Autonomous driving company Pony AI Inc (PONY) posted its second-quarter 2026 earnings on August 18, with total revenue of $36.2 million, a 68.8% year-over-year increase, while GAAP net loss narrowed 14.9% to $45.4 million.

Breaking down the revenue structure, the three business lines showed clear divergence in performance. The standout highlight was the Robotaxi segment, which generated $12.1 million in quarterly revenue, up 691.2% year-over-year, with passenger fare revenue surging 849.3% — both metrics reaching all-time highs.

The Robotruck business delivered $13.3 million in quarterly revenue, up 40.0% year-over-year, primarily driven by freight services in partnership with Sinotrans. According to the company, fourth-generation autonomous trucks have been mass-produced on schedule, and commercial deployment of fourth-generation driverless trucks has commenced at Shenzhen's Mawan Port in collaboration with China Merchants Port.

In contrast, the intelligent solutions segment posted $10.8 million in revenue, up just 3.9% year-over-year. This business was a double-digit growth engine in the same period last year, but has slowed notably this quarter, which the company attributes to fluctuations in the delivery pace of autonomous driving domain controllers (ADC).

The three business lines now display a distinct growth spread of "+691% / +40% / +4%," indicating that the core driver of the company's revenue growth is increasingly concentrating in Robotaxi. Compared to 2025, when growth was jointly propelled by intelligent solutions deliveries and trucking partnerships, Robotaxi passenger fares are becoming an increasingly important component of new revenue.

Operating Leverage — Expense Growth at Just One-Sixth of Revenue Growth

While the 68.8% revenue growth marks Pony AI's strongest quarterly performance to date, what truly deserves attention is the significant slowdown in operating expense growth alongside rapid revenue expansion.

Second-quarter operating expenses totaled $72.1 million, up 11.4% year-over-year — roughly one-sixth of the revenue growth rate. R&D expenses rose 14.7% year-over-year, also notably below the revenue growth pace. On a Non-GAAP basis, operating expenses increased 9.6%, with operating losses of $56.7 million, up 4.9% year-over-year and essentially flat overall.

The clear mismatch between revenue and expense growth is beginning to demonstrate the operating leverage that comes with scale.

This shift also aligns with the company's technology roadmap. The CTO emphasized during the earnings call that PonyWorld 2.0 can support "deploying autonomous taxi fleets across multiple countries and cities simultaneously without requiring proportional increases in engineering resources." Based on Q2 financial data, this technology and operational model is starting to receive initial validation at the financial reporting level.

Specifically, the operating loss margin narrowed from 285.6% in Q2 2025 to 181.5% in Q2 2026 — a 104-percentage-point improvement year-over-year. The net loss margin fell from 248.3% to 125.2%, a 123-percentage-point improvement. Revenue growth is beginning to translate more visibly into margin improvements, with operating leverage gradually emerging.

The Loss Picture — GAAP Narrows, but Attributable Net Loss Widens

However, the "14.9% net loss narrowing" does not fully capture the company's actual loss dynamics.

GAAP net loss for Q2 was $45.4 million, but the net loss attributable to Pony AI Inc reached $59.8 million — a difference of approximately $14.5 million, mainly related to income attributable to non-controlling interests under the co-fleet model.

This means that as the co-fleet business expands, the company's financial statements are seeing more profit distribution impacts related to non-controlling interests, and looking solely at the overall GAAP net loss could easily underestimate this factor.

Additionally, the company recognized a $25 million impairment on long-term investment prepayments during the quarter. Management stated that following proactive strategic adjustments and the introduction of new strategic partners, the company recognized a $25 million impairment provision for certain long-term investment prepayments determined to be unrecoverable.

On a Non-GAAP basis, after excluding $9 million in share-based compensation, $33.4 million in fair value changes of trading financial assets, and the $25 million long-term investment impairment, the company's Non-GAAP net loss for Q2 was $44.7 million, essentially flat compared to the same period last year.

Therefore, this quarter's "loss reduction" needs to be viewed across multiple metrics: GAAP net loss narrowed year-over-year, but Non-GAAP net loss was roughly flat, while non-controlling interests and one-time impairments had notable impacts on final profit performance. Compared to simply focusing on year-over-year net loss changes, the improvement in loss ratios driven by operating cost control and revenue growth deserves more attention.

Fleet and Per-Vehicle Economics — Nearly Doubling in Six Months

Beyond revenue and losses, another core metric the market is watching is Robotaxi fleet size.

As of June 30, 2026, the company's autonomous taxi fleet reached 1,975 vehicles, compared to approximately 1,000 at the end of 2025 — a 97.5% increase in six months.

The company's full-year target is to exceed 3,500 vehicles, covering more than 20 cities both domestically and internationally. This implies an additional ~1,525 vehicles needed in the second half, averaging ~250 new vehicles per month — requiring a significant acceleration compared to the ~160 vehicles per month pace in the first half.

Meanwhile, positive signals are emerging in Robotaxi per-vehicle economics. The company disclosed in its Q1 2026 earnings report that Robotaxis in Guangzhou and Shenzhen have achieved "city-level per-vehicle profitability" (referred to as UE by the company). Notably, Shenzhen's daily net revenue per vehicle reached ¥394 on March 22, with 25 orders that day.

During the Q2 earnings call, the company maintained this position but has yet to provide a clear timeline for achieving company-wide profitability.

The CFO previously stated during the Q1 earnings call that for the Robotaxi business to move from "point" to "surface," the fleet needs to expand from the thousand-vehicle scale to the ten-thousand-vehicle scale to further dilute R&D and management expenses.

Therefore, the core transmission chain the market needs to verify remains clear: per-vehicle UE turning positive → continued fleet scale expansion → scale effects being released → overall company profitability improvement.

Overseas and Uber — Co-Fleet Model Beginning to Deliver

International progress in Q2 results is also noteworthy. The company disclosed that it has signed contracts with Uber to deploy over 2,000 autonomous taxis in Europe. Combined with other overseas partners, the total number of vehicles in the negotiation phase has now exceeded 4,000.

The core of the co-fleet model lies in transferring part of the heavy-asset vehicle purchase costs to partners such as Uber and OnTime, while the company participates in operations through technology licensing and fare sharing. Compared to a fully self-operated model, this approach reduces the pressure of per-vehicle expansion on the company's balance sheet and provides conditions for rapid overseas replication.

Currently, the company is advancing deployment in Luxembourg in partnership with Bolt and Stellantis. In Singapore, mobility services are being offered to the public through ComfortDelGro's Zig app. Public passenger services have been opened in Zagreb, Croatia, and driverless testing has been launched in Dubai.

More notably, the Q2 earnings release explicitly mentioned that revenue contributions from co-fleet models in both China and overseas grew on a quarter-over-quarter basis.

The significance of this statement is that the co-fleet model is transitioning from prior contract signings and scale planning into actual operations and revenue contribution. If overseas fleets continue to materialize, the company's growth logic will evolve from relying solely on domestic Robotaxi fleet expansion toward a dual-engine approach of "domestic scaling + overseas asset-light replication."