy's "Invest Like the Best" podcast:
"Going from Hopper to Blackwell, first you go from air-cooled to liquid-cooled. The rack goes from weighing round numbers 1,000 pounds to 3,000 pounds. Goes from round numbers 30 kilowatts, which is 30 American homes, to 130 kilowatts, which is 130 American homes."
Baker offered an analogy that stuck with me: "Imagine if to get a new iPhone you had to change all the outlets in your house to 220 volt, put in a Tesla Powerwall, put in a generator, put in solar panels, put in a whole home humidification system, and then reinforce the floor."
That's Blackwell. You essentially have to rebuild your entire data center.
This is why deployment speed matters more than money.
Alphabet's $(GOOG)$ $(GOOGL)$ Google has infinite money. Microsoft $(MSFT)$ has infinite money. But money doesn't build data centers. Execution builds data centers. And Huang, who sees every customer's timeline, says no one executes faster than Musk.
Here's the number that should get your attention.
Here's the number that should get your attention: xAI announced recently that it ended 2025 with more than 1 million H100 GPU equivalents across its Colossus I and II supercomputers. The company just raised $20 billion in Series E funding backed by Nvidia and Cisco Systems $(CSCO)$ to expand that infrastructure.
Musk has publicly spoken about his companies "converging" on AI. One ecosystem. One infrastructure. One entry point for public investors.
Whoever deploys first, trains first. Whoever trains first, releases the best model first. Better models attract more users. More users generate more data. More data trains better models. In AI, first-mover advantage isn't a luxury. It's a moat.
The data advantage nobody talks about
Tesla has a data "moat" hiding in plain sight that almost nobody is pricing.
Every Tesla on the road is a data-collection device. The cameras aren't just for safety. They're training AI systems in real-time. According to Tesla's FSD safety page, the fleet currently has accumulated close to 7.3 billion miles of real-world driving data, with 2.63 billion of those miles on city streets. Musk has acknowledged that Tesla needs roughly 10 billion miles of training data before he's confident in unsupervised FSD.
A caveat: Not every Tesla buyer has FSD. The take rate is about 12% of the fleet as of the October earnings call. And Tesla's system remains SAE Level 2 (supervised), while Waymo operates at Level 4 (fully autonomous).
But here's what matters for the AI training argument: Waymo, Alphabet's self-driving subsidiary with a 15-year head start, recently expanded to highway driving. Its fleet of roughly 1,500 to 2,500 vehicles operates in controlled, geofenced environments in a handful of cities.
Tesla has more than 5 million vehicles on the roads collecting data. Every Tesla sold is a data-collection device that customers pay for. Waymo has to buy its own cars and hire safety drivers.
The systems are different. The data advantage is real.
There's also what engineers call the "checkpoint problem." Baker explained it: "Every one of those labs has a more advanced checkpoint internally of the model. They're using that model to train the next model. And if you do not have that latest checkpoint, you're behind. It's getting really hard to catch up."
Tesla's checkpoints contain billions of miles of learning. A competitor starting today begins at checkpoint zero. They're not behind by months or years. They're behind by miles. Billions of them.
Rockefeller's stack, updated
Rockefeller's genius was vertical integration. Own every layer of the stack, and your competitors become tenants in your building. They can drill all the oil they want. If you own the only pipeline to market, they're paying you either way.
Let's map Musk's empire the same way. Rockefeller owned oil wells; Musk has xAI, a frontier AI lab. Rockefeller owned pipelines; Musk has Starlink, a 9,500-satellite global network. Rockefeller owned refineries; Musk has Colossus, already at 200,000 GPUs and expanding to 1 million.
Rockefeller owned distribution networks; Musk has Tesla's 5.1 million vehicles and Optimus robots. Rockefeller used oil profits to fund expansion; Musk uses Tesla's cash flow. In the third quarter of 2025, Tesla generated almost $4 billion in free cash flow, a new quarterly record. Total cash and investments: $41.6 billion. That's the war chest funding the AI empire.
Training data comes from X, formerly Twitter, with 600 million users generating real-time content that no competitor can replicate through web scraping. Training compute comes from Colossus. Global distribution comes from Starlink. Physical deployment comes through Tesla's vehicles, robots and millions of touchscreens already in customers' hands.
The parallel to Standard Oil is exact. Rockefeller controlled oil from wellhead to gas pump. Musk controls AI from training cluster to robot hand. And like Rockefeller, he's built it while everyone was focused on the obvious business.
Why Wall Street keeps missing this
Jordi Visser, former CIO of Weiss Multi-Strategy Advisors, coined a useful term: the "Musk discount." When Warren Buffett makes a prediction, the market reprices immediately. Analysts nod sagely. CNBC runs segments. When Musk makes the same prediction, analysts roll their eyes and mutter about missed timelines.
The pattern is remarkably consistent: Musk is optimistic on timing - but accurate on direction.
Musk said electric vehicles would dominate transportation. Analysts laughed. EVs now command most major automakers' strategy. He said reusable rockets were possible. Aerospace engineers scoffed. SpaceX lands them routinely and has made NASA a customer. He said neural networks could drive cars. Skeptics pointed to edge cases. Tesla's Full Self-Driving is on public roads in multiple countries.
