Every few years, a founder with a big reputation and a bigger appetite for capital picks a frontier and the money follows. Right now, that frontier is robotics and autonomy — and one of the names attached to it is familiar enough to make even skeptics look twice.
The pitch is seductive: machines that see, move, and act in the physical world, powered by a new generation of AI. The déjà vu is unavoidable. We have watched this movie before, in 2016, when autonomous vehicles were going to reshape cities within 24 months. They didn’t. So the useful question isn’t whether history is rhyming — it clearly is — but what’s genuinely different this time, and what founders should carry forward from the last cycle.
The comeback
Per TechCrunch, Travis Kalanick — the co-founder ousted from Uber in 2017 — is reportedly back, building a new robotics company as capital and talent wars in autonomous systems intensify in a way reminiscent of the 2016 autonomous-vehicle hype cycle. The report is worth flagging as still developing, but the signal it sends is louder than the details: a marquee operator is returning to a capital-hungry hardware frontier at exactly the moment the money is already rushing in.
The company is Atoms, which Kalanick brought out of roughly eight years of stealth in March 2026. It grew out of his real-estate firm City Storage Systems — the parent of ghost-kitchen operator CloudKitchens — and targets what he calls the “digitization of the physical world”: industrial robotics, autonomous mining and transport, built around a common “wheelbase for robots.” Crucially, Kalanick has been explicit that Atoms is not chasing humanoids but “specialized robots” for industrial-scale work. He is also on the verge of acquiring Pronto, the autonomous-haulage startup founded by former Uber colleague Anthony Levandowski, whose systems already move limestone at quarry sites with Komatsu (TechCrunch).
And it is rushing in. Autonomy — humanoids, warehouse robots, delivery bots, self-driving stacks — has become one of the loudest magnets for venture and strategic capital, and the competition for a thin bench of specialised engineers has turned into a genuine talent war. Robotics researchers, perception specialists, and reinforcement-learning talent are being courted with the kind of packages that used to be reserved for pure-software AI labs.
If you were paying attention a decade ago, none of this feels new. In 2016, ride-hailing giants, carmakers, and a swarm of startups poured billions into self-driving, hoovered up robotics PhDs, and promised near-term deployment. The cycle we’re in now has the same shape: a charismatic thesis, a land grab for talent, and valuations that price in a future still several breakthroughs away. History rhyming isn’t automatically a problem. But it is an invitation to ask harder questions than the last cohort did.
What’s different this time
The honest answer is: quite a lot, and it matters.
The most important shift is the underlying AI. The 2016 cycle ran on hand-engineered perception stacks and brittle rule-based systems that struggled with the messy long tail of the real world. Today’s resurgence, as TechCrunch has argued, is fuelled by advances in physical AI and world models — AI systems that learn representations of how objects, forces, and environments behave, rather than relying on rigidly coded assumptions. That gives robots a shot at generalising across unfamiliar situations instead of failing the moment reality deviates from the script.
The supporting hardware has matured too:
- Better models: foundation-model approaches that transfer learning across tasks, closing the gap between one-off demos and adaptable behaviour.
- Cheaper, better sensors: lidar, cameras, and compute that cost a fraction of what they did in 2016, improving the odds — though not the certainty — of workable unit economics.
- Simulation and data: the ability to train in simulated worlds before touching physical hardware, compressing iteration cycles.
All of which is real progress. But here is the caveat that should govern every valuation in this space: as TechCrunch put it in framing 2026 as the year AI moves ‘from hype to pragmatism,’ the persistent gap between an impressive demo and reliable, cost-effective deployment remains the key risk. A robot that folds laundry in a controlled lab is a fundamentally different thing from a fleet that works profitably, safely, and unattended across thousands of unpredictable real-world sites. The demo is the easy 80%. The deployment is the brutal, expensive, years-long 20% that decides who survives.
The cautionary notes
The last cycle left a clear set of lessons, and they’re worth stating plainly because the current enthusiasm has a way of drowning them out.
Hype outruns reliability and cost. The 2016 AV boom collapsed not because the vision was wrong but because the timeline was fantasy. Reliability at the edge cases — the child chasing a ball, the faded lane marking, the unusual weather — turned out to be the whole game, and it took far longer and cost far more than anyone admitted. The same trap is open now. A robot that succeeds 95% of the time sounds impressive and is commercially useless if the remaining 5% requires a human babysitter.
Timelines are long and capital is punishing. Robotics is not software. You cannot ship an update over the air and forget a hardware recall. Building, iterating, and scaling physical systems consumes enormous capital over many years before revenue arrives — and in a tighter funding environment, patient capital is harder to find than it was in the zero-rate exuberance of 2016. Founders should assume the road is longer and more expensive than their deck says.
Who survived, and why. The companies that came through the last AV cycle largely weren’t the ones chasing the grandest ‘robotaxi everywhere’ vision. They were the ones that narrowed: firms that picked constrained, high-value domains — mapped highway corridors, fixed warehouse routes, specific industrial tasks — where the operating environment was controlled enough to make reliability and economics tractable. The survivors also tended to have deep-pocketed backers or strategic parents willing to fund the long slog. Breadth killed; focus and endurance survived. That’s the pattern to internalise before committing to a moonshot framing.
The India read
For Indian founders and operators, this cycle reads differently than it does in San Francisco — and that difference is an advantage if it’s used well.
India is unlikely to win the capital-intensive, winner-take-all humanoid race being waged with Silicon Valley cheque sizes. But that was never where the near-term value sat. The opportunity here is in applied robotics and autonomy aimed at concrete, unglamorous problems — the domains where reliability in a bounded environment beats general-purpose ambition.
Consider where the constrained, high-value use cases actually are:
- Warehousing and logistics: as e-commerce and quick-commerce scale, structured environments like fulfilment centres are ideal for robots that operate within known layouts.
- Manufacturing and inspection: India’s manufacturing push creates demand for automation on repeatable, high-volume tasks where the ROI is measurable.
- Agriculture and infrastructure: autonomous monitoring, spraying, and inspection where labour is scarce and the environment, while outdoors, is more predictable than open-road driving.
- Enterprise and defence-adjacent autonomy: drones and ground systems for surveying, mapping, and security in controlled operating zones.
The lesson from the last hype cycle applies with extra force in a market that can’t rely on infinite capital: pick a narrow domain, prove reliability and unit economics there, and expand outward — rather than pitching a general-purpose robot and hoping the cost curve saves you. Cost discipline, which Indian startups are often forced into anyway, turns out to be exactly the muscle this cycle rewards.
Kalanick’s reported return will pull attention, and inevitably money, toward the biggest, boldest robotics visions. That’s fine — frontier bets are how breakthroughs happen. But the founders most likely to still be standing when the hype settles are the ones who treat 2026 the way the survivors treated 2016: as a moment to build something narrow that actually works, not something sweeping that merely demos well. The technology is genuinely better this time. The temptation to overpromise is exactly the same.
