What happened
Nvidia will invest $26 billion over five years to build open-source artificial intelligence models, confirmed by a 2025 financial filing and executive interviews. This initiative includes the release of Nemotron 3 Super, a 128-billion-parameter open-weight model. Nvidia claims Nemotron 3 Super scores 37 on the Artificial Intelligence Index (GPT-OSS scored 33) and ranks first on PinchBench for OpenClaw control, incorporating architectural and training innovations for enhanced reasoning and long-context handling.
Why it matters
Nvidia's substantial investment and open-model releases reshape the open-weight AI landscape, offering a US-made alternative to growing Chinese model adoption. For platform engineers and CTOs, this provides access to high-performance models optimised for Nvidia hardware, strengthening existing infrastructure dependencies. Procurement teams must evaluate the long-term cost implications of building on these models, balancing open-source flexibility with hardware lock-in, especially as rival open models emerge from non-Nvidia hardware ecosystems.
