TOKYO — The garage is in Tokyo, the ambition is global, and the chips are increasingly from Santa Clara’s other chipmaker. Turing, the Japanese startup developing fully autonomous driving software, has added AMD Ventures to its investor list and begun running part of its artificial-intelligence work on AMD graphics processors, according to Bloomberg and the company. The five-year-old company, which built its entire computing stack on Nvidia hardware, now handles roughly 10% of its AI training on AMD chips, executives said.
The shift is both technical and strategic. Turing’s executives said AMD offered two things Nvidia could not: a second source of supply, and lower costs. Self-driving development is among the most compute-hungry businesses in technology, with each new version of a driving model requiring thousands of GPUs to train, and the company said cheaper computing is a direct path to price competitiveness in a market where capital is the moat.
The money behind the move is now public. Turing raised $79 million in an extension of its Series A round, comprising $43 million in equity from AMD Ventures, Mitsubishi Corporation, MUFG Bank, Super Micro Computer, and others, plus a $36 million loan facility from MUFG. Combined with the $95 million first close announced in November, the round totals $174 million, and the company is valued at about $600 million, according to people familiar with the matter.
The scale is modest by the standards of the self-driving industry, where rivals have raised billions, but Turing argues that its approach is different. The company builds end-to-end neural networks that map camera and sensor inputs directly to driving decisions, an approach that requires less hand-engineered logic and, its founders say, less data infrastructure than traditional stacks. Founder and chief executive Issei Yamamoto said the company’s technology is the most advanced of its kind in Japan, and that the country’s automotive industry is at stake.
Japan is the terrain where the bet will be tested. Turing is targeting a commercial launch of its software in consumer vehicles and driverless robotaxis as early as 2028, and it has partnered with the auto supplier Denso to prepare for production. The competitive pressure is building: a partnership of Nissan, the British startup Wayve, and Uber plans to pilot a self-driving taxi service in Tokyo by the end of this year, and global players are circling a market that Japan’s government wants to open to robotaxis.
For AMD, the investment is a foothold in an industry Nvidia has dominated. Nearly every self-driving project in the world runs on Nvidia’s chips and its CUDA software platform, and AMD has been working for years to break that grip, from data-center accelerators to the automotive market. An equity stake buys AMD something a chip sale alone cannot: a seat at the table as Turing decides which computing platform will run its next generation of models.
The 10% share of training is small, and Turing has not said whether AMD will take a larger role in inference — the computing that runs inside the car in real time, where Nvidia’s automotive platform has been entrenched for a decade. The company said the split reflects a deliberate diversification strategy rather than dissatisfaction with Nvidia, and that it will keep both options open as volumes grow.
Autonomous driving has been Nvidia’s market to lose. The company’s automotive platform runs everything from advanced driver assistance to full robotaxi fleets, and its chips are the default choice for startups and incumbents alike. That concentration is exactly why Turing’s shift matters beyond the company itself: every design win AMD takes in this industry chips away at the assumption that Nvidia is the only viable supplier for safety-critical AI.
Japan’s government has made robotaxis a national project, easing rules and funding pilots in cities from Tokyo to the country’s smaller municipalities, and the market is attracting global players. The Nissan-Wayve-Uber partnership plans a Tokyo pilot by the end of this year, and Chinese robotaxi operators have signaled interest in Japanese cities. Turing’s path to market runs through partnerships like the one it signed with Denso, which gives it access to automotive-grade manufacturing and the certification expertise that software companies lack.
The investment round also signals where AMD is placing its bets. The company has pushed into automotive computing with its own platform, and its ventures arm has backed a string of AI companies, from the data center to robotics. An equity stake in Turing gives AMD influence over a startup’s roadmap at the moment it is choosing its computing foundation — the kind of position that historically has belonged to Nvidia by default.
The question for AMD is whether the foothold grows. Every percentage point of a fast-growing startup’s compute that moves from Nvidia to AMD is a data point in the argument that alternatives are viable, and Turing is a credible reference in a region where AMD has little presence. For Turing, the AMD relationship is a hedge against the concentration risk that has defined the AI industry — and a bet that the cost of computing, not the prestige of the brand, will decide who builds the cars of the future.
