Waymo Proved Robotaxis Work. The Harder Question Is Whether They Actually Pay.

The debate over whether autonomous vehicles could drive themselves is largely over. The debate that decides the industry is just beginning — and it runs straight through the road beneath the wheels.

By Robert James  ·  Founder & CEO, RJ Innovations  ·  Father of Connected Automated Vehicles

500,000+ Waymo paid rides per week

$30–35B Estimated invested over 15+ years

~$16B Recently raised to keep scaling

10–30cm URGNs™ all-weather positioning

Google has spent an estimated $30–35 billion building Waymo over the last fifteen years, and the result is genuinely remarkable: more than 500,000 paid rides every week, real customers, real journeys, no driver behind the wheel. By almost any engineering measure, it is one of the defining technology achievements of our generation.

But I want to ask the question too few people in this industry are willing to ask out loud: is it actually a business?

For a decade, the argument about autonomous vehicles was whether the car could drive itself. That argument is largely settled. The technology works. The argument that matters now is different, and in my view harder: can a robot move a passenger for less than a human driver can — everywhere, in every condition, on a balance sheet that closes?

The technical debate is over. The economic debate is just beginning.

Waymo is now serving north of half a million paid robotaxi rides a week and has set a public target of one million. Those are extraordinary numbers, and they prove the perception, planning, and control problems are tractable at scale. Nobody serious still argues that driverless cars are science fiction.

And yet the company recently raised roughly $16 billion in fresh capital, most of it from Alphabet, and by most accounts still loses money in its most mature markets. Everyone talks about autonomy. Almost nobody talks about returns.

Per-vehicle profitability? Probably reachable, and improving as operators learn which levers to pull on hardware and utilization. A meaningful return on $30–35 billion of investment plus ongoing multi-billion-dollar operating costs? I am far less certain. That gap — between an engineering triumph and a durable enterprise — is the most important open question in mobility today.

Everyone assumes the hardest problem was getting the car to drive itself. What if the harder problem is proving that a robot is actually cheaper than a driver?

The hidden cost stack behind every driverless mile

A human driver is an astonishingly efficient machine. Low capital cost. Self-calibrating. Adapts to the unexpected without a software update. Proven economics at massive scale. That is the benchmark autonomy has to beat — not match, beat — before the unit economics of a robotaxi tilt in its favor.

Now look at what sits behind every autonomous mile. Lidar, radar, and camera arrays. Onboard compute. High-definition maps that must be built and continuously refreshed. Connectivity. Charging and energy management. Depreciation on expensive hardware. Insurance. And a remote-operations team standing by to assist when the vehicle hesitates.

Uber built a platform. Waymo built a platform, a vehicle ecosystem, an operations model, and a transport network — all at once. That is a very different equation, and every element of it shows up as a line item on the cost side of the ledger.

The cost nobody puts on the slide: weather and edge cases

Here is the uncomfortable truth about today’s self-driving stack: it is strongest exactly where driving is easiest. Clear skies. Sunbelt cities. Well-mapped, gently curving grids. The demonstrations that made the world believe were filmed in the friendliest possible conditions.

Move the same vehicle into fog, snow, heavy rain, a tunnel, or an urban canyon and the picture changes. Falling snowflakes can be mistaken by lidar for solid objects. Rain and fog scatter and degrade the sensor returns the vehicle depends on. GPS — still the backbone of most positioning stacks — drifts or drops out entirely between tall buildings and underground.

None of that stops the car. But it makes the car expensive. Every degraded mile means more cautious driving, more disengagements, more remote-operator interventions, higher insurance exposure, and slower claims when something does go wrong. “The car can drive itself” and “the car can drive itself profitably, in any weather, anywhere” are two very different claims — and the distance between them is measured in dollars.

Why this is an economics problem, not just a safety problem Interventions, remote operators, insurance premiums, and downtime are not edge-case footnotes. They are the operating-cost levers that decide whether a fleet clears its cost of capital. Harden the vehicle’s understanding of the road in bad conditions and you move every one of those levers in the right direction at once.

Why a robot isn’t automatically cheaper than a driver

This is the part the headlines skip. Removing the driver removes a wage, but it adds a balance sheet — sensors, compute, mapping, energy, insurance, and a control room. To win, autonomy cannot simply reach parity with a human on a sunny afternoon in Phoenix. It has to be structurally, reliably cheaper across the entire operating envelope, including the messy ten percent of conditions where the technology is weakest and the liability is highest.

Alphabet can afford to wait a decade for that math to resolve. Most operators — the fleets, the OEMs, the mobility companies who will actually deploy this at scale — cannot. They need the economics to work now, in the real weather of real cities.

Fixing robotaxi economics from the road up

This is where I have spent the last several years, and where I believe the industry has a blind spot. We have poured tens of billions into making the vehicle smarter. We have invested almost nothing in making the road legible to the vehicle.

