MW I got hate mail for being a Tesla stock bull - but here's the No. 1 thing many are missing about Elon Musk's AI ambitions
By Charlie Garcia
Tesla stock is the only portal right now to xAI and Musk's other empire-building plans
Many readers torched Charlie Garcia's column on Tesla's potential. They should take a closer look.
Editor's note: Columnist Charlie Garcia shares select emails and online comments from his virtual mailbag each week.
Dear Charlie,
No thanks. Hard pass on Tesla.
Tesla's stock $(TSLA)$ returned 3.3% per year over the past four years, with no dividends paid. You would have done better sitting in cash. A 300 P/E? The company would have to increase profits 900% without any increase in stock price just to make the valuation somewhat reasonable.
There are much better companies out there with better valuations and nice dividends. TSLA is one of the biggest legal pyramid schemes ever created by humans. Other than bitcoin.
Bradley
Dear Bradley,
You're pricing a car company. That's the mistake.
300 P/E is absurd for automotive. It's not absurd for the only liquid access point to a vertically integrated AI monopoly with 7 billion miles of training data, 1 million GPUs and $4 billion in quarterly free cash flow.
"Past four years" measures the construction phase. Colossus. Shanghai. The neural network. Robotaxis launched six months ago. You're grading the blueprint, not the building.
"Pyramid scheme" is a strong phrase for a company printing $4 billion a quarter in free cash flow. Pyramid schemes require new money to pay old investors. Tesla pays itself.
Charlie
P.S.: Dividends are what companies pay when they've run out of ideas. Tesla hasn't.
Read: I refused to invest in Tesla for years - but now's the time to bet on Elon Musk
Dear Charlie,
You really drank that Kool-Aid, didn't you? Good luck, I guess. I'll never give a penny to Tesla or X, so Musk had better own all the infrastructure if he wants to profit off of me.
Mike
Dear Mike,
You don't have to buy a Tesla. But if you use GPS, internet, or anything trained on AI compute, you're already in the tollbooth.
That's the point.
Charlie
P.S.: Rockefeller's customers didn't like him either. They still bought kerosene.
Dear Charlie,
You were waiting for a 296 P/E ratio? Seriously? For every $296 you invest, you're getting one dollar in earnings. That's not investing, that's speculation dressed up in a nice suit. How do you justify that math to your readers?
Randall
Dear Randall,
Tell me you didn't read the piece without telling me you didn't read the piece.
296 P/E for a car company? Insane. I said that.
296 P/E for the only liquid access point to training data, compute infrastructure, physical deployment and cash generation at scale? Different math.
You're dividing price by earnings. I'm dividing price by optionality on AI, robotaxis and robotics that aren't in the earnings yet.
Amazon had no earnings for a decade. Bezos was building infrastructure. The P/E guys missed a 200-bagger.
You're watching the rearview mirror. I'm watching the road.
Charlie
P.S.: For every dollar in Tesla's current earnings, there's $10 in infrastructure not yet monetized. That's the bet.
Dear Charlie,
Your Amazon.com analogy doesn't hold water. Amazon had no substantial competition during its unprofitable years. They were building in open space. Tesla is facing fierce competition from players that do almost everything better. Everything with cars is done better or cheaper or both by competitors. BYD (CN:002594) is eating their lunch in China. Legacy automakers are catching up fast.
And the AI space? It's stacked to the rafters with superior talent that leads Tesla by miles. Alphabet's $(GOOGL)$ $(GOOG)$ Google, OpenAI, Meta (META), Microsoft $(MSFT)$. Competition shrinks profit margins. Amazon (AMZN) is not a workable analogy here.
Hiram
Dear Hiram,
Amazon had no competition?
Walmart: $500 billion revenue, best logistics on Earth. Microsoft: unlimited capital, Azure launched same year as AWS. Barnes & Noble: 700 stores, dominant brand.
Every analyst said Amazon stock was overpriced and about to get crushed by "real" companies. Sound familiar?
"Tesla faces fierce competition from players that do almost everything better." Better at what? Cars? Agreed. Toyota (JP:7203), BMW (XE:BMW), BYD all make better cars. But they don't have 7 billion miles of training data. They don't have a million GPUs. They don't have robotaxis deployed.
You're grading a basketball team on baseball stats.
"The AI space is stacked with superior talent." Google has 30,000 AI researchers. Where's their robotaxi fleet? Their training data moat? Their humanoid robot?
Talent is an input. Infrastructure is an output.
Charlie
P.S.: Google's been working on self-driving since 2009. Tesla passed it in deployed miles three years ago. Talent isn't the bottleneck.
