Musk told Davos self-driving is “solved.” NHTSA has 80 violations on file.

Elon Musk has a problem with numbers.

Not the big, impressive ones. He's great at those. $430 billion net worth. 9,000 satellites in orbit. 100 gigawatts of solar manufacturing capacity, coming soon, any day now… just you wait.

No, his problem is with the small, verifiable numbers. The ones that show up in federal investigations. The ones on factory safety reports. The ones that count how many people he's fired while promising robots will make work "optional."

At Davos 2026, sitting across from BlackRock CEO Larry Fink, Musk spent half an hour painting a picture of technological utopia. Robots outnumbering humans. Artificial general intelligence by Christmas. Self-driving, he declared, is "essentially a solved problem."

I put his claims through the gauntlet.

Four were verifiably false. Three were misleading. Two can't be checked until the future arrives (convenient, that). Only two held up.

What Musk said

Here's the highlight reel:

Self-driving is "essentially a solved problem"
China deploys "over 1,000 gigawatts per year" of solar
Tesla's robotaxi service will be "widespread by year end"
AGI arrives by end of 2026, superintelligence by 2030
There will be "more robots than people"
AI and robotics are "the path to abundance for all"
Space will be the cheapest place for AI computing "within 2–3 years"
A 100x100 mile solar farm could power all of America

Some of this is aspirational. Some is defensible. And some of it is the kind of thing you'd expect to be corrected by a fact-checker before it made it to air. It wasn't.

The false claims

Let's start with the ones that don't require interpretation.

"Self-driving is essentially a solved problem"

Here's what "solved" looks like at Tesla:

NHTSA has an active investigation covering 2.4 million vehicles. As of December 2025, they've documented over 80 traffic safety violations. Red lights run. Lanes changed into oncoming traffic. The usual stuff you'd expect from a solved problem.

The investigation isn't new. It's just the latest. Since July 2021, NHTSA has logged 956 crashes involving Tesla's automated driving systems. That's more than every other manufacturer combined. The Washington Post counted at least 40 deaths.

Tesla's response to the latest probe? They asked for a five-week extension to respond. Granted.

Here's my definition of "solved": you don't need extensions to explain why your solution keeps killing people.

The experts aren't buying it either. Missy Cummings, former NHTSA senior safety advisor: "We are nowhere close to solving autonomous driving. The edge cases are essentially infinite."

Tesla's Full Self-Driving remains SAE Level 2. That means the driver is responsible, always. The car is an assistant. Tesla's own manual says drivers must "remain attentive" and be "ready to take over." That's not autonomy. That's a very expensive cruise control with good marketing.

VERDICT: ❌ FALSE

"China is deploying over 1,000 gigawatts per year of solar"

China deployed 277 gigawatts of solar in 2024. That's a world record. It's also not 1,000.

Musk inflated the number by 360%. I don't know why. The real figure is impressive enough. China added more solar capacity last year than any country in history. Their cumulative installed base hit 887 GW by early 2025. These are genuine achievements.

Maybe 277 didn't sound dramatic enough. Maybe he confused manufacturing capacity with deployment. Or maybe when you're worth $430 billion, a factor of 3.6 feels like a rounding error.

VERDICT: ❌ FALSE

"The lowest cost place for AI will be space within 2–3 years"

This one's fun because it contradicts physics.

NASA has spent decades learning that space is a terrible environment for electronics. Cosmic rays fry circuits. Temperature swings are extreme. And you can't exactly send someone up to swap a failed GPU.

Lumen Orbit, the startup actually trying to build space data centres, raised $11M last December. Their target? 2027 or 2028 for first systems. And they're positioning it for edge computing, not bulk AI training, because they understand the physics.

VERDICT: ❌ FALSE

"There will be more robots than people"

Current count of humanoid robots worldwide: somewhere between 10,000 and 50,000. Most are prototypes. Tesla has about 100 Optimus units deployed, doing what the company calls "simple tasks" in its factories.

Current count of humans: 8.1 billion.

Goldman Sachs, who are generally bullish on robotics, project 1.4 million humanoid robots by 2035 in their base case. Their optimistic case: 5–6 million. At those rates, robots outnumbering humans is a multi-generational proposition. We're talking late 21st century at the earliest, if it happens at all.

This isn't a false claim so much as a meaningless one. Saying there will eventually be more robots than people is like saying the sun will eventually engulf the Earth. Technically possible. Not useful for planning purposes.

VERDICT: ❌ FALSE (for any timeline that matters)

The misleading claims

Some claims weren't technically false. They were just deployed in ways that created false impressions.

"Robotaxis will be widespread by year end"

Tesla launched its robotaxi service in Austin on June 22, 2025. By January 2026, it had expanded to… the San Francisco Bay Area. That's it. Two cities.

And here's what Musk didn't mention: every Tesla robotaxi has a human safety monitor in the vehicle. These aren't autonomous rides. They're supervised test drives that customers pay for.

The supervised vehicles have still managed eight reported accidents. For comparison, Waymo provided 14 million fully autonomous rides last year across five metro areas. No safety monitors. No human hands hovering over a kill switch.

Musk's original 2025 goal was to serve "half the US population." By October, that became "8–10 metro areas." By January, it was two cities with babysitters.

VERDICT: 🟡 MIXED — the service exists, but "widespread" is doing marathon-level lifting

"Insurance companies are offering half-price insurance for FSD"

Technically true! Lemonade Insurance announced exactly that on January 21, 2026. One day before Musk's Davos appearance. Interesting timing.

What Musk didn't specify:

It's one insurance company, not "companies" plural
The 50% discount is on per-mile rates, not total premiums
Lemonade operates in only 10 states
The discount is based on Tesla's self-reported safety data (not independently verified)

Tesla's own insurance offers a 10% FSD discount. State Farm, Progressive, Geico, Allstate? No FSD discount at all. They've seen the crash data.

