This web summary separates established evidence from the outcome still to be tested. The full thesis PDF retains the complete analysis and source register.
- Research assessment
- Open · monitoring continues
- Record type
- Market catalyst
- Initiation
- Post-event ecosystem baseline
Analyst view
My take,
Driverless transport has reached real service, but its commercial case cannot be settled by adding robotaxi rides to truck miles. Waymo has the strongest reviewed historical safety evidence in its established surface-street domains. Freight operators have paid driver-out activity in selected lanes. The unresolved question is whether each operator can repeat that work safely and economically as its operating domain expands.
Summary
What the evidence establishes
Waymo reports 271.3 million rider-only miles through June 2026. California deployment records show about 4.22 million completed trips in Q2. Kodiak and Atlas report paid driverless work in the Permian, while Aurora has started driverless highway freight. These are operating records, not proof of fleet-wide profitability or unrestricted autonomy. Supervised demonstrations, factory capacity and future vehicle orders remain separate.
How the evidence ties to the question
The decision question
Can driver-out passenger and freight service deliver repeatable, safe and useful transport within a named operating domain, at a sustainable full cost after remote support, empty movement, maintenance and fleet replacement?
Best reported evidence
Reported evidence
The thesis follows six domains separately: U.S. surface-street ride-hailing; freeways and airports; Chinese urban ride-hailing; Permian sand hauling; middle-mile and regional freight; and open-highway Class 8 freight. Each retains its own operator, geography, period and safety definition.
Aurora reported $2 million of Q2 revenue against $7 million of cost of revenue; the revenue includes supervised operations. Kodiak reported $3.499 million of revenue against $10.649 million of truck and freight operating expenses. Neither comparison isolates driverless-domain contribution. Kodiak’s positive net income also included a warrant-related gain, so it is not evidence of operating profitability.
| Month | Rider aboard · million miles | Other service · million miles |
|---|---|---|
| April 2026 | 5.58 | 3.83 |
| May 2026 | 5.75 | 3.95 |
| June 2026 | 5.24 | 3.78 |
Interpretation boundary
N-Ledger calculations and rationale
California records show 28.14 million total Waymo miles against 16.58 million occupied miles: roughly 1.70 total miles per occupied mile, with about 41.1% outside the passenger-carrying period. These figures expose work behind each ride; they do not disclose its cost. A fared-ride flag is not a collected fare.
Economics requires collected revenue and all relevant costs for the same fleet, domain and quarter. Safety requires comparable incident definitions and exposure. Missing remote-assistance frequency, active fleet hours or domain profit stays unknown rather than becoming zero.
| Q2 2026 measure | Value | Basis |
|---|---|---|
| Completed trips | 4,220,075 | Reported |
| Total vehicle miles | 28,143,284.70 | Reported |
| Period 3 miles | 16,579,724.92 | Reported |
| Periods 1 + 2 miles | 11,563,559.78 | Total minus Period 3 |
| Non-Period-3 share | 41.09% | 11,563,559.78 / 28,143,284.70 |
| All miles per Period 3 mile | 1.697 | 28,143,284.70 / 16,579,724.92 |
Independent validation
Independent cross-check
IIHS independently analyzed Waymo’s 2021–2024 experience and found 68% fewer police-reportable crash involvements per mile overall and 81% fewer injury crashes than the matched human benchmark. The study used about 50 million driverless miles, with meaningful city differences and a small Austin sample. Its finding supports that historical scope; it does not validate new platforms, winter weather, freeways or competing operators.
| Q2 2026 line · $M | Aurora | Kodiak |
|---|---|---|
| Revenue | 2.000 | 3.499 |
| Direct operating cost line | 7.000 | 10.649 |
| Loss from operations | 266.000 | 43.677 |
Challenge the thesis
Counterargument
Local success may depend on mild weather, limited routes, remote labor or promotional demand. Scaling could increase empty movement and intervention cost. A software remedy also needs follow-up: school-bus incidents after the December 2025 Waymo recall warrant investigation, but the preliminary record does not establish which software build was involved or a proven recurrence of the same defect.
Peer and market context
Comparable context
Tesla, Waymo and Chinese operators disclose different mile, intervention and injury measures. Raw crash totals without mileage cannot rank them. Aurora and Kodiak operate different freight mixes, while Gatik’s middle-mile work is not interchangeable with open-highway trucking. Vehicle makers, chip suppliers, charging partners and ride platforms belong in the ecosystem map, but their transactions must not be counted again as additional transport output.
Failure conditions
Thesis risks
Unresolved risks include domain expansion, adverse weather, remote-support dependence, fleet replacement, regulatory restrictions and economics hidden inside company totals. Passenger and freight results must remain separate. Suppliers’ production targets, customer memoranda and capacity ambitions cannot establish delivered driverless service.
Why the view holds
Conclusion
Driver-out transport is operating in several bounded settings, and historical Waymo surface-street safety has independent support. The wider safe-and-economic scale-up remains open. The next useful evidence is a repeatable operating series with domain-specific exposure, service quality, intervention burden and contribution after full costs.
Forward measurement
Next expected outcome assessment
Evidence current through October 22, 2026. Revisit quarterly operator filings, CPUC deployment records and Waymo’s next safety release. Year-end fleet and ride targets are company targets, not completed outcomes. Track NTSB findings when published; no release date is assumed. Expected significance is three stars against U.S. commercial road transport. No outcome or realized Depth grade is issued, and there is no fixed terminal date.
Evidence register
Sources
Selected sources for this web summary. The full thesis PDF retains the complete analysis, tables and source register.
Uniform standard
Source-use standard
Reported fact: Directly stated in the cited record.
N-Ledger calculation: Reproducible arithmetic from disclosed inputs.
Scenario assumption: An explicit input used to bound outcomes, not a reported fact.
Supported inference: Sourced links with the remaining weak link identified.
Uniform disclosure
Research assistance
Software- and AI-assisted workflows may support public-record discovery, evidence organization and preliminary checks. Automated output is not an authoritative source. The Founder & Research Lead reviews material evidence and approves published research conclusions.
Uniform disclosure
Opinion and investment-risk disclosure
This report is an independent opinion piece for informational and educational purposes. It is not personalized investment advice, an offer, solicitation, recommendation, price target, rating, or promise of any commercial, market, or investment outcome. Facts are drawn from cited public records; calculations, scenarios, and interpretations are the author's. Views may change as new evidence emerges. Investing involves risk, including possible loss of principal.