25 giants hold high the banner of open source. Why are Open AI and Anthropic silent?
On July 24, local time, 25 American technology companies and institutions jointly signed an open letter explaining the importance of the open source model. The signature list includes not only established technology giants such as Nvidia, Microsoft, IBM, Meta, and Dell, capital institutions such as a16z and YC, AI upstarts such as HuggingFace, Mistral, and Perplexity, but also backbone forces such as the Linux Foundation and Mozilla that adhere to the foundation of open source.
On July 24, local time, 25 American technology companies and institutions jointly signed an open letter explaining the importance of the open source model. The signature list includes not only established technology giants such as Nvidia, Microsoft, IBM, Meta, and Dell, capital institutions such as a16z and YC, AI upstarts such as Hugging Face, Mistral, and Perplexity, but also backbone forces such as the Linux Foundation and Mozilla that adhere to the foundation of open source. NVIDIA CEO Jen-Hsun Huang posted his first personal tweet endorsing this open letter on a social platform, and Microsoft CEO Nadella, Elon Musk and others have forwarded responses one after another. The content of the open letter emphasizes that the open source model can strengthen security and network protection capabilities, accelerate technological innovation and industry popularization, and achieve independent and controllable technology. The status of AI in the United States should not be measured by the success or failure of a certain cutting-edge model, but by whether it can build a strong and open ecosystem and allow it to penetrate into various industries. Open letter exposes differences of interest In addition to Huang Renxun, Microsoft Chairman and CEO Nadella also issued a statement saying: The open weight model is crucial to a healthy AI ecosystem. Zuckerberg, chairman and CEO of Meta, said: Open source is a positive and important force that can both empower people and prevent the concentration of power. It's an honor to support open source. The most intriguing aspect of the open letter is the stark contrast between the signatories and the absentees. Huang Renxun proposed that "the world needs both cutting-edge closed source and cutting-edge open source." As the world's largest seller of GPU computing power, the more prosperous the open weight model is and the larger the developer base is, the more exponential the demand for computing power will be. Also in March this year, Nvidia announced plans to invest a total of US$26 billion in the next five years to develop open-weight AI large models, including model development, computing infrastructure, research and talent, and ecosystem construction. The ecological part refers to building a developer and application ecosystem around Nvidia hardware, and trying to make its own hardware architecture a de facto industry standard through an open source model. Defending open source means defending its hardware ecological barriers. As the largest shareholder of OpenAI, Microsoft still signed, mainly out of risk hedging considerations: holding OpenAI with its left hand to anchor high-end enterprise-level APIs, and supporting the open ecosystem with the help of Azure computing power with its right hand to prevent being kidnapped by a single closed-source supplier. Meta has been betting on the Llama series for a long time, weakening the subscription premium of closed source opponents by making the underlying model free. For a16z and Y Combinator, the vast majority of startups cannot afford the cost of pre-training models with hundreds of billions of parameters. Once the opening weight is subject to administrative restrictions, the capital efficiency of entrepreneurship in the entire AI application layer will plummet, and the exit channel for venture capital will also be significantly narrowed. Compared with the huge momentum of 25 manufacturers, the absence of OpenAI and Anthropic exposed the deep worries of the closed source camp. Both companies have secretly submitted S-1 documents to the U.S. Securities and Exchange Commission at this time, preparing for an IPO with a valuation approaching US$1 trillion. Its valuation logic is based on cutting-edge closed-source models and high unit price API authorization. If the open weight model can approach or even surpass the closed source level in the short term, its fee moat will instantly face revaluation pressure. Although OpenAI CEO Sam Altman and President Greg Brockman have repeatedly stated in public that they “support the democratization of AI,” their refusal to sign revealed their defense of commercial barriers. The AI competition behind the open-source and closed-source debates The open letter incident broke out at this time, and the direct trigger was the recent accelerated rise of China’s AI and the US government’s stress response. In July 2026, the Chinese startup Dark Side of the Moon Kimi released the Kimi K3 open weight model, which achieved performance comparable to the top closed-source models in the United States in multiple benchmark tests, while the inference cost was only a fraction of the latter. The data shows that the average performance gap between open-weight models and top closed-source models has narrowed to 3.3%. Faced with this change, the White House Office of Science and Technology Policy and the U.S. Department of the Treasury tried to restrict the release of open source models on the grounds of preventing technology spillover and national security, and accused some Chinese teams of absorbing the ability of U.S. closed source models through knowledge distillation.
