Pony.ai Gen-7 Robotaxi

What the machine is

The Gen-7 Robotaxi is the autonomous driving system and vehicle combination Pony.ai unveiled in April 2025, and it now makes up the overwhelming majority of its operating fleet. Thirty-four sensors in six categories: 9 lidar, 14 cameras, 4 mmWave radar, detection out to 650 metres, with integrated self-cleaning. Compute is four NVIDIA Orin-X chips at 1,016 TOPS, which Pony calls the world’s first L4 domain controller built on four Orin-X. The kit is fully automotive-grade and designed for a 600,000 km, ten-year service life. In April 2026 Pony announced a next-generation platform on NVIDIA DRIVE Hyperion with dual DRIVE Thor over NVLink at 4,000 FP4 TFLOPS.

The real innovation is not in the sensor stack. It is on the cost line: the Gen-7 kit’s bill of materials is 70% below Gen-6, with compute down 80% and solid-state lidar down 68%. That number determines whether this car shows up in your city far more than any performance spec does.

The platform question needs care. The Toyota bZ4X is the only platform with a stated volume commitment — 1,000-plus units in 2026 — and the only one mass-produced on an OEM assembly line, at GAC Toyota under the Toyota Production System, with the first unit off the line on 2026-02-09. But the November 2025 fully driverless commercial launch ran on the BAIC ARCFOX Alpha T5 and GAC Aion V; the bZ4X only reached start of production in February 2026. So this entry carries “Gen-7 Robotaxi” as the product with the bZ4X as lead platform and the other two disclosed alongside.

Two more notes. Pony.ai’s public website still has no dedicated Gen-7 product page and continues to illustrate sixth-generation vehicles. PonyPilot is the ride-hailing service brand, not the machine; PonyAlpha is a retired name.

Why this should already have been here

Because the catalogue already carries WeRide (1,300 vehicles), Zoox (fleet size undisclosed, only starting to charge for rides in Las Vegas in August 2026), Aurora (USD 2M of quarterly revenue) and Gatik — but not the company with 1,700-plus vehicles running fully driverless paid service in three Chinese tier-1 cities and operating Europe’s first commercial robotaxi line. Ranked by vehicles actually deployed, this was the largest hole in the autonomous-vehicle category.

What “1,700 robotaxis” does and does not mean

That figure is the deployed and produced fleet. It is not the number of vehicles in fully driverless paid service. Pony.ai does not publish a supervised-versus-driverless split, and we could not find one in any independent source. This is the biggest evidentiary gap in the entry, and it should not be papered over.

What is verifiable: Guangzhou, Shenzhen and Beijing have run fully driverless fare-charging service since November 2025, and Zagreb is commercial. Shanghai’s Pudong permit explicitly requires a human in the passenger seat. The Uber Middle East pilots carry a safety operator. And Beijing’s Yizhuang district is, in Pony’s own framing, still in “commercialisation validation” — a Jiemian reporter on the ground found thin ridership, seven of eight riders surveyed saying they would not use it, waits over fifteen minutes, and service suspended entirely during snowfall. One fleet, wildly different maturity by city. Compressing that into a single number misleads.

Profitability is the same story. Per-vehicle breakeven has been reached in exactly two cities, Guangzhou in November 2025 and Shenzhen in February 2026, while the company posted a USD 53.5M net loss in Q1 2026 that widened year over year. The single profitable quarter, Q4 2025, came from securities valuation gains rather than operations.

Whose work it takes

Ride-hail drivers, taxi drivers, airport and rail-station transfer drivers, private-hire chauffeurs.

On scale: China’s Ministry of Transport ride-hail regulatory system recorded licensed driver permits rising from 3.086 million in January 2021 to 7.483 million by October 2024, after which the series stopped being published. A separate 2025 study of China’s blue-collar workforce puts ride-hail driving at roughly 37.24 million people, the second-largest blue-collar occupation in the country after domestic service. That is the largest exposed workforce attached to any machine in this catalogue.

The income evidence has stopped being a forecast. A Wuhan taxi driver, Yao Xinnong, saw his wages fall about 40% after Baidu’s Apollo Go arrived in 2022, and recover when Apollo Go was pulled off the road in spring 2026. A Wuhan taxi company wrote publicly that autonomous ride-hailing “allowed technology to monopolise resources and take away the livelihoods of people at the bottom of society.” Coverage of the 2024 Wuhan driver protests was censored.

Attribution has to stay honest. Guangzhou’s fall from RMB 3.46 to RMB 2.66 per km and Chongqing’s 20.8% drop in daily driver revenue are driven mainly by oversupply: Guangzhou and Shenzhen each added more than 40,000 licensed ride-hail drivers a year, roughly doubling driver counts in four years, and Shenzhen’s regulator now calls the market already saturated. A new rule effective 2026-06-01 caps passenger-transport driving at eight hours in any 24, which will compress incomes further. Robotaxis are an accelerant on that, not the sole cause.

The jobs created belong in the ledger too. Pony CFO Wang Haojun says one remote operator can manage 15 to 20 vehicles at scale. Against a 3,500-vehicle fleet that implies roughly 175 to 235 remote operators per shift — our arithmetic, not a disclosure — or about one new remote role for every 15 to 20 driver jobs displaced, plus maintenance, cleaning, charging and depot staff. That ratio is itself the finding.

Why we care for LostJobs

Because Pony.ai’s CEO Peng Jun has already said it more bluntly than we would: “a robotaxi with a driver is meaningless.” That is not a hazy technology vision. It is a stated intention to remove the seat.

Meanwhile Beijing is trying to hold two incompatible goals at once. The state-run Workers’ Daily has editorialised on labour rights “facing new challenges” in the AI age; the government has pledged to “comprehensively address the impact of new technologies such as artificial intelligence on employment”; Bloomberg framed the tension precisely — how China leads in AI without fuelling unemployment. When a country’s industrial policy and its employment policy start fighting over the same machine, that machine has stopped being merely a transport product.

The practical takeaway for a reader is this: to judge when robotaxis reach you, do not watch disengagement rates or demo videos. Watch revenue per vehicle per day. That RMB 394 in Shenzhen is the number recorded on the day the curve crossed the human wage line.