Visser nails the asymmetry: "If Musk is early, these sectors still compound. If he is on time, they rerate. If he is late, they eventually rerate. In every version, the asymmetry works in the investor's favor."
Wall Street has a temporal bias. If something won’t show up in next quarter’s earnings call, it functionally doesn’t exist. But competitive moats aren’t built in quarters. They’re built over years, quietly, while analysts debate delivery numbers and gross margins.
The stock catalyst has a date
Nvidia’s Blackwell chips began shipping in volume in the first quarter of 2025, and xAI, with its substantial allocation, will be among the first to deploy them at scale. When xAI releases the first frontier model trained primarily on Blackwell architecture, the market will be forced to re-evaluate where Tesla sits in the AI hierarchy.
This isn’t speculation about some distant, hazy future. It’s a catalyst with a date. The models trained on that infrastructure will emerge by the middle of this year. And when they do, the question “Is Tesla an AI company?” will answer itself.
The receipts
When people tell you something is impossible, what they usually mean is impossible for them.
Here's what Musk has actually built.
Jensen Huang, Nvidia's (NVDA) CEO, said something remarkable about Musk on the "Bg2 Pod" podcast in October 2024. Nvidia makes the chips that power virtually every AI system on the planet. The company is worth about $4.5 trillion. Huang has been running Nvidia since co-founding it 32 years ago. When he talks about who's building what in AI, he's not speculating. He's reading the purchase orders.
Huang's assessment of Musk: "As far as I know, there's only one person in the world who could do that. Elon is singular in his understanding of engineering and construction and large systems and marshaling resources. It's just unbelievable."
What prompted that praise? Musk's xAI's Colossus data center went from empty dirt in Memphis, Tenn., to 100,000 Nvidia GPUs operational in 19 days. Huang said this process typically takes several years.
Several years is normal. Musk did it in 19 days. That's not a rounding error. That's not even the same sport. That's a different species of execution.
Likewise, Tesla's Shanghai Gigafactory went from breaking ground on Jan. 7, 2019 to producing cars in under 11 months. The automotive industry said it couldn't be done.
Industry standard for a new factory is four to five years, minimum. By December 2019, Tesla Model 3s were rolling off the line. When people tell you something is impossible, what they usually mean is impossible for them.
Gavin Baker, a hedge-fund manager who has covered Nvidia for more than 20 years and is considered one of the most technically informed investors in AI, explained the challenge on Patrick O'Shaughnessy's "Invest Like the Best" podcast:
"Going from Hopper to Blackwell, first you go from air-cooled to liquid-cooled. The rack goes from weighing round numbers 1,000 pounds to 3,000 pounds. Goes from round numbers 30 kilowatts, which is 30 American homes, to 130 kilowatts, which is 130 American homes."
Baker offered an analogy that stuck with me: "Imagine if to get a new iPhone you had to change all the outlets in your house to 220 volt, put in a Tesla Powerwall, put in a generator, put in solar panels, put in a whole home humidification system, and then reinforce the floor."
That's Blackwell. You essentially have to rebuild your entire data center.
This is why deployment speed matters more than money.
Alphabet's (GOOG) (GOOGL) Google has infinite money. Microsoft (MSFT) has infinite money. But money doesn't build data centers. Execution builds data centers. And Huang, who sees every customer's timeline, says no one executes faster than Musk.
Here's the number that should get your attention.
Here's the number that should get your attention: xAI announced recently that it ended 2025 with more than 1 million H100 GPU equivalents across its Colossus I and II supercomputers. The company just raised $20 billion in Series E funding backed by Nvidia and Cisco Systems (CSCO) to expand that infrastructure.
Musk has publicly spoken about his companies "converging" on AI. One ecosystem. One infrastructure. One entry point for public investors.
Whoever deploys first, trains first. Whoever trains first, releases the best model first. Better models attract more users. More users generate more data. More data trains better models. In AI, first-mover advantage isn't a luxury. It's a moat.
The data advantage nobody talks about
Tesla has a data "moat" hiding in plain sight that almost nobody is pricing.
Every Tesla on the road is a data-collection device. The cameras aren't just for safety. They're training AI systems in real-time. According to Tesla's FSD safety page, the fleet currently has accumulated close to 7.3 billion miles of real-world driving data, with 2.63 billion of those miles on city streets. Musk has acknowledged that Tesla needs roughly 10 billion miles of training data before he's confident in unsupervised FSD.
A caveat: Not every Tesla buyer has FSD. The take rate is about 12% of the fleet as of the October earnings call. And Tesla's system remains SAE Level 2 (supervised), while Waymo operates at Level 4 (fully autonomous).
But here's what matters for the AI training argument: Waymo, Alphabet's self-driving subsidiary with a 15-year head start, recently expanded to highway driving. Its fleet of roughly 1,500 to 2,500 vehicles operates in controlled, geofenced environments in a handful of cities.