Dear Charlie,
Your audacious comparison set a volatile crowd on fire, but it makes sense. This kind of vertical integration has been in the books for a while, and it's been applied often, though never at this scale. Historically, it's often been a success.
While the story makes sense, my main question is whether the flow of expected future profits will ever make it to the Facebook crowd of Tesla investor fans. That's the disconnect I keep coming back to.
P.S. I hate Elon, but I still tried to buy TSLA at 175 not too long ago. Thanks for your high-profile opinions. Always a delight.
Pierre
Dear Pierre,
You've identified the real risk.
The thesis requires the profits to flow through to shareholders. Musk could spin off xAI. He could dilute Tesla to fund new ventures. He could get bored and go to Mars.
I'm betting he won't. The vertical integration only works if Tesla remains the liquid access point. Spin off the pieces and you destroy the moat.
But you're right to question it. "Will shareholders see the value?" is a better critique than "cameras can't work."
Thanks for engaging with the actual argument.
Charlie
P.S.: Hating the man and owning the stock is intellectually honest. Most people can't separate the two. I don't hate anyone. It clouds the judgment.
Dear Charlie,
I like the argument. You made me think about these issues in a new way. But I'm not buying.
Too many Wall Streeters think science works the same as economics. FSD will never work no matter how much data Tesla collects, because camera data represents less than 30% of all available sensory data. Ask yourself why so many predatory animals can see perfectly on a moonless night. They evolved beyond the visible spectrum for a reason.
Google is a premier science and engineering center, and they recognized that gathering 100% of 30% is inferior to gathering more than 30% of 100%. That's why Waymo uses lidar.
And as far as Rockefeller's success goes, being first is the same as being lucky.
Thomas
Dear Thomas,
Owls don't have lidar. They have better brains.
Evolution ran this experiment for 200 million years and landed on the same answer Tesla did: Amplify visible light through neural processing. Same spectrum, smarter interpretation.
The handful of animals that sense beyond visible wavelengths? Pit vipers finding mice in burrows. Handy for locating snacks. Less handy for reading a construction worker's hand signals at highway speed.
And Rockefeller wasn't first. He was four years late to a party with 30 competitors. Crude was selling for less than the barrels holding it.
He won because he saw the chokepoint wasn't underground. It was in the infrastructure.
Pipeline beats drill bit. Every time.
Charlie
P.S.: Thanks for reading closely enough to argue. That's the only compliment that matters.
Dear Charlie,
Tesla FSD has a huge disadvantage being camera-based only. Level 4 is never going to happen, while Waymo keeps expanding. The singular dependency on Musk making all decisions means they have millions of cars that will never reach Level 4.
Probabilistic data-based decision-making is not going to work when lives are at risk, no matter how much data you collect. Most driving data is repetitious and useless, and exception data without lidar isn't useful either.
FSD still hasn't been delivered after folks paid $5K to $8K almost 10 years ago. I want to support Musk, but it's always more hype built on more hype. As soon as Tesla as an EV company started getting overtaken in the market, suddenly Tesla is an AI company and Musk is an excellent data-center builder? LOL. It's always magic happening in the future.
Good luck to the author on their shares.
Matthew
Dear Matthew,
"FSD still not delivered after 10 years." Robotaxis are running in Austin, Texas, right now. Unsupervised. No safety driver. You're welcome to go take a ride.
"Camera-based only. Level 4 not gonna happen." Humans drive with two cameras and a neural network. No lidar. The road system was literally designed for vision-based navigation.
Waymo costs $150,000 per vehicle. Tesla costs the cameras already on the car. One scales. One doesn't.
"Most driving data is repetitious and not useful." That's how neural networks work. Repetition is the training signal. Edge cases emerge from baseline patterns. Tesla has 7 billion miles. Waymo has 160 million.
"It's always magic in the future." The 1 million GPU cluster exists in the present.
"Good luck on your shares." Thanks. But I'd rather have the data than the luck.
Charlie
P.S.: The LOL is noted. I'll check back in a year.
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. Follow him on X here.
MW I got hate mail for being a Tesla stock bull - but here's the No. 1 thing many are missing about Elon Musk's AI ambitions
By Charlie Garcia
Tesla stock is the only portal right now to xAI and Musk's other empire-building plans
Many readers torched Charlie Garcia's column on Tesla's potential. They should take a closer look.
Editor's note: Columnist Charlie Garcia shares select emails and online comments from his virtual mailbag each week.
Dear Charlie,
No thanks. Hard pass on Tesla.
