NVIDIA bets on a 'U.S. version of DeepSeek' with plans for open-weight models

A report carried by the Chinese-language outlet 文学城 and surfaced through Google News says NVIDIA is betting on what it describes as a 'U.S. version of DeepSeek' and intends to release open-weight models. The headline frames the move as a strategic bet by the chip giant.

The core signal is directional: NVIDIA is turning toward the open-weight model route rather than only supplying chips and compute to others. Open-weight models are released so that their parameters can be downloaded, deployed and further developed by outside users.

The phrase 'U.S. version of DeepSeek' points to the competitive backdrop. Chinese teams such as DeepSeek previously drew attention with open-weight models that paired strong performance with low cost, and the report places NVIDIA's move in that comparison.

For NVIDIA, moving from compute supplier toward model publisher means competing for ecosystem influence beyond hardware. If open-weight models are widely deployed, they may also pull demand back toward the company's platforms and software stack.

At this stage the disclosure is mostly a headline. NVIDIA has not yet published model scale, release timing or licensing terms, and those details will decide whether this is a positioning statement or a genuine product launch.

The next things to watch are when the model ships, under what license it is opened, and how NVIDIA balances the move against existing closed-model vendors. Each of those factors will shape the competitive map of the open-weight camp.