The robotaxi industry has reached an inflection point: technical progress, deep pockets and favourable policies are pushing driverless fleets from pilots to city streets, even as a procession of high‑profile accidents is eroding public trust.
Global players and a crowded field of Chinese firms are racing to scale. Alphabet’s Waymo reported a surge in orders and raised roughly $16 billion in early 2026 to fund international expansion; Tesla removed safety drivers from its Austin pilot and is pushing CyberCab production in 2026. At home, Baidu’s Robocar service operates in more than 20 Chinese cities and reports millions of passenger trips and well over 100 million driverless kilometres. Legacy automakers, new energy carmakers and mobility platforms—from Geely and Xpeng to Didi and Hello—have announced partnerships and fleet plans that together signal a broad commercial push in 2026.
The economics that underpin this push are seductive. Several operators say they are approaching single‑vehicle profitability in dense urban markets; analysts at Goldman Sachs and local company results suggest the unit economics of robotaxis could flip positive this year in prime cities. That potential profit horizon helps explain why vehicle makers, ride‑hailing platforms and autonomous‑driving suppliers are bundling technologies, forming joint ventures and committing thousands of vehicles to trials and initial rollouts.
Yet the industry’s safety record has become an unavoidable counterweight. Between 2025 and late 2025, China recorded several alarming incidents: a Robotaxi fire in Beijing traced to the battery compartment, a vehicle in Chongqing driving into an unmarked municipal construction trench and a Level‑4 vehicle in Hunan striking pedestrians at a zebra crossing. In the United States, Tesla’s Austin driverless fleet logged multiple crashes in a few months. These episodes have highlighted vulnerabilities in perception systems, the difficulty of handling ‘‘long‑tail’’ scenarios such as temporary worksites or sudden pedestrian behaviour, and weaknesses in fleet operations and incident response.
Technical, regulatory and operational gaps intersect to explain why these failures occur. Autonomous stacks still struggle with rare or novel situations—construction zones, extreme lighting or unusual human behaviours—that were under‑represented in training data. Sensors have blind spots and algorithms can mispredict intent; some operators have reportedly economised on sensors or testing as they race to scale. At the same time, regulatory frameworks are incomplete. China is moving toward a national L4 testing and operation standard, but many local authorities have relaxed entry requirements to win investment, leaving inconsistent oversight and hazy accountability when accidents occur.
The industry’s hurry to commercialise also exposes build‑and‑operate tensions. Rapid fleet expansion can outpace maintenance, incident handling and the creation of robust remote‑assistance systems. Several firms have acknowledged weaknesses in on‑the‑ground operations and post‑crash remediation. Observers say the remedy is not simply stricter rules but better governance: mandatory data reporting, clearer accident‑liability rules, standardised testing and stronger third‑party audits of safety claims.
What this means for users and regulators is a test of trade‑offs. Robotaxis promise to relieve urban congestion, cut the cost of mobility and ultimately reduce crashes caused by human error. But those benefits will not be realised if public confidence collapses after repeated, preventable incidents or if regulatory vacuum allows unsafe models to scale unchecked. Policymakers, investors and operators face a choice between a prudent, data‑driven ramp and a headlong commercial sprint that risks backsliding public support.
For now, the answer to whether passengers can ‘‘sit with confidence’’ is: not universally. In a handful of cities, with thoroughly validated software stacks, conservative operational design and strong oversight, robotaxis can already operate at acceptable safety margins. But the broader transition from ‘‘can sit’’ to ‘‘feel safe’’ requires clearer national standards, mandatory transparency from operators, sustained investment in long‑tail scenario handling and an industry culture that prioritises redundancy and slow, verifiable progress over headline‑driven expansion.