Tesla's stock (TSLA) returned 3.3% per year over the past four years, with no dividends paid. You would have done better sitting in cash. A 300 P/E? The company would have to increase profits 900% without any increase in stock price just to make the valuation somewhat reasonable.
There are much better companies out there with better valuations and nice dividends. TSLA is one of the biggest legal pyramid schemes ever created by humans. Other than bitcoin.
Bradley
Dear Bradley,
You're pricing a car company. That's the mistake.
300 P/E is absurd for automotive. It's not absurd for the only liquid access point to a vertically integrated AI monopoly with 7 billion miles of training data, 1 million GPUs and $4 billion in quarterly free cash flow.
"Past four years" measures the construction phase. Colossus. Shanghai. The neural network. Robotaxis launched six months ago. You're grading the blueprint, not the building.
"Pyramid scheme" is a strong phrase for a company printing $4 billion a quarter in free cash flow. Pyramid schemes require new money to pay old investors. Tesla pays itself.
Charlie
P.S.: Dividends are what companies pay when they've run out of ideas. Tesla hasn't.
Read: I refused to invest in Tesla for years - but now's the time to bet on Elon Musk
Dear Charlie,
You really drank that Kool-Aid, didn't you? Good luck, I guess. I'll never give a penny to Tesla or X, so Musk had better own all the infrastructure if he wants to profit off of me.
Mike
Dear Mike,
You don't have to buy a Tesla. But if you use GPS, internet, or anything trained on AI compute, you're already in the tollbooth.
That's the point.
Charlie
P.S.: Rockefeller's customers didn't like him either. They still bought kerosene.
Dear Charlie,
You were waiting for a 296 P/E ratio? Seriously? For every $296 you invest, you're getting one dollar in earnings. That's not investing, that's speculation dressed up in a nice suit. How do you justify that math to your readers?
Randall
Dear Randall,
Tell me you didn't read the piece without telling me you didn't read the piece.
296 P/E for a car company? Insane. I said that.
296 P/E for the only liquid access point to training data, compute infrastructure, physical deployment and cash generation at scale? Different math.
You're dividing price by earnings. I'm dividing price by optionality on AI, robotaxis and robotics that aren't in the earnings yet.
Amazon had no earnings for a decade. Bezos was building infrastructure. The P/E guys missed a 200-bagger.
You're watching the rearview mirror. I'm watching the road.
Charlie
P.S.: For every dollar in Tesla's current earnings, there's $10 in infrastructure not yet monetized. That's the bet.
Dear Charlie,
Your Amazon.com analogy doesn't hold water. Amazon had no substantial competition during its unprofitable years. They were building in open space. Tesla is facing fierce competition from players that do almost everything better. Everything with cars is done better or cheaper or both by competitors. BYD (CN:002594) is eating their lunch in China. Legacy automakers are catching up fast.
And the AI space? It's stacked to the rafters with superior talent that leads Tesla by miles. Alphabet's (GOOGL) (GOOG) Google, OpenAI, Meta (META), Microsoft (MSFT). Competition shrinks profit margins. Amazon (AMZN) is not a workable analogy here.
Hiram
Dear Hiram,
Amazon had no competition?
Walmart: $500 billion revenue, best logistics on Earth. Microsoft: unlimited capital, Azure launched same year as AWS. Barnes & Noble: 700 stores, dominant brand.
Every analyst said Amazon stock was overpriced and about to get crushed by "real" companies. Sound familiar?
"Tesla faces fierce competition from players that do almost everything better." Better at what? Cars? Agreed. Toyota (JP:7203), BMW (XE:BMW), BYD all make better cars. But they don't have 7 billion miles of training data. They don't have a million GPUs. They don't have robotaxis deployed.
You're grading a basketball team on baseball stats.
"The AI space is stacked with superior talent." Google has 30,000 AI researchers. Where's their robotaxi fleet? Their training data moat? Their humanoid robot?
Talent is an input. Infrastructure is an output.
Charlie
P.S.: Google's been working on self-driving since 2009. Tesla passed it in deployed miles three years ago. Talent isn't the bottleneck.
Dear Charlie,
Your audacious comparison set a volatile crowd on fire, but it makes sense. This kind of vertical integration has been in the books for a while, and it's been applied often, though never at this scale. Historically, it's often been a success.
While the story makes sense, my main question is whether the flow of expected future profits will ever make it to the Facebook crowd of Tesla investor fans. That's the disconnect I keep coming back to.
P.S. I hate Elon, but I still tried to buy TSLA at 175 not too long ago. Thanks for your high-profile opinions. Always a delight.
