Pony AI unveils self-improving AI upgrade as it plans 3,000 robotaxi expansion across 20 cities

Reviewed byNidhi Govil

2 Sources

Share

Pony AI has launched PonyWorld 2.0, an upgraded autonomous driving AI engine that enables self-diagnosis and targeted performance improvements. The company plans to deploy over 3,000 robotaxis across 20 global cities, with nearly half overseas, while recently launching Europe's first commercial robotaxi service in Zagreb, Croatia.

Pony AI Introduces Self-Improving Autonomous Driving AI Engine

Pony AI has launched PonyWorld 2.0, a significant AI upgrade to its autonomous driving AI engine that fundamentally changes how autonomous driving systems learn and improve

1

. The upgraded model enables autonomous systems to self-diagnose weaknesses and drive targeted performance improvements, marking a shift toward AI systems that increasingly guide their own training. This development matters because it addresses one of the industry's most persistent challenges: how to scale autonomous driving technology efficiently without relying solely on human-directed improvements.

Source: Benzinga

Source: Benzinga

The system adds three key capabilities that distinguish it from previous iterations: self-diagnosis, targeted data collection in underperforming scenarios, and more efficient training on complex edge cases

1

. Using a structured intention layer, the model evaluates its decisions against outcomes and identifies performance gaps. It then directs human teams to collect relevant real-world data to refine and improve the training process. This approach represents a fundamental evolution in how autonomous driving systems develop, moving from passive learning to active participation in their own improvement cycles.

Enhanced Self-Diagnosis Capabilities Transform Development Process

The self-diagnosis feature enables Pony AI's autonomous driving systems to identify their own weaknesses without waiting for human engineers to detect problems

1

. This capability accelerates the development cycle by allowing the AI to recognize scenarios where it underperforms and automatically flag them for additional training. The system's ability to guide data collection represents a significant efficiency gain, as it focuses resources on the specific situations where improvement is needed most.

Founder and CTO Dr. Tiancheng Lou emphasized the strategic importance of this shift: "PonyWorld 2.0 is an important step toward a more self-improving approach to autonomous driving development. As AI systems become more capable, they can play a larger role not only in learning to drive, but also in guiding their own improvement -- making L4 development more scalable over time"

1

. The upgrade builds on Pony AI's reinforcement learning framework, which has been developed since 2020, demonstrating the company's long-term commitment to advanced AI methodologies.

Robotaxi Expansion Accelerates With 3,000-Vehicle Deployment Plan

PonyWorld 2.0 is now deployed across Pony AI's driverless fleet and R&D operations, improving safety, ride comfort, and traffic efficiency

1

. The launch aligns with ambitious robotaxi expansion plans to deploy over 3,000 robotaxis across 20 global cities, with nearly half of them overseas. This represents one of the most aggressive expansion strategies in the autonomous vehicle industry and signals Pony AI's confidence in its technology's readiness for large-scale commercialization.

The company recently launched Europe's first commercial robotaxi service in Zagreb, Croatia, in partnership with Verne and Uber

1

. This launch extends Pony AI's robotaxi services into Europe under its dual-engine growth strategy and follows breakeven milestones for Gen-7 unit economics in two Chinese tier-one cities. The achievement of unit economics breakeven in major Chinese markets provides a financial foundation for international expansion and suggests the business model can scale profitably.

Diverse Business Portfolio Beyond Robotaxi Services

While robotaxi services represent Pony AI's most visible operation, the company maintains a diversified business model across three core segments

2

. The robotruck services business develops robotrucks with safety drivers to fulfill freight orders within its existing nationwide logistics network. The licensing and applications business offers personally owned vehicles intelligent driving solutions and other value-added technological services. This diversification strategy reduces dependency on any single revenue stream while allowing the company to apply its autonomous driving technology across multiple commercial applications.

Stock Performance Reflects Market Caution Despite Technical Progress

PONY shares traded at $9.66 during premarket trading, up 3.09%, though the stock faces technical headwinds

1

. The stock is trading 5.5% below its 20-day simple moving average of $10.22 and 21.8% below its 50-day SMA of $12.36, indicating bearish short-term and intermediate-term trends. The relative strength index stands at 37.90, signaling neutral momentum, while the MACD shows a bullish crossover that may indicate potential upward movement if buying interest strengthens.

Despite recent stock performance challenges, analyst ratings remain optimistic with a Buy Rating and an average price target of $21.17

1

. HSBC initiated coverage with a Buy rating and $16.60 target on March 31, while Barclays maintains an Equal-Weight rating with a $10.00 target. The stock carries significant weight in specialized ETFs, including a 4.36% weight in the Roundhill Robotaxi, Autonomous Vehicles & Technology ETF and 2.34% in the SPDR S&P Kensho Smart Mobility ETF, meaning institutional flows through these funds could trigger automatic buying or selling pressure.

Today's Top Stories

TheOutpost.ai

Your Daily Dose of Curated AI News

Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Spend less time searching for the latest in AI and get straight to action.

© 2026 Triveous Technologies Private Limited
Instagram logo
LinkedIn logo