When Tesla revealed the Cybercab in 2024, many people were baffled by the automaker’s decision to make it a two-seater. I had similar thoughts when I first saw it at the LA Auto Show later that year: What kind of taxi has only two seats? Once you get above a tuktuk, that’s an asinine idea no one will ever want.

For 18 months, that seems to have been the prevailing attitude. On Reddit and other social media platforms, users have weighed in on the utility, or lack thereof, of a two-seat robotaxi. “You could strap a few people to the roof for a special discount,” quipped one commenter. Nevertheless, the first Tesla Cybercab was put into production last month, and now prototypes have been spotted testing both on public roads and on the grounds of Tesla’s assembly plant in Austin, Texas.

But in case you thought the Cybercab would be alone in absorbing all the anti-two seater vitriol, now there are two such vehicles.

The Lunar eclipse

At Lucid Motors’ Investor Day in New York City this month, its executive team laid out the EV maker’s plan to reach profitability. It includes three new models on a less expensive midsize platform — the first will arrive next year — an all-new electric powertrain, and a continuing focus on semi-autonomous driving functions and robotaxi partnerships.

Then, in a “just one more thing” moment, Lucid’s acting CEO Marc Winterhoff revealed a concept car hidden behind a curtain at one end of the hall. The two-seat robotaxi concept, dubbed the Lucid Lunar, was displayed without doors to show off its interior space and large luggage bay. Visions of the Tesla Cybercab, different but the same, flashed through my mind.

In a “fireside chat” with Uber’s Andrew Macdonald, Winterhoff revealed the rideshare company has pledged to buy 20,000 Gravitys fitted with robotaxi sensors and software from Nuro. The two firms plan a similar deal with an upcoming Lucid midsize EV model as well.

It took a few hours with Lucid Motors executives for me to understand why, in fact, I was wrong about two-seat robotaxis. And why ridehailing services the world over might buy giant fleets of them. Presuming, of course, that robotaxis prove to be safe, reliable, not a major contributor to urban congestion, and notably cheaper for ridehail fleets than are today’s human drivers using a wide variety of EV and internal combustion models.

At minimum, deploying two-seat robotaxis would require an added step during the hail process: Every user would have to say how many people were traveling, to ensure a vehicle arrived with a sufficient number of seats. Users would most likely accept this added friction in the process, especially if two-seaters were cheaper than other alternatives.

The Lucid Lunar concept uses the same wide horizontal display on the dash as the Lucid Cosmos, but no steering wheel. It would be built on a shortened version of the midsize platform that will underpin the Cosmos and its two siblings, which would deliver major cost savings on the basic architecture. The Lunar is both lower to the ground and smaller than the Cosmos, and — my “aha!” moment — it was designed to be as energy-efficient and low-cost as possible for buyers.

Smaller, lighter, CHEAPER!

In their fireside chat, Winterhoff and Macdonald noted that more than 90 percent of the rides Uber provides today have just one or two passengers. Other studies put the proportion slightly lower — though still a strong majority. The logic of the vehicle, Lucid chief engineer Zach Walker later explained, was that the needs of ridehail fleet operators differ substantially from those of individual drivers. 

Physics dictates a two-seat EV will be smaller and lighter than one with more seats. That will make it cheaper to buy and to operate, primarily because it can provide the necessary range from a battery with lower capacity — which will cost less and recharge faster for the same range added. Ridehail companies will need robotaxis with the absolutely lowest possible lifetime cost, to make the expensive tech practical versus human-driven vehicles.

According to chief engineer Walker, every 1-kWh reduction in battery size will save a robotaxi operator $1,000 per year in recharging costs, presuming it covers 100,000 miles a year. How efficient could that two-seat robotaxi be versus a four-seat compact EV? Walker said Lucid projects efficiencies of 5.5 miles per kilowatt-hour, perhaps as high as 6 mi/kWh, in typical use. (For comparison, the most energy-efficient EV sold in the US today is the rear-wheel-drive Lucid Air Pure, with an EPA efficiency rating of 146 MPGe, which translates to 4.4 mi/kWh.)

Beyond that, he added, there’s a “virtuous circle” of not just size and weight reduction, but engineering to purpose. Walker noted Lucids sold to individual buyers must have stellar roadholding and handling capabilities; it’s part of the brand. But the preset behavior of a robotaxis’ driving algorithm, and the limits of the performance it will demand, are known in advance, and won’t impose handling loads as extreme as a few private owners will. That means suspension structures can be tuned for comfort, using softer and less complex bushings. It could even let Lucid reduce or eliminate certain reinforcements and braces required to keep the structure stiff during extreme handling maneuvers, Walker said.

