Turing turns part of its AI training to AMD, indicating a small but growing pushback against Nvidia’s dominance

Turing, a Tokyo-based self-driving technology startup, said it is running a portion of its AI training on AMD chips instead of Nvidia’s. According to a report, the change reduces Nvidia’s share of Turing’s AI training workload to roughly 90%, with the remainder shifting to AMD.

The report also says AMD Ventures has taken a stake in Turing. While details on the investment size were not provided in the coverage, the combination of an equity link and a partial chip swap points to a practical, engineering-focused relationship rather than a purely commercial arrangement.

Turing’s effort appears to be experimental and incremental. The coverage characterizes the company as about five years old and frames the work as testing whether AMD hardware can match Nvidia’s role as the industry default for training large AI models.

In the same report, the key point for chip vendors is not the scale of Turing’s workloads, but what the trial represents. Even if Turing’s deployment is small, it reflects a broader pattern in the AI supply chain, where customers increasingly want options, leverage, and the ability to manage costs and availability across multiple hardware platforms.

For Nvidia, the story is a reminder that its position in AI training and accelerated computing is not locked in permanently. Nvidia has built a dominant market position around its data-center GPU ecosystem, but as AI adoption broadens, more buyers are incentivized to validate alternative silicon for both technical performance and procurement risk.

For AMD, the development fits with its long-running strategy to win more share in AI compute. By backing a startup that is actively evaluating hardware, AMD can gather real-world indicates about how its GPUs perform for training pipelines, an area where benchmark results do not always translate cleanly to day-to-day model development.

Still, major details are missing from the public reporting. The coverage does not disclose the investment amount, whether AMD’s role extends beyond training (such as inference at the edge or in vehicles), or the training outcomes Turing is seeing, including any accuracy, speed, power usage, or total cost-of-ownership comparisons.

What to watch next is whether Turing expands the AMD share beyond the reported 10%, and whether any technical benchmarks or operational metrics emerge. A further indicator will be whether other robotics and autonomous-driving teams follow a similar path, turning chip evaluation trials into longer-term purchasing commitments.