Hope you enjoyed last week's issue on Pop Mart: The collectibles giant nobody is talking about.
That got lots of great feedback, thanks for the kind words!
As usual, right after we published we noticed mainstream media suddenly starting to talk about it... Rolling Stone, News.com, and The Guardian all published stories about it last week.
I love when that happens 😏
Speaking of under-discussed stories, today we're talking Tesla.
Everyone’s talking about how sales are sliding, Elon has potentially torched the brand, and the stock is bouncing all over the place.
But today, Brian Flaherty and I teamed up to shine a light on two large, important, and misunderstood parts of Tesla’s business:
In Part 1: Brian explores Tesla's interesting relationship with carbon credits
In Part 2, I'll explore sensor technologies (i.e. "Tesla Vision") and the LiDAR situation
But the company also has an underappreciated side hustle: selling carbon credits.
You can think of carbon credits like a ‘permission slip’ to emit carbon. In areas like California and Europe, regulators make automakers acquire these permission slips to sell gas-powered cars.
The idea is to incentivize the transition to EVs – and thus mitigate climate change.
The use of carbon credits via ‘emissions trading systems’ has grown rapidly over the past few years. Still, less than a quarter of global emissions are covered by ETS or a carbon tax. Source: World Bank
As one of the two leading EV companies in the world, Tesla generates a massive amount of carbon credits simply by selling their vehicles.
The company can then sell these permission slips to their gas-focused competitors, netting a tidy profit in the process.
In 2024, Tesla sold $2.8 billion in these carbon credits. That side hustle helped supplement the company’s $74.3 billion in regular sales + leasing revenue.
Ironically for Elon Musk, there’s some danger that this side hustle could actually come under threat thanks to Republican efforts to kill climate credits. But for the time being, it’s a solid extra revenue source.
Is Tesla a carbon credits company in disguise?
Tesla’s carbon credits sales catch a lot of flack – but only because they’re so misunderstood.
Lately, some very loud pundits have spouted off, arguing that Tesla’s “real product” is carbon credits and that the EV business is just incidental.
In this section, I want to briefly explain why these arguments are so flawed.
Full disclosure: I’m no fan of Elon Musk. But many criticisms of Tesla’s carbon credit business are overblown.
Carbon credits are highly profitable, but not costless
On the surface, carbon credits might seem like an insanely lucrative profit stream.
After all, these credits are manufactured out of thin air by governments, and Tesla faces almost no costs to sell them — so this $2.8 billion looks a lot like ‘free money.’
But here’s the thing: Tesla needs to make and sell EVs in order to earn these credits. It’s absolutely not free money!
Headlines arguing that some massive percentage of Tesla’s profits come from carbon credits are an abuse of accounting — they implicitly assume that selling credits is pure profit.
As an analogy, look at silver supply. Most of the world's silver is produced as a byproduct of mining other metals. We’d never argue that mining silver is "costless."
Carbon credits act in the same way! Tesla generates carbon credits as a byproduct of selling EVs. There's still a cost.
Tesla is not overly reliant on carbon credits
Nor is Tesla’s EV business overly reliant on these credits. In Q1 this year, Tesla turned a gross profit of $1.7 billion in their automotive segment on EV sales + leasing revenue alone.
Yes, carbon credit sales helped boost that to $2.3 billion. But they amounted to just 3.1% of total revenue.
By my calculations, Tesla has turned a gross profit on automobiles without credits since at least 2013.
2012 was the last year in which Tesla relied on carbon credits to turn a gross automotive profit, posting a $27m loss on sales + leasing revenue alone.
Carbon credit sales are a nice little business for Tesla – but flawed accounting metrics make people dramatically overestimate their importance.
Are carbon credits "real" profit?
It’s also common for critics to argue that carbon credits aren’t real profit for Tesla, since they’re nothing more than government subsidies.
There’s some merit to this argument. But let's get real: the entire point of carbon credits is to reward companies that help avoid emissions.
When Tesla earns carbon credits, the system is working exactly as it’s supposed to — the company is monetizing the social benefits of mitigating climate change.
Sure, you may think that the social damage caused by making Elon Musk richer far outweighs the social benefits of Tesla’s EVs.
But arguing that carbon credits aren’t ‘real profit’ is just wrong. It’s the same flawed logic as thinking that carbon emissions don’t impose real costs/externalities on society.
Tesla’s hidden liability?