Tesla has more than 5 million vehicles on the roads collecting data. Every Tesla sold is a data-collection device that customers pay for. Waymo has to buy its own cars and hire safety drivers.
The systems are different. The data advantage is real.
There's also what engineers call the "checkpoint problem." Baker explained it: "Every one of those labs has a more advanced checkpoint internally of the model. They're using that model to train the next model. And if you do not have that latest checkpoint, you're behind. It's getting really hard to catch up."
Tesla's checkpoints contain billions of miles of learning. A competitor starting today begins at checkpoint zero. They're not behind by months or years. They're behind by miles. Billions of them.
Rockefeller's stack, updated
Rockefeller's genius was vertical integration. Own every layer of the stack, and your competitors become tenants in your building. They can drill all the oil they want. If you own the only pipeline to market, they're paying you either way.
Let's map Musk's empire the same way. Rockefeller owned oil wells; Musk has xAI, a frontier AI lab. Rockefeller owned pipelines; Musk has Starlink, a 9,500-satellite global network. Rockefeller owned refineries; Musk has Colossus, already at 200,000 GPUs and expanding to 1 million.
Rockefeller owned distribution networks; Musk has Tesla's 5.1 million vehicles and Optimus robots. Rockefeller used oil profits to fund expansion; Musk uses Tesla's cash flow. In the third quarter of 2025, Tesla generated almost $4 billion in free cash flow, a new quarterly record. Total cash and investments: $41.6 billion. That's the war chest funding the AI empire.
Training data comes from X, formerly Twitter, with 600 million users generating real-time content that no competitor can replicate through web scraping. Training compute comes from Colossus. Global distribution comes from Starlink. Physical deployment comes through Tesla's vehicles, robots and millions of touchscreens already in customers' hands.
The parallel to Standard Oil is exact. Rockefeller controlled oil from wellhead to gas pump. Musk controls AI from training cluster to robot hand. And like Rockefeller, he's built it while everyone was focused on the obvious business.
Why Wall Street keeps missing this
Jordi Visser, former CIO of Weiss Multi-Strategy Advisors, coined a useful term: the “Musk discount.” When Warren Buffett makes a prediction, the market reprices immediately. Analysts nod sagely. CNBC runs segments. When Musk makes the same prediction, analysts roll their eyes and mutter about missed timelines.
The pattern is remarkably consistent: Musk is optimistic on timing — but accurate on direction.
Musk said electric vehicles would dominate transportation. Analysts laughed. EVs now command most major automakers’ strategy. He said reusable rockets were possible. Aerospace engineers scoffed. SpaceX lands them routinely and has made NASA a customer. He said neural networks could drive cars. Skeptics pointed to edge cases. Tesla’s Full Self-Driving is on public roads in multiple countries.
Visser nails the asymmetry: “If Musk is early, these sectors still compound. If he is on time, they rerate. If he is late, they eventually rerate. In every version, the asymmetry works in the investor’s favor.”
Wall Street has a temporal bias. If something won’t show up in next quarter’s earnings call, it functionally doesn’t exist. But competitive moats aren’t built in quarters. They’re built over years, quietly, while analysts debate delivery numbers and gross margins.
The stock catalyst has a date
Nvidia's Blackwell chips began shipping in volume in the first quarter of 2025, and xAI, with its substantial allocation, will be among the first to deploy them at scale. When xAI releases the first frontier model trained primarily on Blackwell architecture, the market will be forced to re-evaluate where Tesla sits in the AI hierarchy.
This isn't speculation about some distant, hazy future. It's a catalyst with a date. The models trained on that infrastructure will emerge by the middle of this year. And when they do, the question "Is Tesla an AI company?" will answer itself.
The investment thesis
The car business isn't the point. The car business is the cash register that funds the empire.
Here's the uncomfortable reality for investors who want exposure to Musk's AI empire: xAI is private. SpaceX is private (though planning an IPO in 2026 that could value it at $1.5 trillion). Starlink is private. The Boring Co. is private. Neuralink is private.
Today, Tesla is the only public stock in the entire constellation. If SpaceX goes public later this year, investors will have two entry points instead of one. Until then, there's only one door.
When you buy Tesla, you're not buying a car company trading at an absurd multiple to automotive earnings. You're buying the only liquid access point to a vertically integrated AI empire that spans training data, compute infrastructure, global distribution, physical deployment and cash generation. The car business isn't the point. The car business is the cash register that funds the empire.
Wall Street consensus ranks Tesla seventh out of seven in "Magnificent Seven" AI exposure. Nvidia first, obviously. Microsoft second for its OpenAI partnership. Alphabet is third for DeepMind and Gemini. Then Meta Platforms (META), Amazon.com (AMZN) and Apple $(AAPL)$. Tesla is last - a car company, not an AI play.
That list is upside-down.
Rockefeller's competitors didn't see Standard Oil coming until they were buying kerosene from him. By the time Wall Street figures out what Tesla actually is, the moat will be filled with water - and Musk will own the only bridge.
Charlie Garcia is founder and a managing partner of R360, a peer-to-peer organization for individuals and families with a net worth of $100 million or more. He owns Tesla shares.