Self-driving's next contest is who can actually build it at scale

Build the car and the driver together

Toyota and Waymo announced the next phase of their autonomous driving collaboration, the companies say, following a preliminary agreement signed in April 2025. Initial testing of the autonomous Toyota bZ4X fleet will begin in Japan in 2027, with deployment to the United Kingdom planned for the years following 2027, and the companies aim to scale Level 4 autonomous vehicles globally. The timeline is the story here. A 2027 test start, with a second market after that, is a production schedule, not a demo date, and it forces the usual questions about manufacturing, servicing, and regulation to be answered in order. m:…1218

Toyota will build the autonomous vehicles at its Takaoka plant in Toyota City, Aichi Prefecture, Japan, the companies say, with vehicle assembly and AV hardware integration occurring under one roof. The custom-outfitted bZ4X includes rear legroom, automatic doors, a power liftgate, and ambient lighting, details that signal a vehicle designed for passengers rather than for engineers. The companies also say they will explore autonomous driving technology for personally owned Toyota vehicles and collaborate with Woven by Toyota on safety initiatives. Waymo's Dmitri Dolgov claims Waymo is up to 20 times safer than human drivers regarding serious crash injuries in areas where it operates, a company claim that will be tested by exactly the kind of deployment this partnership is planning. m:…1218

Or sell the stack to everyone else

Mobileye represents the other route to scale: do not operate the fleet, supply the makers who will. The company says it reported Q3 2026 revenue of $492 million, a 2% decline from Q3 2025, with a Q3 2026 GAAP operating loss of $82 million and adjusted operating income of $89 million, and that Q3 operating income benefited from a research and development Law incentive grant. Full-year 2026 revenue guidance is $1,995 million to $2,020 million. Revenue down while the company guides to roughly $2 billion for the year is not a growth story in the headline number, but the volume story underneath it points the other way. m:…8670

EyeQ shipment volume increased 11% in the first nine months of 2026 compared with 2025, the company says, against an estimated 1% decline in global automotive production over the same period, and the company expects worldwide volume in the L2+ highway hands-free segment to grow at a 20% compound annual growth rate through the end of the decade. It also says it secured new design wins for Cloud-Enhanced ADAS with a Top 5 global automaker from Asia and a Top 10 global automaker from Japan, that its contract to supply Cloud-Enhanced ADAS for Stellantis was expanded to include European platforms, and that its customers began autonomous ridepooling for the public in Hamburg, Germany. The supplier model wins slowly, one design win at a time, and loses slowly too: each win locks in a platform for years. m:…8670

The chip underneath both routes

Black Sesame Technologies shows the third route: supply the compute that both the partnership model and the supplier model need. The company says its Huashan A2000 chip family was honored as an AITX 2026 Hot Innovative Technology Case, covers an equivalent computing power range of 200 to 1000 TOPS, and meets the ASIL-D automotive safety level. The company also says it showcased the Tianyan L3 autonomous driving domain controller based on dual A2000 chips. An award from an exhibition committee is marketing, but the safety level and the dual-chip controller are engineering claims that matter for any automaker deciding whether a chip can be trusted with a driving task. m:…3677

The company says the Jiushao NPU supports multi-precision mixed operations including INT4, FP8, and FP16, features a 100MB-level dedicated high-speed on-chip cache with 8TB/s bandwidth, and that the Shanhai toolchain supports VLA and world models, with its Xingmou ISP maintaining stable imaging under extreme lighting conditions including backlighting, extreme darkness, and glare. The company's Deng Kun argues that perception, decision-making, and control algorithms will run on a unified computing foundation. Whether that unification happens on one chip or across several, the implication for buyers is the same: autonomy's bottleneck is increasingly a bill of materials and a factory schedule, not a research breakthrough. Toyota builds the car, Mobileye sells the stack, Black Sesame ships the silicon, and the winner will be whoever turns those pieces into vehicles on roads first. m:…3677