As I’ve argued, many of the common criticisms of Tesla’s carbon credits business don’t really make sense. But that doesn’t mean there aren’t potential issues on the horizon.
In fact, Tesla may have managed to dig itself into a multi-billion dollar carbon credits hole...
Historically, Tesla has been a dependable supplier of carbon credits to other auto companies. As a result, they’ve negotiated long-term contracts to sell credits to those same companies in the future.
But as of Q1 2025, Tesla has unsatisfied carbon credit contracts of at least $4.6 billion, with $1.5 billion of that amount due over the next 12 months
(For comparison, recall that the company sold $2.8 billion in total credits in 2024.)
The amount of credits Tesla generates is directly tied to the amount of EVs they can sell — and as you've probably heard, sales have been falling like crazy.
How much would sales have to fall to create a real problem for Tesla’s carbon credit obligations?
By comparing Tesla’s vehicle deliveries to credit sales over the past eight quarters, I’d estimate that the company currently earns roughly $1,333 in credits per vehicle.
Determining Tesla’s average credit per vehicle is tricky, since the company may not recognize credit revenue in the same quarter that the underlying vehicle delivery occurs.
That would mean to satisfy Tesla’s $1.5 billion in credit commitments over the next year, the company would need to sell around 1.1 million vehicles.
Even with the dip this quarter, Tesla has a little breathing room at their current pace. But if sales continue to fall, the risk is that Tesla can no longer generate sufficient credits to meet their obligations.
Unfortunately, Tesla offers very little disclosure about their carbon credits business. The company’s financial statements are notoriously opaque.
I’d really need to know more contract details to understand if this could turn into a real problem.
Unfortunately, we just don’t have the answers to these questions. But it’s certainly a risk I’m paying attention to.
Part 2: Tesla’s blind spot — LiDAR and the future of vision
Tesla’s vision for autonomy has always looked a little different. Literally.
While most autonomous vehicle employ a combination of sensors — radar, ultrasonic, and especially LiDAR — Tesla has gone the opposite direction.
In fact, Tesla has deliberately eliminated these additional sensors from its vehicles, choosing instead to rely solely on a camera-based system they call “Tesla Vision”
While that’s an elegant idea in theory, it’s also a bit stubborn. As the rest of the autonomous vehicle (AV) industry is adding sensing for environmental perception, Tesla is removing it.
So what’s going on here? Is Tesla ahead of the curve, or is it about to hit something it can’t see?
To understand this, let's first do a quick primer on how AVs “see” the road.
How self-driving cars perceive the world
There are three primary technologies in play here: Radar, LiDAR, and Cameras.
Most AV companies have sensor fusion systems which combining one or more of the above.
But not Tesla. Their "Tesla Vision" system uses uses eight cameras and a neural net to infer depth from 2D images. No backup, no fallback. At least for now.
What is the advantage of LiDAR?
Alts Community member Patrick Bybee is a Director of Electrical Engineering with 20 years of experience, including Raytheon and Teledyne. He has developed both LiDAR and Thermal/IR camera systems professionally.
He explains why LiDAR is advantageous:
"The advantage of LiDAR is getting an exact distance to the target based on the time-of-flight (TOF) of the photons from the laser pulse. We know how fast light travels, so if we know when we fire the laser, we can deduce the exact distance to the target."
Put simply, LiDAR doesn’t guess — it measures. It doesn’t care about light, color, or contrast. It just maps the world around the car in real-time.
Tesla’s bet is that cameras plus AI is enough. No need for lasers and redundancy — just collect enough data, train the model hard enough, and let the neural nets do the rest.
It’s a bold claim. And not one most of the industry agrees with.
Per Patrick:
"For vision-based systems [like Tesla’s], distance is estimated based on Convolutional or Deep Neural Network (CNN/DNN) classifiers. These approximate distances based on the number of pixels used for a classified target. For example, a stop sign is usually 6 feet off the ground. The CNN/DNN knows the size of the pixel in the camera, and how many pixels are covered by the stop sign on the sensor. Based on that data, the distance to the stop sign can be quite usefully estimated.
Much of the difference seems to come down to edge cases.
Most technologies & companies can handle the easy obstacles. But it's the rare stuff — the white mattress on the snowy highway, the construction cone tipped over, the person in black crossing the street at night — that separates safe from sorry.
