Jensen Huang backs open-source AI, while Bank of America argues chipmakers will profit either way

NVIDIA CEO Jensen Huang has thrown his weight behind open-source AI models, arguing that the path forward in artificial intelligence likely involves models that can be widely accessed and built upon. In a market report carried by Yahoo Finance, the discussion is framed as less about which AI assistant is “best” today and more about whether open models can ultimately reduce the need for the kind of massive, ongoing spending on new AI systems that has characterized the current boom.

Bank of America’s perspective, as summarized in the same report, is that chip and hardware suppliers could still get paid under either outcome. The core idea is that even if more AI capability becomes available through open models, deploying, fine-tuning, and running those models at scale still requires high-performance compute, creating demand for the underlying infrastructure.

For NVIDIA, the practical question is not only whether developers choose open or proprietary models, but how those models are trained, optimized, and served. Open-source models can broaden who participates in building and adapting AI applications, which can expand the overall market for GPUs and related networking and software stacks that power training and inference workloads.

The “either way” argument matters because it addresses a risk that investors have debated since open-source momentum picked up: that widely available models might lessen the competitive advantage of closed models and potentially slow down the pace of incremental, platform-driven spending. Bank of America’s framing suggests that, at least from the chipmaker’s standpoint, the underlying compute consumption may remain robust even as the software ecosystem becomes more open.

Huang’s stance also highlights how leadership messaging is evolving in the AI industry. Rather than treating openness as a threat to specialized platforms, the public tone from major AI companies has increasingly emphasized that open models can accelerate adoption, which then increases the number of end users and enterprises that run inference and require scalable hardware.

Still, the details of what Huang intends to support, and how quickly open approaches could translate into measurable spending changes across the industry, were not spelled out in the headline-level report. Specifics such as which model releases are in view, the terms under which they would be opened, or how NVIDIA would structure product or software support for them were not included in the information provided here.

Investors and industry participants will likely watch for follow-on statements from NVIDIA clarifying the scope of its openness position, as well as any evidence that open-source model adoption is materially changing how often new accelerator capacity is purchased. Until clearer disclosures arrive, the debate will remain centered on expectations rather than confirmed shifts in orders or revenue composition.