'Unfinished Business': FT Reveals Kalanick's Atoms Is Building Robotaxi Tech With Levandowski

Nine years after investors forced him out of the company he founded, Travis Kalanick is going back to robotaxis. A Financial Times investigation published in the early hours of September 6 reports that Atoms — the physical-AI holding company Kalanick has been assembling in stealth since his 2017 ouster — is actively developing robotaxi technology, and that the effort is being led by none other than Anthony Levandowski, the controversial autonomy pioneer Atoms hired after acquiring his startup Pronto earlier this year.

The FT, citing sources familiar with the matter, also confirmed that Uber has invested $100 million in Atoms. The company Kalanick was dragged out of is now helping to bankroll his return to the exact market he once bet the company on — and lost.

The story in one paragraph

In March 2026, Kalanick emerged from eight years of stealth to rename his holding company (formerly City Storage Systems, parent of CloudKitchens) as Atoms, organized around three divisions: Atoms Food, Atoms Mining, and Atoms Transport. In April, Atoms completed its acquisition of Pronto, the autonomous haulage specialist Levandowski co-founded in 2018, making it the core technology engine of Atoms Mining. In July, the company raised a $1.7 billion round led by Andreessen Horowitz, with Uber participating. And now, per the FT, the Transport division is quietly building robotaxi software — a full-circle moment for a founder whose original self-driving unit at Uber became one of the most expensive failures in Silicon Valley history.

Why this matters

The robotaxi market Kalanick is re-entering bears little resemblance to the one he left. In 2017, Uber’s autonomous unit was racing Waymo (then a Google project) in a winner-take-all land grab that ended in a trade-secrets lawsuit, a $245 million settlement, and eventually the 2018 death of Elaine Herzberg in Tempe, Arizona — the first pedestrian fatality involving a self-driving test vehicle. Uber sold its self-driving division, ATG, to Aurora in 2020.

Since then, the market has matured into a real business. Waymo now runs commercial driverless services across multiple US cities. Tesla launched its Cybercab robotaxi in Austin, registering 45 vehicles with the Texas DMV before a Gigafactory Texas debut. Uber itself has committed $10 billion to robotaxi fleets per FT reporting, striking hardware partnerships with Wayve, Rivian, and others rather than building autonomy in-house. That strategic reversal — from building to buying — is precisely the gap Atoms appears to be aiming at: if Uber won’t develop its own stack, someone will sell it one.

The FT report gives the first hard confirmation that Atoms Transport’s “wheelbase for robots” vision extends to passenger robotaxis, not just the logistics platforms the company has publicly described. It also resolves a question that has hung over the company since March: why hire Levandowski — an engineer who pleaded guilty to stealing Google’s self-driving trade secrets before receiving a presidential pardon in 2021 — if you only intended to automate mines and ghost kitchens?

The Kalanick playbook, version 2

Atoms’ structure reads like a retrospective of Kalanick’s career, assembled into a single entity. CloudKitchens — the ghost-kitchen business he started in 2018 — became Atoms Food, digitizing food preparation with real estate and robotics. Pronto’s autonomous haulage system, proven across more than 100 trucks at Heidelberg Materials operations including over two million tons hauled at a Texas quarry in under eight months, anchors Atoms Mining. And Atoms Transport is building what the company calls a “wheelbase for robots” — foundational vehicle platforms for logistics that, we now know, include passenger autonomy.

Kalanick has been explicit about the framing. When Atoms announced its $1.7 billion round in July, he wrote on X: “On many levels, this round is a bit of unfinished business. Fuel to complete the bits-to-atoms story arc we started at Uber, continued at CloudKitchens.” The FT’s reporting suggests “unfinished business” was more literal than the market assumed.

The leadership bench reinforces the pattern. In August, Atoms hired Uber’s former finance chief as CFO, continuing what TechCrunch called “getting the band back together.” Add Levandowski — the engineer whose Otto acquisition brought trucking autonomy into Uber in the first place — and the org chart starts to look like a 2016 Uber reunion with better funding and no board fights.

The $100 million question

Uber’s investment is the detail that transforms this from founder nostalgia into strategic news. Uber spent roughly $2 billion developing its own self-driving cars between 2015 and 2020, then sold ATG to Aurora for $4 billion after the Tempe fatality and a protracted safety review. Since then it has pursued an aggregator model: partner with every autonomy developer, put their cars on Uber’s demand network, and let them fight over utilization.

But 2026 has complicated that strategy. Waymo — Uber’s most important AV partner in several markets — has been steadily verticalizing, and Uber’s shift toward large capital investments in robotaxi fleets (the Rivian deal, the $10 billion commitment) signals that pure aggregation no longer feels sufficient. Investing $100 million in a startup building robotaxi software, founded by the man Uber’s board removed, is either a hedge or an admission that the aggregator moat is thinner than hoped.

For Atoms, the check buys more than cash. Uber’s demand network is the distribution layer every autonomy developer needs, and a commercial relationship with the world’s largest ride-hailing platform would solve the demand-side problem that sank most first-generation AV startups.

Analysis: the crowded field and the long odds

Atoms enters a market with well-funded incumbents. Waymo has a multi-year head start and a safety record regulators trust. Tesla is manufacturing its own purpose-built Cybercab at gigafactory scale. Chinese players like Baidu’s Apollo Go operate large fleets at lower cost. And a wave of well-capitalized startups — Nscale-adjacent compute deals, Figure’s robotics partnerships, the Wayve-Uber alliance — are all targeting the same intersection of physical AI and transport.

What Atoms has that most challengers lack is a proven operator with capital, a clear industrial thesis (“gainfully employed robots” — specialized machines doing paying work, not spectacle humanoids), and now a team whose autonomy credentials are undeniable, if chequered. What it lacks is time: every competitor in the space has been iterating on passenger autonomy for years while Kalanick built ghost kitchens.

The Levandowski question will also shadow the effort. His conviction for trade-secret theft (later pardoned) is the industry’s cautionary tale about moving fast with other people’s IP. Atoms will need to demonstrate — to partners, insurers, and regulators — that its robotaxi stack is clean-room engineering, not recycled Waymo lore.

Still, the FT’s reporting marks the moment the robotaxi story’s most dramatic personal arc turned circular. The founder who was forced out over the chaos of his self-driving program is back, funded by his old company, with his old engineer, building the technology that nearly destroyed Uber once — and might yet define its next decade.