The $0.25 Ride: How Robotaxis Will Cut Transport Costs by 90%

The Price Trajectory: Wright's Law Applied to Transport

The $0.25-per-mile robotaxi projection is not speculation. It is the output of a cost curve that has precedent across every technology that has ever been manufactured at scale.

Wright's Law — first articulated by aeronautical engineer Theodore Wright in 1936 — states that for every cumulative doubling of units produced, costs fall by a predictable percentage. It held for aircraft manufacturing. It held for semiconductor production. It held for solar panels, which declined from $100 per watt in 1976 to $0.20 per watt by 2023: a 99.8% cost reduction over 47 years of compounding.

Robotaxis are now on that curve.

The current cost structure, per ARK Investment Management LLC (2026):

Platform	Incremental Cost Per Mile	Status
Human-driven ride-hail (US)	$2.80	2025 actuals
Personal car (US, all-in)	$0.80	2025 actuals
Human-driven ride-hail (China)	$0.50	2025 actuals
Waymo 5th Gen	~$1.05	2025, early launch
Tesla Model Y (refreshed)	~$0.68	2025, early launch
Waymo 6th Gen	~$0.30	2030 projection, at scale
Tesla Cybercab	~$0.17	2030 projection, at scale
Robotaxi global average	$0.25	2035 projection

Source: ARK Investment Management LLC, 2026.

The trajectory runs from $2.80 to $0.25 — an 89% reduction. Every doubling of cumulative robotaxi miles driven moves costs further down that curve. The miles are already accumulating.

What Is Already Happening in San Francisco

Market share data is a more honest indicator than technology demonstrations. Demonstrations prove capability. Market share proves adoption.

Waymo's San Francisco trajectory:

August 2023: ~0% of San Francisco operating zone rides
May 2025: ~25% of San Francisco operating zone rides

In the same period, Uber declined from approximately 65% to 55% market share in the same geography.

This is not a pilot programme. A pilot programme does not move incumbents' market share by ten percentage points in their own flagship market. This is commercial competition — and Waymo has gone from market entrant to top-quartile player in under two years.

San Francisco is not a representative sample of global transport demand. It is a specific, high-density, high-income, regulation-permissive urban environment that Waymo chose as its proving ground. The conditions that enabled 25% share in San Francisco are not present everywhere simultaneously. But the mechanism — hardware cost reduction, software improvement, utilisation efficiency — is geography-agnostic. What happened in San Francisco will happen in other cities on a lagged timeline.

The Unit Economics: Three Sources of Cost Decline

Waymo's 5th Generation vehicle costs approximately $1.05 per incremental mile to operate. That is still above human-driven ride-hail in the US. At $0.17 per mile, the Cybercab at scale would be 94% cheaper than today's ride-hail and 79% cheaper than driving your own car.

The cost decline from $1.05 to $0.17 comes from three compounding sources.

1. Cheaper hardware. Waymo's 6th Generation sensor suite costs approximately 80% less to manufacture than its 5th Generation equivalent. LiDAR, cameras, and compute hardware all follow semiconductor-style cost curves. The specialised hardware that cost $150,000 per vehicle five years ago is approaching $15,000 per vehicle today and will continue declining.

2. Higher utilisation. A human Uber driver works approximately 8 hours per day. A robotaxi, absent regulatory constraints, can operate 20+ hours per day. The same hardware asset generates 2.5x more revenue per 24-hour period than a human-operated equivalent. Depreciation, financing, and fixed costs are spread across far more revenue miles.

3. No labour cost. Labour — the human driver — represents approximately 60–70% of the cost of a human-driven ride-hail trip. Remove it, and the cost base restructures entirely. The remaining costs are hardware depreciation, software licensing, fleet management, insurance, and energy. All of these are declining or manageable at scale.

These three forces are not independent. They compound. Cheaper hardware enables deployment at scale. Scale increases cumulative miles. More miles drive software improvement. Better software reduces incidents. Fewer incidents lower insurance costs. Lower insurance costs improve unit economics, which enables further price reduction, which drives demand, which increases miles. Wright's Law does not require management intervention to operate. It requires production.

Where the Economics Land for Households

At $0.25 per mile, a 10-mile trip costs $2.50.

The current US average Uber trip is approximately $28 — a blended figure across distances, surge pricing, and markets. The math on a $0.25-per-mile world is not incremental. It is structural.

Consider the annual transport cost for an urban household that currently owns one car:

US average vehicle operating cost: ~$12,000 per year (depreciation, insurance, fuel, maintenance, financing)
At $0.25/mile and 15,000 annual miles: $3,750 per year
Net annual household saving: ~$8,250

Those figures are approximate and context-dependent — rural households, multi-vehicle families, and cargo users face different economics. But the directional conclusion holds: the car ownership model that has structured American household budgets since the 1950s ceases to be the economically rational choice for urban households when per-mile costs fall below roughly $0.35.

Transport is 15–17% of US household spending. A 90% reduction in the per-mile cost of urban transport is one of the largest household cost deflation events since the smartphone eliminated the market for standalone cameras, GPS units, alarm clocks, and music players simultaneously.

The Competitive Landscape: Software Wins

The robotaxi ecosystem will not distribute value the way the automotive industry does.

ARK's analysis of the robotaxi value chain projects that technology platform operators — the companies controlling the AI stack and fleet management software — will capture approximately 76% of robotaxi revenue, 97% of EBIT, and 98% of enterprise value in the ecosystem at maturity. Automakers, which provide the physical vehicles, are projected to capture approximately 8% of revenue.