“It’s not about driving on a sunny day in a straight line. It’s about the mattress that falls off a truck, or the kid that darts out between cars. That’s where sensor fidelity matters." — Raquel Urtasun, Waabi CEO
So why doesn’t Musk use LiDAR?
Musk’s long-standing belief is that LiDAR is a crutch. He has called it expensive, unnecessary, and inelegant.
His philosophy is simple: humans drive with eyes and brains, not lasers. So that’s how AI should do it too.
And frankly, there’s something admirable about that purity. The Tesla approach is clean. Scalable. Minimal.
“Anyone relying on LiDAR is doomed. Doomed! Expensive sensors that are unnecessary...like having a whole bunch of expensive appendices… It’s a fool’s errand” - Elon Musk, Tesla Autonomy Day 2019
What's fascinating about reading Elon’s comment today is how outdated it has become.
Back in 2019 LiDAR was indeed the expensive option, and Tesla Vision seemed like a brilliant, forward-thinking, cost-conscious move.
But recently, the cost of LiDAR sensors has absolutely plummeted in recent years, making the technology far more accessible for automotive use.
The price of a LiDAR unit in China has plummeted from approximately $4,100 to just $138, representing a dramatic decrease.
Furthermore, there’s also something very risky about Musk’s approach. Because while human eyes and brains can drive, we also crash. A lot.
The US average is one crash every 670,000 miles. So on the surface, Tesla is over twice as safe as average. Hooray!
But here’s the thing: 1) This is autopilot, not accidents from "regular driving," and 2) This is Tesla’s own data. There’s no third-party verification. No context on fault.
Just like with their financial statements, it’s all very opaque.
Now, automakers are only required to report crashes if they result in a fatality, a vulnerable road user being struck, or certain other serious outcomes.
The lack of transparency and third-party verification means the public must largely rely on Tesla’s own reporting, with minimal independent oversight.
Competitors Waymo and Cruise have also reported significantly fewer collisions than human drivers — including data showing improved performance in edge cases.
So while Tesla’s system definitely seems safe, it’s increasingly difficult to know for sure.
Meanwhile, Tesla still has the highest accident rate of any automobile brand. And when it comes to edge cases, the data suggests Tesla may be a step behind.
Is thermal imaging a sleeper technology?
Tesla seems to be at a critical juncture here. It can either:
Double down on its current “Vision-only” philosophy
Pivot toward LiDAR like nearly every major competitor (unlikely)
Carve a new path — leveraging other sensing technologies that add robustness without betraying the core ideology.
Option 3 is where it gets interesting.
Tesla may not need to embrace LiDAR. Instead, it could quietly shore up the weaknesses in Tesla Vision using low-cost thermal imaging.
Thermal cameras see heat, not light — perfect for detecting pedestrians and animals in darkness, fog, or glare.
Thermal imaging isn't new, but just like LiDAR, the cost is coming down due to better technology, ITAR exemptions, European and Chinese providers.
Watch the first 30 seconds of this YouTube video to see a thermal-enabled car driving in heavy fog. You can't see the pedestrian, but the camera easily can — and it tells the car to stop.
"In my opinion, thermal is the big sleeper in this discussion. One thing Elon might be right on, it is a visible world that our cars and roads are designed for. We operate based on signage and signaling from other motorists or by personally seeing cyclists, pedestrians, animals, etc. With thermal, you are using the same "visible world". However, you have a passive system that can see through dust, fog, at night, and can easily discriminate an object from the background. Distance can be accurately approximated using the same style CNN/DNN systems as visible spectrum cameras. If Elon made any mistake, my opinion is not including thermal." - Patrick Bybee
He continues:
"It will likely not be long until NIHST sees how many lives will be saved with Automatic Emergency Braking (AEB) tied into the classification by thermal camera video feeds. I believe there is a solid probability of NIHST regulating thermal cameras into vehicles. I personally think that with the passive nature of the system, lack of cross talk, and a swiftly falling cost profile, Thermal is the technology to keep an eye on"
That's it for today!
Thanks for reading, We'd love to hear your thoughts, comments, or critiques in the Alts Community.
See you next time, Stefan & Brian
Disclosures
This issue was co-authored by Stefan von Imhof and Brian Flaherty
Special thanks to Alts community member Patrick Bybee for dropping 20 years of sensor technology knowledge bombs into this issue
Alt Assets, Inc. has no holdings in any companies mentioned in this issue.