In response to this, Huang Zhenxin, head of Kimi Enterprise Business, retorted in an interview: Open source models and large Chinese models should not be labeled as low-priced. The core of K3’s performance jump comes from the innovation of the underlying original architecture, not a distillation of existing models. This time, 25 major joint manufacturers also emphasized in an open letter: A blanket ban on distillation and open source models is tantamount to giving up food because of choking. It will not only fail to stop opponents, but will also interrupt the innovation rhythm of the American local ecosystem. Zhang Faen, chief scientist of Innovation Qizhi and senior engineer of Beijing Artificial Intelligence, told reporters that this open letter exposes the illusion that "closed source is safe." If the most advanced AI capabilities are concentrated in the hands of a few closed-source models, the systemic risks to human society will be infinitely amplified. Not only can closed-source models be silently broken or abused, but they can also create single points of failure. Once a fatal flaw occurs in this black box, the outside world will not even have the ability to diagnose it. True security is never "safe only when invisible". In a world where hackers have begun to use advanced AI to launch network attacks, if the defender does not have an equally capable open source model to simulate, detect and defend, it is tantamount to facing a powerful ship with bare hands. At the same time, Zhang Faen analyzed that the giants used this declaration to prudently and clearly rectify the name of model distillation technology: this technology relies on the knowledge of large models to iteratively optimize small models. Closed source parties regard it as an infringement of rights, while the open source system regards it as a reasonable inheritance of "standing on the shoulders of giants" and is a normal technology iteration. If legislation restricts this research and development method across the board and confuses technology research and development with malicious misappropriation, it will eventually stifle the innovation vitality of the industry. The battle for open source weight is essentially a game over the control of AI technology. Zhang Faen said that the era of a few oligopolists monopolizing the AI industry is outdated. Local open source forces such as DeepSeek, GLM, and Kimi are reshaping the industrial order. A country's AI leadership relies on an open ecosystem that deeply penetrates all walks of life, rather than a top closed-source model. Open source code built the foundation of the Internet more than 40 years ago. Now Nvidia, Microsoft, Meta and other companies have once again joined hands to defend the freedom, security and future of the AI era with open weight.
In response to this, Huang Zhenxin, head of Kimi Enterprise Business, retorted in an interview: Open source models and large Chinese models should not be labeled as low-priced. The core of K3’s performance jump comes from the innovation of the underlying original architecture, not a distillation of existing models. This time, 25 major joint manufacturers also emphasized in an open letter: A blanket ban on distillation and open source models is tantamount to giving up food because of choking. It will not only fail to stop opponents, but will also interrupt the innovation rhythm of the American local ecosystem. Zhang Faen, chief scientist of Innovation Qizhi and senior engineer of Beijing Artificial Intelligence, told reporters that this open letter exposes the illusion that "closed source is safe." If the most advanced AI capabilities are concentrated in the hands of a few closed-source models, the systemic risks to human society will be infinitely amplified. Not only can closed-source models be silently broken or abused, but they can also create single points of failure. Once a fatal flaw occurs in this black box, the outside world will not even have the ability to diagnose it. True security is never "safe only when invisible". In a world where hackers have begun to use advanced AI to launch network attacks, if the defender does not have an equally capable open source model to simulate, detect and defend, it is tantamount to facing a powerful ship with bare hands. At the same time, Zhang Faen analyzed that the giants used this declaration to prudently and clearly rectify the name of model distillation technology: this technology relies on the knowledge of large models to iteratively optimize small models. Closed source parties regard it as an infringement of rights, while the open source system regards it as a reasonable inheritance of "standing on the shoulders of giants" and is a normal technology iteration. If legislation restricts this research and development method across the board and confuses technology research and development with malicious misappropriation, it will eventually stifle the innovation vitality of the industry. The battle for open source weight is essentially a game over the control of AI technology. Zhang Faen said that the era of a few oligopolists monopolizing the AI industry is outdated. Local open source forces such as DeepSeek, GLM, and Kimi are reshaping the industrial order. A country's AI leadership relies on an open ecosystem that deeply penetrates all walks of life, rather than a top closed-source model. Open source code built the foundation of the Internet more than 40 years ago. Now Nvidia, Microsoft, Meta and other companies have once again joined hands to defend the freedom, security and future of the AI era with open weight.