Nvidia CEO Huang led 70+ tech giants (Microsoft, Meta, a16z) in signing an open letter on 'Open Weight and American AI Leadership,' directly confronting Anthropic's closed-source positioning.

The rarely outspoken "leather-jacket ninja" Huang Ren-hao broke his silence. He recently posted his first-ever tweet on X (formerly Twitter), publicly backing an open letter titled "Open Weights and American AI Leadership." Within days, the list of signatories swelled from an initial 25 to over 70 companies, with viewership exceeding 60 million.

This marks the most divisive symbolic moment in Silicon Valley's AI sector over the past year. Behind this roster—featuring Microsoft CEO Nadella, Elon Musk, Yann LeCun, Andrew Ng, and other industry leaders—Silicon Valley has fractured cleanly into two camps: one led by NVIDIA and Meta spearheading an open-source alliance, the other anchored by Anthropic defending closed-source principles. As AI enters its commercial deepwater phase, this ideological clash has transformed into an "arms race" over industry pricing power and control of the compute infrastructure itself.

Looking back over the past six months of AI development, closed-source giants once held absolute commercial dominance through their parameter advantages. Now, the open-source coalition's comprehensive counteroffensive has arrived.

At the storm's center sits Huang's open letter, whose demands are unmistakably clear: open weights models are essential to a healthy AI ecosystem; the industry must unite in support; regulators must not impose blanket restrictions on open-source models.

What appears on the surface as an industry advocacy piece is, in substance, a precisely calibrated anti-monopoly manifesto. The letter articulates four core demands: first, break technical monopoly and prevent startups from being locked into dependence on single suppliers; second, extend competition from the model layer into cloud computing and chips, activating broader competition throughout the value chain; third, reframe the safety narrative by arguing that "transparency is safer than opacity," dismantling black-box systems controlled by a handful of giants; and most crucially, rehabilitate model distillation—asserting that using large-model outputs to train smaller models is a legitimate research practice that must not be criminalized as "theft."

In essence, Huang and his allies are drawing a line for regulators: do not strangle the open-weights path in the name of preventing hypothetical risks.

This sudden "Silicon Valley realignment" is no impulsive contest of egos. So the question becomes: why now?

Parsing the 70-plus signatories reveals a battle over value distribution rules throughout the industry, with each faction fiercely protecting its "moat" and "core interests":

**Hardware and compute giants (NVIDIA, Dell):** As closed-source incumbents accelerate custom chip development to escape compute dependency, NVIDIA must expand the overall compute market through open-source ecosystems. More open-weights models drive global enterprises toward self-hosting and fine-tuning, turbocharging both stock and incremental compute demand—NVIDIA's bedrock concern.

**Cloud and enterprise SaaS (Microsoft, Palantir, others):** These players need both to hedge risk and diversify cloud supply. Open weights free enterprise customers from single-API lock-in, enabling deployment on private and on-premises infrastructure, directly translating into cloud and SaaS subscription revenue.

**Open-source ecosystem and infrastructure (Meta, Mistral, Hugging Face):** Meta seized AI-era standard-setting through open Llama. Emerging players like Mistral penetrate enterprise private clouds via open weights, capturing high-value customers from closed-source incumbents.

**Top-tier VCs and application-layer ecosystem (a16z, YC, others):** Premier VCs have seeded hundreds of application startups. If foundational APIs remain perpetually controlled by a few closed-source giants, application-layer margins will be hollowed out. Only open weights gives the application layer genuine cost control and business flexibility.

In this "great coalition," Anthropic stands out as the most conspicuous dissenter. As the closed-source hardliner, Anthropic not only refused to sign—its lead researcher Julian openly mocked the "open-source faction" on X, directly confronting the open-source camp. This provoked a sharp response from AI luminary Andrew Ng: "Anyone has the right to keep their code private, but no one has the right to prevent others from open-sourcing."

Anthropic's core anxiety stems from a business model built entirely on model scarcity. Once models become open-source commodities, the scarcity premiums fueling the "thousand-model wars" evaporate, and with them, the high API prices and inflated valuations (PE expectations) Anthropic depends on. Regulatory uncertainty remains the biggest obstacle facing the open-source camp.

Stepping back from any single company's perspective and viewing the endgame from today, this already-revealed "battle of the gods" signals that AI has definitively moved past its wild-growth phase.

As open-source competition intensifies, foundational large models are accelerating toward commoditization—becoming cheap infrastructure. This means startups betting on parameter-scale competition face a dead end. Future profit pools will sharply polarize: downward, toward the compute and infrastructure stack NVIDIA controls; upward, toward the application layer holding proprietary data, capable of embedding AI deeply into real business workflows and agent systems.

The more closed-source giants pursue monopoly, the deeper enterprise clients' concerns about data sovereignty and technical security. Vertical industry models offering small parameters, high performance, and private deployment capability—combined with mass adoption of edge compute (AI PCs, AI phones)—will form the most certain monetization path ahead.

Commerce contains no absolute justice, only eternal self-interest. Anthropic's closed shop guards its premium valuations; Huang and allies rally to protect the compute empire's unlimited growth. But for the vast AI developer community, open-source has demolished technological barriers. Today's real premium lies not in R&D itself, but in closing the information gap between technical solutions and actual market needs. Don't become a wage-earner enslaved to closed APIs. Embrace open-source, identify your landing scenario, and secure your ticket to the next era.