This is not a car business. It is a software and AI business that happens to require cars as hardware input.

The companies currently accumulating the data, the miles, and the software iterations that drive cost curves — Waymo, Baidu Apollo Go, Pony.ai, WeRide — are not building vehicle brands. They are building route coverage, safety records, and AI training datasets that become progressively harder to replicate. Daily robotaxi miles currently in operation (approximate, 2025):

Operator	Daily Miles
Waymo	340,000
Baidu Apollo Go	320,000
Pony.ai	150,000
WeRide	90,000

Sources: Company disclosures; ARK Investment Management LLC, 2026.

Those miles are not just revenue. They are training data. Each mile makes the AI better, reduces intervention rates, improves safety metrics, and lowers insurance costs. The operator with the most miles has the most defensible position — not because of brand loyalty, but because the AI compound interest accrues to whoever accumulates data earliest.

"Waymo has 25% of San Francisco ride-hail. Uber has been watching its market share fall for two years in its own flagship market and is simultaneously its biggest distribution partner. The incumbent and the disruptor are the same company in this transition — which tells you everything about how confident Uber is that it can own the platform layer even as human drivers disappear from it." — The Cricket

Trucking and Last-Mile Delivery: The Same Curve, Larger Market

The cost reduction mechanics that apply to passenger robotaxis apply with equal or greater force to freight.

ARK projects autonomous trucking will reduce truckload delivery cost from approximately $0.07 per ton-mile to $0.03 per ton-mile — a reduction of approximately 57%. The driver cost component of long-haul trucking is higher, as a proportion of total cost, than in ride-hail. Its elimination has an even larger proportional impact on unit economics.

Last-mile delivery — the final leg from distribution hub to door — faces similar economics. Current cost: approximately $15 per order on average. ARK's projection at autonomous last-mile scale: under $1 per order. An 85–93% reduction, depending on density and route efficiency.

For context on the labour displacement implications:

Geppetto tracks delivery robots and autonomous vehicle-adjacent platforms across the full market. Browse the delivery robot category for currently available and commercially deployed systems.

The Broader Economic Impact

Transport cost deflation at the scale ARK projects is not an automotive story. It is a macroeconomic story.

US household spending on transport — including vehicle purchase, fuel, insurance, maintenance, financing, and fares — runs at 15–17% of total household expenditure. For the median US household, that is approximately $12,000–$15,000 per year. A 90% reduction in the per-mile cost of urban transport, fully realised, would be the equivalent of returning $10,000+ in annual purchasing power to every urban household that transitions away from car ownership.

The deflationary effect is not confined to households. Logistics costs are embedded in the price of every physical good. A 60% reduction in truckload cost and a 90% reduction in last-mile cost restructure the delivered cost of goods across the entire retail and e-commerce supply chain. The downstream effects — on real estate values, on retail formats, on urban density — are second and third-order, but they follow directly from the first-order cost reduction.

The $0.25 per mile figure is where Wright's Law puts the robotaxi cost curve by 2035 if production and miles continue to compound. It is already underway in San Francisco. The mechanism is not new — it is the same compounding cost reduction that has restructured every other technology market. Transport is the next one.

FAQs

What does $0.25 per mile mean for a typical ride? At $0.25 per mile, a 10-mile urban trip costs $2.50. The current US average Uber fare for a comparable trip runs $18–$28 depending on market and time of day. The $0.25 figure represents the 2035 projected global average at scale, per ARK Investment Management; near-term costs are higher — Waymo currently operates at ~$1.05/mile incremental cost.

Is the $0.25/mile projection credible? It is consistent with Wright's Law applied to cumulative robotaxi miles and hardware cost reduction curves. Waymo's 6th Generation sensor hardware costs ~80% less than 5th Generation to manufacture. The Tesla Cybercab is projected at $0.17/mile at scale. The $0.25 global average is ARK's blended projection across operators and geographies at maturity. It requires sustained production and deployment — it is not guaranteed — but it is within the range established by comparable technology cost curves including solar, semiconductors, and genome sequencing.

Which robotaxi companies are currently operating commercially? Waymo (San Francisco, Phoenix, Los Angeles, Austin), Baidu Apollo Go (multiple Chinese cities), Pony.ai, and WeRide are all operating commercial services without safety drivers as of 2025. Together they are logging approximately 900,000 miles per day. Tesla's Cybercab service launched in limited form in Austin in 2025.

Does Waymo's 25% San Francisco share mean robotaxis have won? No — it means the technology has proven commercial viability in a permissive, high-density urban environment with specific regulatory support. Scaling to global transport requires hardware cost reduction, regulatory approvals in new geographies, and infrastructure investment that will take years. The market share data establishes the mechanism; the timeline is a separate question.

What happens to human drivers? Approximately 3.5 million ride-hail and taxi drivers operate in the US; the global figure is substantially higher. Full displacement is a long-horizon scenario — regulatory constraints, geographic limitations, and the complexity of edge-case urban environments will preserve some human-driven transport for years. The displacement will be progressive and uneven across geographies. For detailed analysis, see the Geppetto Jobs Index profiles for taxi and ride-hail drivers, delivery drivers, and truck drivers.

How does $0.25/mile affect car ownership economics? At sub-$0.35/mile robotaxi pricing, urban car ownership ceases to be economically rational for households below a certain annual mileage threshold. The US average vehicle operating cost runs ~$12,000/year. At $0.25/mile for 15,000 annual miles, the cost is $3,750 — a saving of ~$8,250 per year. The transition is a function of price, density, and habit change. The economics cross a threshold; the question is when, not whether.