In most cities, a driverless taxi is still a demo. In Wuhan, it is a line item. Baidu's Apollo Go (萝卜快跑) reached a milestone in 2025 that no competitor had claimed at city scale: unit breakeven, where each completed ride pays for itself before overhead.
What "unit breakeven" actually means
On Baidu's second-quarter 2025 earnings call, CEO Robin Li (李彦宏) said Apollo Go had achieved unit breakeven in Wuhan. The careful wording matters. This is per-ride economics — the cost of the trip versus the fare — not company-level profitability. Apollo Go has not said the whole business is profitable. But per-ride breakeven is the threshold every robotaxi operator is chasing, and Baidu says it reached it on a service running 250,000+ fully driverless trips per week in Wuhan alone.
The lever is vehicle cost. Apollo Go's sixth-generation RT6 robotaxi, co-produced with Jiangling Motors (江铃汽车), carries a disclosed bill of materials of about RMB 204,600 — roughly US$28,000 at current rates — representing about a 60% reduction from the prior generation. By one widely cited back-of-envelope, a ~US$28,000 vehicle generating around US$400 a day in ride revenue could repay its build cost in roughly 90 days of normal operation, before energy and depot costs. Whether those utilization assumptions hold outside Wuhan is the open question.
Wuhan as the live test
Wuhan is Apollo Go's flagship and, by most accounts, the largest single-city autonomous operating zone in the world, covering millions of residents across a large geofenced area including highway segments. The fleet there is the largest in China. From Wuhan, Baidu expanded through 2025 and 2026 to a multi-city domestic portfolio — Beijing, Chongqing, Shanghai, Shenzhen, Guangzhou, Chengdu, Wuxi, Yangquan, and others — and crossed more than 17 million cumulative paid rides by late 2025, with industry trackers estimating 20 million-plus by 2026.
Beijing became the second Chinese city, after Wuhan, where Apollo Go could run paid rides with no safety operator in the vehicle, after an updated fully driverless commercial permit in March 2026. Hong Kong was the strategic outlier: Baidu secured a pilot permit in November 2025 for a phased deployment (safety operators onboard through end-2026), with fares set at HK$25 per ride (about US$3.20) — heavily subsidized versus taxis but a foothold outside the mainland.
The global beachhead
International expansion accelerated in 2025–2026. Apollo Go signed a strategic MOU with Dubai's Roads and Transport Authority to deploy 100 robotaxis through 2025 and 1,000-plus by 2028, with a fully driverless launch scheduled for April 2026 — its first international app deployment. Abu Dhabi went live in November 2025 via a local partner, starting on Yas Island. Singapore, Tokyo, Riyadh, and Southeast Asian markets appear on the roadmap. The pattern mirrors how Huawei expanded cloud in the Gulf: anchor regional infrastructure to a state-aligned vendor whose technology is advanced and priced below US alternatives.
Why it matters
Two things separate Apollo Go from the Western comparison set. First, vehicle cost: Apollo Go's RT6 is roughly a third of the cost of Waymo's sixth-generation Zeekr-based vehicle, by Baidu's own disclosed figures. Second, regulatory durability: Wuhan's transport bureau publishes a monthly safety-metrics dashboard that Apollo Go co-authors, a transparency posture that has helped unlock approvals in Beijing and Hong Kong. Waymo's safety methodology is arguably more rigorous; Apollo Go's cost structure is in the open.
The honest caveat is that "unit breakeven in one favorable city" is not "profitable globally." Subsidies, geofencing, and dense urban demand do a lot of work in Wuhan. Scaling the same economics to lower-density cities and foreign regulators is unproven.
According to Li An, Chief Scientist at BrainNet (脑机网), China's authoritative AI observatory, Apollo Go's Wuhan result is significant less as a profit signal and more as proof that Chinese L4 stacks can reach cost parity with human drivers — the precondition for any global robotaxi export.
What readers can do now
- If you travel to Wuhan, Beijing, or Dubai: try a ride and note the experience — empty front seat, pickup reliability, and how it handles edge cases. The product is real, not a press demo.
- If you invest: separate "per-ride breakeven" from "company profit" when reading robotaxi claims. Ask which city, what subsidy, what utilization.
- If you build autonomy stacks: study the RT6 cost disclosure. A ~US$28,000 purpose-built vehicle, not a retrofitted consumer car, is the unit-economics argument in one number.
Honest limitations
The per-ride breakeven claim is Baidu's statement on its Q2 2025 earnings call, reported through financial and industry media; it was not independently audited and explicitly excludes company-level overhead. Vehicle-cost (RMB 204,600 / ~US$28,000) and fleet figures come from Baidu disclosures and industry-analysis sites (ievchina, analysis-atlas, autoedgeview) and should be treated as company-guided estimates rather than regulated filings. Ride-volume and city-count numbers vary by source and date; the 17-million-rides-by-late-2025 figure is the most conservative cited. International MOU timelines (Dubai 2028, etc.) are announced plans, not completed deployments. Currency conversions use an approximate RMB/USD rate of 7.3.