Pierre
Dear Pierre,
You've identified the real risk.
The thesis requires the profits to flow through to shareholders. Musk could spin off xAI. He could dilute Tesla to fund new ventures. He could get bored and go to Mars.
I'm betting he won't. The vertical integration only works if Tesla remains the liquid access point. Spin off the pieces and you destroy the moat.
But you're right to question it. "Will shareholders see the value?" is a better critique than "cameras can't work."
Thanks for engaging with the actual argument.
Charlie
P.S.: Hating the man and owning the stock is intellectually honest. Most people can't separate the two. I don't hate anyone. It clouds the judgment.
Dear Charlie,
I like the argument. You made me think about these issues in a new way. But I'm not buying.
Too many Wall Streeters think science works the same as economics. FSD will never work no matter how much data Tesla collects, because camera data represents less than 30% of all available sensory data. Ask yourself why so many predatory animals can see perfectly on a moonless night. They evolved beyond the visible spectrum for a reason.
Google is a premier science and engineering center, and they recognized that gathering 100% of 30% is inferior to gathering more than 30% of 100%. That's why Waymo uses lidar.
And as far as Rockefeller's success goes, being first is the same as being lucky.
Thomas
Dear Thomas,
Owls don't have lidar. They have better brains.
Evolution ran this experiment for 200 million years and landed on the same answer Tesla did: Amplify visible light through neural processing. Same spectrum, smarter interpretation.
The handful of animals that sense beyond visible wavelengths? Pit vipers finding mice in burrows. Handy for locating snacks. Less handy for reading a construction worker's hand signals at highway speed.
And Rockefeller wasn't first. He was four years late to a party with 30 competitors. Crude was selling for less than the barrels holding it.
He won because he saw the chokepoint wasn't underground. It was in the infrastructure.
Pipeline beats drill bit. Every time.
Charlie
P.S.: Thanks for reading closely enough to argue. That's the only compliment that matters.
Dear Charlie,
Tesla FSD has a huge disadvantage being camera-based only. Level 4 is never going to happen, while Waymo keeps expanding. The singular dependency on Musk making all decisions means they have millions of cars that will never reach Level 4.
Probabilistic data-based decision-making is not going to work when lives are at risk, no matter how much data you collect. Most driving data is repetitious and useless, and exception data without lidar isn't useful either.
FSD still hasn't been delivered after folks paid $5K to $8K almost 10 years ago. I want to support Musk, but it's always more hype built on more hype. As soon as Tesla as an EV company started getting overtaken in the market, suddenly Tesla is an AI company and Musk is an excellent data-center builder? LOL. It's always magic happening in the future.
Good luck to the author on their shares.
Matthew
Dear Matthew,
"FSD still not delivered after 10 years." Robotaxis are running in Austin, Texas, right now. Unsupervised. No safety driver. You're welcome to go take a ride.
"Camera-based only. Level 4 not gonna happen." Humans drive with two cameras and a neural network. No lidar. The road system was literally designed for vision-based navigation.
Waymo costs $150,000 per vehicle. Tesla costs the cameras already on the car. One scales. One doesn't.
"Most driving data is repetitious and not useful." That's how neural networks work. Repetition is the training signal. Edge cases emerge from baseline patterns. Tesla has 7 billion miles. Waymo has 160 million.
"It's always magic in the future." The 1 million GPU cluster exists in the present.
"Good luck on your shares." Thanks. But I'd rather have the data than the luck.
Charlie
P.S.: The LOL is noted. I'll check back in a year.
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. Follow him on X here.
(MORE TO FOLLOW) Dow Jones Newswires
January 20, 2026 16:40 ET (21:40 GMT)
MW I got hate mail for being a Tesla stock bull - -2-
Agree? Disagree? Share your comments with Charlie Garcia at charlie@R360Global.com. Your letter may be published anonymously in the weekly "Dear Charlie" reader mailbag.
By emailing your comments to Charlie Garcia, you agree to have them published on MarketWatch anonymously, or with your first name if you give permission. You understand and agree that Dow Jones & Co., the publisher of MarketWatch, may use your story, or versions of it, in all media and platforms, including via third parties.
More from Charlie Garcia:
I refused to invest in Tesla for years - but now's the time to bet on Elon Musk
'Trump doctrine' demands the world share riches with the U.S. - Venezuelan oil is the latest bargaining chip
China is using silver as an economic weapon. What that means for investors and prices.
Quantum computing works - now investors will see if the stocks do too
-Charlie Garcia
This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.
(END) Dow Jones Newswires
January 20, 2026 16:40 ET (21:40 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
