VESSEL GLOBAL CAPITAL / GLOBAL CAPITAL INSIGHTSISSUE 02 · JULY 2026

AUTONOMOUS MOBILITY · CHINA → GLOBAL

Autonomy leaves the lab—and meets the income statement.

PONY AINASDAQ: PONY / HKEX: 2026

Pony.ai’s next chapter is not about proving that a car can drive itself. It is about proving that a driverless network can earn attractive economics at repeatable scale.

For an autonomy founder, the hardest handoff is not from human driver to software. It is from technical possibility to commercial accountability.

Pony.ai was founded in 2016 around a technically audacious goal: build a virtual driver capable of operating across vehicles and geographies. Years of testing, permits and partnerships created technological credibility. Public markets require an additional layer—the ability to connect fleet deployment, paid rides and utilization to margins and cash.

The company’s 2025 results show the business entering that transition. Revenue grew, robotaxi revenue more than doubled, and management reported city-level unit-economics breakeven in Guangzhou and later Shenzhen. Yet consolidated losses remain meaningful. The story has advanced from feasibility to economics, but it has not reached the end of the proof cycle.

Pony.ai’s next chapter is not about proving that a car can drive itself. It is about proving that a driverless network can earn attractive economics at repeatable scale.

02 / THE MACRO TREND

Robotaxis are moving from permit-by-permit pilots to fleet economics.

The industry’s constraint is shifting. Technical validation and regulatory access still matter, but investors increasingly focus on vehicles per city, paid orders per vehicle, revenue per day, remote-assistance cost, fleet uptime and the capital intensity required to add the next thousand cars.

PONY AI, IN PLAIN ENGLISH

Pony.ai sells and operates a software-defined driver across robotaxis, robotruck services, licensing and autonomous hardware applications.

Build

Develop the virtual driver, simulation and vehicle integration stack.

Deploy

Put fleets into approved cities with OEM and mobility partners.

Operate

Earn fares, transport fees, engineering and licensing revenue.

Improve

Use road data and scale to raise utilization and reduce intervention cost.

SCALE + FINANCIAL PROOF

Most recent fiscal year · Company-reported

2025 REVENUE

$90.0m+20.0% year over year

ROBOTAXI REVENUE

$16.6m+128.6% year over year

ROBOTAXI FLEET

1,400+as of March 2026

GAAP NET LOSS

$(76.8)mimproved from $(275.0)m

Can local unit-economics wins become a scalable network—and eventually a profitable company?

WHAT IS WORKING

Robotaxi revenue more than doubled in 2025.
Fare-charging revenue grew close to 400% for the year.
Management reported citywide unit-economics breakeven in Guangzhou and Shenzhen.
The fleet surpassed 1,400 vehicles by March 2026.

WHAT NEEDS PROOF

Consolidated GAAP and adjusted losses remain substantial.
Revenue mix still relies heavily on robotruck and licensing activities.
Scaling fleets requires vehicles, operations, maintenance and regulatory execution.
City-level economics need consistent definitions and repeatable disclosure.

THE DATA, VISUALIZED

The operating evidence behind the thesis.

Comparable periods, explicit units and primary-source definitions. Bar length shows magnitude; labels provide exact values.

REVENUE · US$M

Revenue has grown, but the mix still matters

Company revenue by fiscal year; US$ millions.

2023$71.9m

2024$75.0m

2025$90.0m

READ-THROUGHTotal revenue rose 25.2% from 2023 to 2025. That growth is real, but the endpoint remains small relative to the capital and operational complexity of a scaled autonomous network.

Source: Pony AI 2025 Form 20-F

ROBOTAXI REVENUE · US$M

Robotaxi is growing from a low base

Robotaxi services revenue by fiscal year; US$ millions.

2023$7.7m

2024$7.3m

2025$16.6m

READ-THROUGHRobotaxi revenue more than doubled in 2025, but represented only about 18.5% of total revenue. Commercial autonomy is becoming material without yet becoming the whole company.

Source: Pony AI 2025 Form 20-F

06 / LONG-FORM ANALYSIS

What the numbers change—and what they still cannot prove.

The commercial unit is not the vehicle. It is the city network.

A robotaxi can complete a safe paid ride while the network around it remains uneconomic. The investable unit therefore includes vehicle utilization, dispatch density, charging, cleaning, maintenance, insurance, remote assistance, local operations and the regulatory limits placed on hours or service areas. A headline fleet count is useful only when investors can see how much productive capacity each vehicle generates.

Pony.ai’s city-level breakeven statements are strategically important because they shift the debate from technical possibility toward operating repeatability. But citywide breakeven can mean several things. The market needs to know whether the measure includes vehicle depreciation, headquarters research, stock compensation and the cost of entering the next city. Without a consistent bridge, a milestone can be directionally encouraging while remaining difficult to compare.

FOUNDER TAKEAWAYDefine contribution economics once, reconcile it to the financial statements and use the same definition in every city.

The company has several revenue engines—and each deserves its own economics.

In 2025, robotruck services and licensing and applications still contributed most of Pony.ai’s revenue. That mix can provide useful diversification and fund learning, but it also makes a single consolidated growth rate an incomplete measure of robotaxi progress. Fare revenue, engineering services, transport services, licensing and hardware carry different gross-margin profiles, capital needs and levels of recurring visibility.

