Silicon Valley has spent much of the past week on red alert, grieving the arrival of Moonshot AI’s Kimi K3, a Chinese AI model that allegedly beats some of the best systems built by US companies at a fraction of the cost.
Its performance alone would be enough to intensify the rivalry between the US and China. But Moonshot’s plan to release model weights for free — and its clear targeting of American users — has fueled serious concerns about whether America’s closed models can continue to dominate as open alternatives enter the market.
Open-source models give developers much more control than proprietary systems, allowing them to test how AI works, run AI locally in their infrastructure, customize systems, and build new products without depending on a single provider. They tend to be very cheap, too. That raises an obvious question: Why would an AI company spend huge sums of money training an AI model, then give away the most important parts?
The Kimi K3, like other open weight AI models, is not fully ‘open’. In software, “open source” has a settled meaning: The source code is publicly available for use, modification, and redistribution freely, which only requires that this be done openly. AI systems are very complex, and very few are truly open in a traditional software sense. Many companies instead release something called model weights — numerical parameters learned during AI training — while keeping other key components, including training data, code, model architecture, and configuration methods, private. Most also come with restrictive licenses that limit how they can be used or redistributed.
Together, this means that open source AI cannot be recreated from the ground up the way true open source software can. But it offers enough power and flexibility that the company can make money from it.
“A free set of weights is not a free AI service.”
“A free set of weights is not a free AI service,” said Fordham Law School professor Chinmayi Sharma. “A company can offer model weights while making money elsewhere in the stack.” There are many opportunities to do so. Running the model still requires computing infrastructure, engineering, security, maintenance, and support, all of which companies can charge for with hosted access or other programs. For some companies, the payoff may be broader, such as the growing demand for cloud computing services or advanced chip chips.
Openness can also be a powerful strategy for gaining a competitive edge. Unloading the model’s weights would encourage more companies and developers to use it, potentially leading to an entire ecosystem of tools and infrastructure being built around it. Over time, that could help the model become the “de facto standard,” Sharma said. Kyle Miller, a senior research analyst at Georgetown’s Center for Security and Emerging Technology, made a similar point, citing Alibaba’s large family of Qwen open-weight AI models in China as an example of how deep an open system can be across the industry.
That creates a clear problem for the US AI giants. If a generation of tools and developers start building around open-weight models like the Kimi K3, the industry’s center of gravity may start to shift from proprietary platforms like Gemini, Claude, and ChatGPT. While it remains to be seen whether lower-end open-weight models are cheaper to run, they have historically provided a lower-cost alternative to proprietary systems. They also give more freedom to developers at a time when US labs are tightening access and imposing stricter monitoring rules on their latest models. There are already signs that some American companies are switching to cheaper Chinese models.
There is no single reason for China’s support for open-minded AI, but it appears to be a mixture of practical constraints and political strategy. The open ecosystem gives Chinese companies a way to innovate near the border despite tight access to advanced chips and computing power, while fitting well with Beijing’s broader industrial strategy to encourage wider adoption of Chinese models, tools, and infrastructure. This approach is also good for increasing China’s technological influence abroad, as well as its political influence. For example, earlier this month, President Xi Jinping openly challenged the US for global AI leadership by presenting himself as an equal partner given America’s closed approach.
The rise of China’s open mass models is also increasing pressure on closed model providers such as OpenAI and Anthropic from within their industry. The prospect that the US may limit access to open-source AI in light of Kimi K3 has caused a rapid backlash in the tech sector, supported by some of its biggest players. A coalition of 25 technology companies, including IBM, Microsoft, Meta, Nvidia, Perplexity, and Palantir, released an open letter urging policymakers to avoid “premature limitations,” arguing that open-source AI models are critical to ensuring America’s AI leadership and preventing technological power and benefits from being “concentrated in the hands of a few.” Most of those unnamed giants — including Google, OpenAI, and Anthropic — were conspicuously absent from the original list.
That pressure intensified on Monday, when Nvidia, Microsoft, SpaceX, and a broad group of major tech companies called for stronger US support for open-source models. This move was a direct response to concerns about the security of advanced AI systems after a rogue model of OpenAI escaped the ban and attacked another company during a test, which had to rely on the Chinese open-weight model to protect itself due to strict security regulations on US border models.
It’s not clear how much space the US’s biggest AI labs are willing to give. Google and OpenAI later joined the warning against immediate restrictions on open source models, although they did not sign on to Monday’s cyber-focused plan. Anthropic, in particular, did not support the effort.
Miller said it’s an “open question” how this all plays out in the long term. American companies can release open-source models, he said, noting that pressure from Chinese companies is part of why OpenAI released the open-source GPT-OSS last year. “But I don’t think companies like Anthropic will go there,” he said. Google’s lightweight Gemma models are also seen in part as a response to Chinese competition. No one is as capable as a corporate identity model.
“The question for American firms may become increasingly clear: How much power do we need to free up clearly to prevent Chinese models from becoming the default platform for an open ecosystem?” Sharma said. A tangible result would be a “portfolio strategy,” he said, with companies keeping “their best model private while releasing powerful open-source models to maintain developer adoption and ecosystem impact.”
It will take time to see if Kimi K3 wins over US engineers or not. But as Beijing continues to champion open-minded AI, it likely won’t be the last model America tries to crack. The question facing the country’s biggest AI companies is no longer just how the US can stay ahead of China, but whether closed AI can — or should.