VERDICT: 🟡 MIXED, depends on good faith

"Solar is 5x more effective in space"

The physics here are real. No atmosphere means no absorption. No night means continuous collection. You can get 5–10x the energy per panel compared to Earth's surface.

The problem is everything else. Launching panels costs thousands of dollars per kilogram. Transmitting power back to Earth loses 50–80% of the energy. Maintenance is impossible. And you still need massive ground infrastructure to receive the power.

Net efficiency after transmission: 2–4x, not 5x. And at costs that make terrestrial solar look like a bargain.

VERDICT: 🟡 MISLEADING — true in isolation, useless in context

The unverifiable claims

"AGI by end of 2026"

Musk has been predicting imminent AGI for years. In 2024, he said it would arrive in 2025. When that didn't happen, he pushed it to 2026. This is a pattern.

AI researchers surveyed on the topic place median AGI arrival around 2040. The Metaculus prediction market estimates October 2027 for "weakly general AI." At the same Davos event, DeepMind's Demis Hassabis said 5–10 years.

I can't tell you Musk is wrong. I can tell you he's at the extreme optimistic end of expert forecasts, and his previous predictions on this topic have all been wrong.

VERDICT: 🌑 UNVERIFIABLE — we'll know by December

"Superintelligence by 2030–2031"

If we don't have AGI yet, superintelligence predictions are pure speculation. No serious research supports this timeline.

VERDICT: 🌑 NO SCIENTIFIC SUPPORT

The claims that checked out

SpaceX has approximately 9,000 satellites in orbit

Satellite tracking sources confirm 9,350–9,500 operational Starlink satellites as of January 2026. Musk said "9,000." Close enough.

VERDICT: ✅ TRUE

100x100 miles of solar could power the US

The math works for electricity. US consumption is about 4,000 TWh annually. Modern solar panels across 10,000 square miles in the Southwest could generate that.

The caveats: this doesn't include storage, transmission losses, or the fact that electricity is only 40% of total US energy use. But as back-of-napkin calculations go, this one's defensible.

VERDICT: ✅ APPROXIMATELY TRUE

The abundance problem

Here's where it gets uncomfortable.

"With robotics and AI, this is really the path to abundance for all."

Let me show you what "abundance" looks like for the people who actually build Musk's companies.

The layoff ledger:

Twitter went from 7,500 employees to about 1,500 in the span of months. Workers learned they'd been fired when they couldn't log into Slack. Tesla's April 2024 layoffs eliminated 10% of the global workforce, including the entire team responsible for building charging infrastructure. The infrastructure Musk says is essential to the EV transition.

The pay gap:

Tesla workers earn 20–40% less than their UAW counterparts at other automakers. No pension. No union. Stock options that take four years to vest, assuming you survive the culture that long.

The culture:

Musk on working hours: "Nobody ever changed the world on 40 hours a week." He's described working 80–120 hour weeks and expects the same from employees.

Musk on remote work: "The laptop class is living in la-la land… it's morally wrong."

Musk on commitment: At Twitter, he emailed staff demanding they commit to "extremely hardcore" work or take severance. Hundreds took the severance.

So here's my question: if work is about to become "optional" thanks to AI and robots, why demand 80-hour weeks right now? If abundance is coming, why the urgency? Why the layoffs?

The safety record:

Tesla's Fremont factory accumulated more OSHA violations from 2014–2018 than the ten largest US auto plants combined. 54 citations versus 11 for everyone else. Workers have described passing out from heat, limbs crushed by equipment, pressure to skip safety protocols to meet production targets.

One Tesla worker, Michael Sanchez, told Reveal News: "Everything feels like the future but us."

That quote has stayed with me.

The pattern

Here's what Musk has promised over the years:

2016: Full autonomy coming next year
2019: A million robotaxis on the road by 2020
2019: Cybertruck deliveries in 2021 (arrived late 2023)
2024: AGI by 2025
2026: AGI by end of this year

The cycle is predictable: bold prediction, media coverage, deadline passes, new prediction, more coverage. The accountability never arrives because there's always a new future to promise.

At Davos, Musk acknowledged this, sort of:

"There will be a lot of trauma and disruption along the way."

For the workers at Tesla, SpaceX, and Twitter, the trauma and disruption is here. It's in the mass layoffs and the safety violations and the 80-hour weeks. The abundance is still in the future. Always the future.

The verdict

Four false. Three misleading. Two unverifiable. Two true.

Here's what bothers me most: the pattern works. Musk makes big claims, the press reports them, the timeline passes, and by then there are new claims to cover. The old predictions fade into noise.

(…aaand cut!)

I don't know if Musk believes what he says at these events. Maybe he's genuinely optimistic. Maybe he's learned that predictions without deadlines generate the same headlines as predictions with deadlines, but without the accountability.

What I know is this: when you check the receipts, a different picture emerges. One where "solved" means "under investigation." Where "abundance for all" means layoffs for thousands. Where the future is always arriving and never here.

The robots are coming. The abundance is coming. The self-driving is coming.

Any day now.

Musk told Davos self-driving is “solved.” NHTSA has 80 violations on file.

Contents

What Musk said

The false claims

"Self-driving is essentially a solved problem"

"China is deploying over 1,000 gigawatts per year of solar"

"The lowest cost place for AI will be space within 2–3 years"

"There will be more robots than people"

The misleading claims

"Robotaxis will be widespread by year end"

"Insurance companies are offering half-price insurance for FSD"

"Solar is 5x more effective in space"

The unverifiable claims

"Superintelligence by 2030–2031"

The claims that checked out

The abundance problem

The pay gap:

The culture:

The safety record:

The pattern

The verdict

Watch the interview

References