Only after reading these two news articles did I understand: Why is Huang so aggressively promoting open source?

Anthropic moves extremely fast, building its own computing power while at the same time tying Samsung and SK Hynix to its own bandwagon.

The underlying industry logic revealed here is actually very direct:

1. Closed-source giants want to gain control over the "lowest level"

Self-developed ASIC chips are nothing new, but before it was all about "designing and finding OEM production", while top Labs (such as OpenAI and Anthropic) have realized that chip design alone is not enough; HBM memory and wafer production capacity (Logic/Foundry) are the true bottlenecks.

Anthropic is now directly engaging with Samsung,SK Hynix, and Micron in cross-shareholding/deep integration. Simply put, this is about establishing its own position in the hardware supply chain at the lowest level, no longer willing to pay Nvidia's high "GPU premium".

2. The real intention behind Jensen Huang's strong advocacy of open source

Recently, Jensen Huang has acted like an open-source spokesperson, enthusiastically supporting the open-source community and developers everywhere. The fundamental reason is clear—the large closed-source companies with abundant capital will eventually pursue self-developed chips to reduce costs.

Nvidia must strengthen and expand the open-source ecosystem so that thousands of small and medium-sized enterprises, developers, and cloud providers remain tightly bound to CUDA and N cards, keeping its general-purpose GPU market stable.

3. The overt strategy from "buying graphics cards" to "locking in production capacity"

According to US Stock Investment Network, although in the short term, the absolute main force for training large models is still the N cards, top Labs are no longer content to just be "major card buyers." As the arms race for computing power has escalated directly to wafer foundry and HBM storage origins, the bargaining power for computing resources in the future is shifting from "simply selling GPU graphics cards" further towards "full-stack independent R&D + underlying production capacity lock-in."