Its Blackwell platform swept MLPerf Training 6.0 results, positioning the systems as the fastest and largest performers in the benchmark round. The company framed the results around the rising demands of frontier model training.
Training infrastructure now shapes how quickly AI teams can iterate, how large their models can become, and whether long runs finish reliably. Benchmark wins therefore matter commercially as much as technically.
The results add another data point to the infrastructure side of the AI race, where chips, networking, systems software, and data-center capacity are becoming strategic bottlenecks.