NVIDIA unveils Minerva AIOps and DSX POD containerized AI factory at GTC Berlin

At GTC Berlin, NVIDIA focused on making AI factories faster to build and easier to run. Minerva is an AIOps platform that combines infrastructure telemetry and workload telemetry so operators can find faults and keep expensive clusters running.

DSX POD literally puts the AI factory in a container. It is a prefabricated Vera Rubin reference design that 17 partners will deliver starting in early 2027. NVIDIA says it cuts onsite deployment from roughly six months to 14 days, because integration happens in the factory rather than in the data center.

The pitch targets this week's dominant constraint. AWS, Azure and Google all say demand exceeds capacity into 2027, and community commenters cite power, space and HBM as limits. Faster deployment directly converts NVIDIA's chip supply into usable capacity. NVIDIA also announced an expansion of its Firebird work in Kazakhstan.

The caveat is that DSX POD does not ship until 2027, and containerization does not solve grid power availability. Watch partner pricing and the first live sites.