That is the gap URGNs™ — Ultra-Wideband Roadway Geometry Networks — is built to close. We deploy low-cost ultra-wideband markers embedded in the pavement and mounted in tunnels, on poles, and on streetlights. Together they form a network that broadcasts precise road geometry to any equipped vehicle, delivering 10–30 cm lane-level positioning that works in any weather — GPS-independent, unjammable, and built on the same UWB chips already shipping in modern smartphones and 2025-and-newer vehicles.

What that reliable, all-weather ground truth does to the economics is the whole point:

• Fewer interventions and higher uptime. When the vehicle always knows exactly where the lane is, it hesitates less, hands off less, and keeps earning.
• Lower remote-operations load. The single most human-intensive line item in a driverless fleet shrinks when the road itself resolves ambiguity.
• Lower crash rates, reduced liability. That flows straight into meaningfully lower insurance premiums and faster claims.
• A hardened stack. Precise V2I data pushes Connected Vehicle 2.0 far beyond today’s pilots and gives on-board autonomy a dependable reference in exactly the conditions where its own sensors struggle.

And the deployment model is designed for adoption, not friction: a public-private partnership in which private industry fully funds installation and maintains operations for thirty years at no cost to taxpayers, in exchange for licensed access to the data. Smart infrastructure that pays for itself is how impressive technology becomes sustainable economics.

On-vehicle intelligence and smart infrastructure — not either/or

The question I keep putting to the industry is simple: pure on-vehicle autonomy, or smart infrastructure — what gets us there first? My honest answer is that it was never a competition. The winning stack is both. The smartest vehicle on earth still drives better, and cheaper, when the road tells it exactly where the lane is.

We did not make aviation all-weather-reliable with better avionics alone. We built instrument landing systems and ground infrastructure so aircraft could operate safely in conditions their onboard sensors could not handle by themselves. Roads deserve the same philosophy. The vehicle and the road should carry the load together.

Alphabet can wait. Most operators can’t. Smart infrastructure is how we turn an engineering marvel into a business that closes.

So — future of transportation, or expensive experiment?

Waymo may well be building the future of transportation. But whether it becomes the future depends on economics as much as engineering. The debate has to move from “are autonomous vehicles technically possible” to “are autonomous vehicles commercially superior.” Those are not the same question, and the second one is far from answered.

The hardest problem was never getting the car to drive itself. The harder problem is proving that a robot is actually cheaper than a driver — in the fog, in the snow, in the tunnel, and on the balance sheet. I believe we close that gap fastest by making the road itself part of the solution.

Robotaxi economics: frequently asked questions

Is Waymo profitable?

Not yet. Despite serving more than 500,000 paid rides a week and being the only autonomous vehicle company deployed across multiple U.S. cities, Waymo still operates at a loss and recently raised roughly $16 billion in new capital. Per-vehicle economics are improving, but a return on an estimated $30–35 billion of investment plus multi-billion-dollar annual operating costs remains unproven.

Why are robotaxis so expensive to operate?

Every driverless mile carries a hidden cost stack: lidar, radar, cameras and onboard compute, high-definition maps, connectivity, charging, depreciation, insurance, and a remote-operations team standing by to assist the vehicle. A human driver is a low-capital, self-calibrating system, so autonomy has to overcome a large fixed-cost base before it is cheaper per mile.

Can autonomous vehicles operate in snow and fog?

Only partially, and at a cost. In snow, lidar can mistake falling snowflakes for solid objects; heavy rain and fog degrade cameras and sensors; and GPS drifts or drops out in tunnels and urban canyons. The vehicle can compensate by driving more cautiously or handing off to a remote operator, but each of those responses lowers uptime and raises operating cost.

What is Ultra-Wideband (UWB) positioning for autonomous vehicles?

Ultra-Wideband is a short-range radio technology already built into modern smartphones and 2025-and-newer vehicles. URGNs™ (Ultra-Wideband Roadway Geometry Networks) embed low-cost UWB markers in the roadway and mount them in tunnels and on poles to broadcast precise road geometry, giving equipped vehicles 10–30 cm lane-level positioning that is all-weather, GPS-independent, and unjammable.

How does roadway infrastructure lower robotaxi operating costs?

Reliable all-weather ground truth from the road reduces interventions and remote-operator load, raises fleet uptime, and lowers crash rates. Fewer incidents mean reduced liability, which translates into meaningfully lower insurance premiums and faster claims — the operating-cost levers that decide whether a robotaxi fleet is profitable.

Autonomy makes the car smart. Infrastructure makes it profitable.

See how URGNs™ delivers all-weather, GPS-independent, lane-level positioning that hardens the autonomous stack and rewrites the operating economics of every driverless fleet. Learn more at urgns.com.

Partner & deployment inquiries: info@urgns.com  ·  813-815-URGN