Does lower drag matter?

Both the Tesla and Lucid two-seaters are low and sleek, which minimizes aerodynamic drag. That further boosts efficiency at speeds of 30 mph and above, when the energy required to overcome wind resistance exceeds the energy required to move the vehicle itself. That’s an advantage for airport runs that require travel on highways, less so in trips predominantly based in urban environments, as Reilly Brennan of Trucks VC notes in his weekly newsletter.

Brennan questioned why the form factor of robotaxis mimicked the shape of a two-door coupe, despite the limitations it imposes on entry and exit. He proposed as an alternative a famous study by designer Giorgetto Giugiaro for a modern New York City taxi, created fully 50 years ago. It’s still crisp, modern, upright — and easier to get into and out of than a coupe.

But Giugiaro’s tall, square, upright design would present a challenge for wind resistance. Still, Brennan is convinced it’s the right approach.  “Most of the trips for these will be low-speed and urban, which will make the Cd [drag coefficient] issues largely moot,” he said in an email. 

After seeing the Lucid Lunar, every reporter at Lucid Investor Day had the same second thought I did 18 months earlier, when I saw the Tesla Cybercab: If you just added a steering wheel, that would make an awesome small, fast EV sports coupe. 

Asked directly about such a possibility, Walker laughed. “Oh, I have lots of ideas in my spare time,” he said. “But this [the Lunar robotaxi] is the one we’re talking about today.”

Is the long freeze on Chinese automakers selling cars in the US finally starting to thaw? China is the largest auto market in the world and now the largest car exporter, too. But high tariffs and geopolitical tensions have kept Chinese automakers away from US customers. Many of those cars are ready for primetime and are fully competitive with current offerings in the United States.

Geely is among those Chinese carmakers that want to sell their cars here. It’s already sailed past one of two hurdles any of them need to clear to do that: Its Volvo Cars division already has an assembly plant in South Carolina that would let it build its cars domestically. That plant now builds the Volvo EX90 electric SUV and the Polestar 3 that shares its platform; it will add the Volvo XC60 late this year.

Few US buyers realize it, but Geely has controlled the Swedish automaker for more than 15 years. Is the Volvo factory a possible foothold for Geely’s future in the US?

Software from “countries of concern”

The challenge for Geely is that any Chinese company selling cars in the US must certify that all software for autonomous driving, and all hardware and software for connectivity and telematics, has not been developed in or controlled by a “country of concern” (China or Russia). Those strictures sit within a Commerce Department rule that took effect in March 2025; it’s intended to protect the country from hostile use of advanced auto technologies by adversary countries.

Initiated by Joe Biden’s administration, the rule was finalized by Donald Trump’s second administration. The provisions on software take effect next month for cars in model years 2027 and later. The hardware ban follows in three years, starting with model year 2030 cars. The bans sought to avert several types of national security risk.

One goal was to prevent an adversary government from gaining access to the cars via telematics to control their movements — whether taking over the driving remotely or mass-disabling them. Another was to ensure the vehicles could not gather personally identifiable information on people of interest in the US (corporate executives, members of the military, celebrities, and more). The fear is that the vehicles could record voice and facial recognition data, their travel patterns, their contacts, their phone calls, and much more — all of which could be used to train artificial intelligence programs. A final goal, broadly, was to prevent vehicles produced by makers in those countries from capturing detailed images and data on US roads and the built environment through which the cars traveled.

Sam Abuelsamid, vice president of market research at Telemetry, takes a more jaundiced view of this risk. The actions it guards against are “a very inefficient way of spying,” he said, as compared to spy satellites, mobile-phone data, and commercial map data. For all the risks cited, he suggested Americans should be far more concerned about the “data being gathered by Flock and Palantir than anything Chinese cars might send to Beijing.”

Avery Ash, CEO of Securing America’s Future Energy, or SAFE, emphasized the rules were established solely for reasons of national security, and are not concerned with national economic competitiveness. SAFE is a nonprofit think tank that advocates and lobbies for policies that reduce the US dependence on oil; it sees connected, autonomous EVs as an important pathway toward that goal. Data storage locations and governance remain concerns, Ash said, so the rules take a very broad look at what constitutes “control” by a Chinese entity. Where the software was designed is irrelevant; it’s about effective control of the company that wrote it and the firms that designed the hardware it runs on.