A strong public-market narrative would treat those engines as a portfolio. Robotaxi can be measured through paid orders, revenue per vehicle and contribution economics. Robotruck requires customer concentration, route density and fleet ownership disclosure. Licensing needs contract duration, deployment volume and revenue-recognition clarity. The objective is not to force every activity into one model; it is to let investors understand which activity is creating operating leverage.

FOUNDER TAKEAWAYSegment disclosure should follow economic differences, not merely the company’s internal product names.

Scaling faster is valuable only if the funding model scales with it.

Autonomous fleets require more than software. Vehicles must be manufactured or procured, equipped, insured, maintained and operated. Mapping, local permits and service infrastructure add another layer. The cash burden changes materially depending on whether Pony.ai, an OEM, a mobility platform or a local partner owns the vehicle and absorbs utilization risk.

That creates a capital-allocation question: which cities should receive company-funded fleets, and where should partnerships provide the balance sheet? Public investors will reward acceleration when each deployment has a visible return threshold. They will discount expansion if fleet targets rise without a transparent distinction between committed capital, partner capital and operating cash consumption.

FOUNDER TAKEAWAYReport fleet growth beside ownership, funding source, utilization and expected payback—not as a standalone achievement.

Safety disclosure is part of the business model, not a compliance appendix.

Autonomy companies earn permission to operate repeatedly. A safety event can affect permits, customer adoption, insurer behavior and deployment velocity at the same time. That makes safety performance economically material even when the immediate accounting cost is small. The most useful framework combines miles, intervention rates, incident severity and operating-domain context without turning one metric into a claim of universal superiority.

Management should also explain how operational learning changes the system: what is handled by remote assistance, what triggers a vehicle removal, how software releases are tested and how city-specific edge cases move into simulation. The goal is to demonstrate a governed learning loop. Trust rises when investors can see both the outcome and the control system behind it.

FOUNDER TAKEAWAYIntegrate safety, release governance and utilization into one operating scorecard.

An IPO does not complete commercialization; it makes the proof cycle visible.

Public capital can finance vehicles, technology and geographic expansion, but it also imposes a quarterly reconciliation between strategic claims and reported results. For Pony.ai, the crucial bridge runs from fleet and paid orders to revenue, then from revenue to gross profit, contribution economics and cash. Each link should become more explicit as the network grows.

The best investor-relations program will avoid declaring victory from one metric. A larger fleet with low utilization is not scale. Higher revenue with worsening capital intensity is not operating leverage. City contribution breakeven without consolidated progress is not the end state. The credible narrative is a sequence of measurable gates that become harder—not easier—as the company expands.

FOUNDER TAKEAWAYTreat every reporting period as a chance to close one gap between technical scale and financial scale.

07 / TRANSACTION BREAKDOWN

A U.S. IPO paired public-market capital with strategic validation.

Pony.ai’s November 2024 U.S. offering registered 20.0 million ADSs, each representing one Class A ordinary share, with a $11–$13 indicated range. The listing brought the company into a disclosure regime where fleet and economics claims must reconcile with audited financial outcomes.

CAPITAL / SECURITIES20.0m ADSs

⇄

ISSUER / INVESTORS$11–$13 range

NASDAQ GLOBAL SELECT MARKET · SYMBOL PONY

08 / SIGNATURE IR ANALYSIS

The messages management should keep proving.

Define unit economics

Publish a consistent bridge from paid orders and revenue per vehicle to city contribution economics.

Separate the revenue engines

Help investors distinguish fares, engineering, trucking, licensing and hardware economics.

Show deployment productivity

Track fleet size alongside utilization, uptime and cost per autonomous kilometer.

Reconcile scale with cash

Explain the capital, working capital and partner structure required for each expansion model.

Autonomy becomes investable when operational proof is as rigorous as technical proof.

Founders in frontier technologies often assume that better technology will carry the narrative. Public investors need a measurement system that converts technology into utilization, unit economics, cash needs and risk-adjusted expansion.

Pony.ai’s reported city milestones matter because they narrow the gap between pilot and business. The next credibility step is comparability: the same operating definitions, disclosed consistently, across cities and periods.

01Design the investor KPI set before fleet scale accelerates.
02Distinguish partner-funded, company-funded and asset-light expansion.
03Make safety and economics part of one operating narrative.

ABOUT VESSEL GLOBAL CAPITAL

Capital-market readiness is an operating discipline.

Vessel works alongside founders and management teams across IPOs, SPAC and de-SPAC transactions, cross-border M&A, capital raising and PIPEs, investor relations, governance, strategic advisory, readiness and post-IPO strategy.Discuss your roadmap ↗

Warren WangCEO

Pengfei XieSenior Advisor

Guojian ChenChief SPAC Officer

PRIMARY SOURCES

Source notes

IMPORTANT NOTICE

This publication is for informational and educational purposes only and does not constitute investment, legal, accounting or tax advice, an offer to sell, or a solicitation to buy any security. Vessel Global Capital expresses no securities recommendation or price target. Public information is believed reliable but has not been independently verified. Readers should conduct their own diligence and consult appropriate advisers.

← All Global Capital Insights