I'm at CES in Las Vegas to check out the latest in autonomous driving. Here's what I've learned so far.

Amazon-backed Zoox is unlike any other robotaxi.

This is the first year Zoox, an Amazon-backed robotaxi company, will be giving live demonstrations of its service during CES.

I got to take a ride in one on Monday night in front of Resorts World. (The company tagline that I saw from an ad at the Harry Reid International Airport was: "Don't just do the Strip. Zoox it.")

My immediate thoughts were that Zoox feels unlike any other robotaxi or pseudo-robotaxi on the market. It felt more like I was on a theme park ride than in an everyday car we're familiar with.

Unlike Waymo's robotaxis, Zoox is not a regular car you could buy that's been retrofitted with sensors. The Zoox car is bi-directional — meaning there's no real front or back of the car — and the inside has no steering wheel, just seats.

The robotaxis were clearly a great tourist attraction from what I saw. My Uber driver wasn't too happy about them.

Uber, Lucid, and Nuro have big plans to scale.

Uber, Lucid, and Nuro had a swanky cocktail hour at Fontainebleau Las Vegas, where they quite literally wined and dined a room full of reporters, analysts, and investors: endless glasses of wine and an open bar, lobster tails, jumbo shrimp, too many appetizers to count, and a giant charcuterie board — the works.

Maybe understandably so? 2026 will be a big year for the three companies.

Uber's plan is to roll out a robotaxi service by late 2026. The first market is San Francisco, where Uber will directly compete with Waymo. These two companies are partners in other markets, like Austin.

"We've been moving very, very quickly," Nuro's co-CEO and cofounder Dave Ferguson said. "We signed this partnership last July. We're already testing the production-intent vehicles on public roads. And very soon, we're going to have tens of thousands of them worldwide."

Here's a 60,000-pound John Deere combine for scale.

A quick image to get a sense of how big CES's mobility division is at West Hall of the convention center: There's a 60,000-pound combine from John Deere that's sitting in the middle of the showroom.

The combine is one of the world's largest on the market, according to Julian Sanchez, an engineer at the machinery company.

Even so, John Deere doesn't even have the largest footprint on the floor. This year, it's Hyundai.

The combine isn't autonomous in the way we think about self-driving cars, Sanchez told me, but it is self-steering.

The world got a reality check on self-driving cars since the last hype cycle.

There's a lot of talk of self-driving cars in the automotive industry, but the scope of what it can realistically achieve has narrowed down in the last decade or so.

Paul Costa, an ex-Apple veteran of 25 years who worked on the company's abandoned self-driving car project, gave me a bit of interesting color from what he saw at CES in 2015 — when the driverless car hype was reaching its peak — and what's different now.

"My sense at the time was that people really wanted to focus on Level 5 autonomy," Costa, who now leads Ford's electrical engineering team, told me. Level 5 is the highest level of autonomous driving set forth by the Society of Automotive Engineers. That means full autonomy in all weather conditions and no geofences. Waymo is currently Level 4.

The tone has been brought down to reality, according to Costa. The focus is on highly advanced driver assistance systems and eyes-off driving or Level 3 systems, he said.

"Now, I feel like here in 2026, L3 is extremely interesting," Costa said. "It's interesting for me to see how the industry — its focus has changed over the years."

The big unknown for robotaxis: How long can they last on the roads?

AV operators, like Waymo, Uber, or Lyft, need to know the lifespans of their vehicles to determine the true cost of staying in business.

Stephen Hayes, VP of Lyft's autonomous vehicle program, told me that the "lifetime mileage" the company can accrue on a robotaxi is still unknown.

"How long will an AV last?" Hayes said. "Today, we're in the early innings, obviously, of the autonomous transition and very few AV players have vehicles on the road that have run hundreds of thousands of miles."

Part of the business proposition for ride-hailing companies like Lyft is reselling its fleet of vehicles, also known as remarketing. Lyft owns a fleet through its subsidiary called Flexdrive. For robotaxis, remarketing remains an "unknown," Hayes said.

"There's no market for it," he said. "There's no secondary market where it's like, 'Hey, I want to go buy a used AV.' Ten years from now, it's going to look completely different, but we don't know in what ways."