The rules do not “cover Chinese software developed before the new rules took effect, so long as it was not being maintained by a Chinese firm,” according to the Commerce Department. But, Ash noted, any updates to grandfathered software happening after the effective date would fall under the new rules.

So… what about Volvo?

The ban on Chinese vehicle software leads inevitably to the question of Volvo. The public face of the company is entirely Swedish, but Geely has controlled 80 to 100 percent of the company since 2010. Sources who attended last year’s hearings on the Commerce rules felt Volvo may be able to find a way to address the department’s concerns. Since it’s seen as a European carmaker and a household name in the US, it will likely be able to continue building and selling cars here.

Indeed, those sources suggested, Volvo is likely already in talks with Commerce about its future vehicle plans, the software in those vehicles, and what it would take to get explicit approvals that certify it adheres to the rules for each model year of every model it sells.

“Volvo Cars follows government rules in all markets where we operate and is committed to working with US Commerce on the technical aspects of the rule,” company spokesperson Thomas McIntyre Schultz said in a statement. “We are having a productive dialogue with the Department of Commerce which is ongoing.”

To date, Ash said, no waivers to the rules have been issued. Any company that wanted one would have to request a “specific authorization” for vehicles, software, and/or components that would otherwise be barred — and a separate authorization is required for each separate model, for each model year. Companies self-document and self-attest to their compliance, he noted, but the department can review any such claims, request documentation, and test the vehicle. (This is similar to how US vehicle-emissions certification has worked for 50 years.)

But if future Zeekr or Geely models use the same platforms and software as those Volvo will continue to build in the US, the Chinese company may say: We just want to put different sheet metal on the same car that’s already received a waiver, and sell it under another name. That’s what Polestar now does with its 3, largely a Volvo EX90 underneath. Why can’t we do that?

Software rules in peril?

Concern arose last month when it became public that the architect of the policy, Elizabeth Cannon, recently resigned as executive director of the Office of Information and Communications Technology and Services in the Commerce Department. (Reuters reported she was “pushed out” by the Trump administration.) White House spokesperson Kush Desai subsequently told The Atlantic Cannon’s departure “should not be read that deeply into as reflective of broader Administration thinking or decision-making.”

That followed shortly after a new National Defense Strategy on January 23rd that made little mention of China as an adversary. Reversing long-standing US defense policy, the new strategy uses remarkably conciliatory language toward China, focusing on “mutual respect” and “establishing trade.”

One insider echoed Desai’s take, saying no evidence to date has emerged that suggests the current administration intends to ax the Commerce rules. The longer-term concern, that person said, will be whether Commerce will stick to its guns and require “robust, rigorous” review of the software, where it was developed, what entity controls its source code, and where all US data is stored. Will specific authorizations only be issued once those issues have been thoroughly vetted — or will Chinese makers be let off with a “soft review” that leaves some risks open?

A recent example of soft review, the source said, was the deal under which a controlling interest in TikTok’s US activities passed to a joint venture in which owner ByteDance retained a stake. While the announcement touted “comprehensive data protections, algorithm security, content moderation, and software assurances for US users,” security professionals suggest significant loopholes remain.

Geely has a great deal of work to do before it can sell cars in the US under its own brands, but it remains the Chinese automaker best positioned to build cars in the US. As it stands today, Geely will still need to convince the US government its software isn’t Chinese at all — and was developed by some entity not controlled by a Chinese firm. At a Senate Commerce Committee hearing this week, Waymo officials were grilled over the company’s decision to use a decontented Zeekr RT as its next robotaxi platform, with one Republican senator griping that the Alphabet-owned company’s assurances that the vehicle met all relevant mandates “ridiculous.”

It would take roughly three years to add a new vehicle, from a different brand, to Volvo’s South Carolina factory. Given the fast-moving nature of the current political situation, whether Geely will in fact have to prove its software complies with the Commerce rules during that entire period — for every model, for every model year — remains up to the changing imperatives and whims of the administration.

Meanwhile, Geely’s targets for localized production are hardly limited to the US. While its Volvo unit will have European assembly plants in three countries by the end of the year, Ford and Geely are reportedly in discussions about Geely potentially using excess Ford factory space in Europe to produce vehicles for that market. And the leading autonomous vehicle company in the US, Waymo, is using a decontented Zeekr RT as its next-generation robotaxi.