Jensen Huang landed in Tokyo with a partner list that reads like a map of Japan’s economy: Toyota, Kawasaki, Mizuho, Canon, Fujifilm, RIKEN, SEGA — names covering mobility, robotics, banking, medical imaging, supercomputing, and gaming. This is no ordinary roadshow; for NVIDIA it is a chance to show the world, and investors, what it can do when a market is fully open, the exact opposite of what’s happening in China.

The company chose a precise narrative wrapper, the “sovereign AI” that Huang repeats at every stop: each country should build and run its own AI rather than rent it from abroad. The technical vehicle is the open Nemotron models, released with datasets and training recipes that enterprises can fine-tune on local data and deploy within their own borders. The commercial logic is transparent: locally governable open models still need silicon to run on, and that silicon today is called NVIDIA.

In finance, Mizuho plans what it calls the largest on-premises AI factory in Japan’s banking sector, built on DGX B200 systems — a deliberate choice to keep sensitive data from leaving the building. SMBC and Rakuten Bank are also building foundation models on Nemotron. In healthcare, the Tokyo-1 consortium uses the BioNeMo platform for drug screening and generative design, while Canon and Fujifilm ship next-generation CT systems built on NVIDIA GPUs. Robotics — from Kawasaki’s surgical-support robots to Toyota’s Omniverse simulations — completes the picture of an ecosystem NVIDIA is cementing with the message: your factories, hospitals, and banks won’t need to leave national borders to use AI.

But the most interesting part of the story isn’t on the Tokyo stage; it’s what remains outside the frame. Days before the Japan blitz, NVIDIA tightened compliance for Asian customers, introducing a whitelist that disqualified more than half of past buyers and imposing stricter end-user scrutiny in Singapore, Malaysia, and Japan to prevent chips from being rerouted to China. Japan thus becomes the positive side of the same policy: the market you can freely sell to must be saturated with an offering that ties hardware, models, and services together.

For Southeast Asian markets, courted as data-center hubs but now policed as potential loopholes, this dual role is a warning. The Chinese response, meanwhile, hasn’t stalled: the LineShine system reclaimed the top spot on the June TOP500 list, a reminder that export controls have accelerated — not stopped — domestic development. That explains the urgency with which NVIDIA is locking in allied ecosystems while it still holds a technological lead.

For those tracking deployment logic, the Japanese case offers a concrete testing ground. Banks process financial data on their own premises, hospitals handle diagnostic imaging on-premises, factories train robots on local digital twins. What is presented as sovereignty is also a commercial lock-in strategy: the more an organization adopts the entire stack — from Nemotron models to Metropolis toolkits to Blackwell GPUs — the harder it becomes to migrate elsewhere. How many pilot projects will turn into actual production remains to be seen, but the strongest signals — bank commitments and CT systems already shipping — suggest Japan is now the most advanced laboratory for NVIDIA’s allied-market strategy.