Meta and Nvidia this week released artificial intelligence models that developers can download for free through the open-source ecosystem, as U.S. tech companies push open-weight AI models in a race that has drawn in China, Washington and Silicon Valley. The move looks generous on paper. The power behind it is not.
Who Gets to Set the Terms
A group of American tech giants urged policymakers not to impose what they called "premature restrictions" on open-weight AI models, even if they come from China. That open letter, dated July 24, came from the same corporate class that wants the rules written around its own products, its own market share, and its own version of "openness." The consortium wrote, "The age of AI can be one of prosperity," and added, "With the right choices, open weight AI can expand opportunity, strengthen competition, extend American technological leadership, mitigate risk, and ensure that the benefits of this extraordinary technology are shared broadly across our economy."
That language is doing a lot of work. It frames corporate expansion as public benefit, and it asks policymakers to keep the gates open while the companies inside the gates decide who gets access, who gets locked out, and who gets to call the shots.
Meta on Monday unveiled Muse Glimmer as part of a strategy to release its most powerful AI models to the open-source community. Chief Executive Mark Zuckerberg said the company would open the weights for its latest AI model, Muse Spark 1.2. Weights are the calculations and rules that determine how the AI works and behaves. A day later, Nvidia debuted Nemotron 3.5 Lightning. The model comes from the company's Nemotron 3 family of models released in December. Nvidia said its models are "truly open source" because the company publishes the related "training datasets, techniques, and model weights" for developers to inspect.
Who Pays for the Race
The companies still have to prove there’s an audience for their offerings in a market that already includes popular models from Chinese AI labs such as Moonshot AI and DeepSeek, as well as Alibaba's Qwen. The competition is framed as a contest between national blocs, but the people doing the actual work are developers, enterprises, and users being asked to adapt to whatever the giants release next.
Box Chief Executive Aaron Levie, one of the signatories of the July 24 letter, called Zuckerberg's plan for Muse Spark 1.2 a "very big deal" because it is a powerful model that rivals top foundation models from Anthropic and OpenAI. He said the models from Meta and Nvidia this week are smaller and intended to run on laptops for tasks such as powering on-device digital agents.
"There's a very firm flag in the ground that America will have near-frontier open-source models," Levie said.
That flag is planted by corporations, not communities. The race is about who controls the infrastructure, who sets the standards, and who gets to define what counts as open when the underlying machinery still belongs to the same firms.
Meta has tried this route before with Llama, Zuckerberg's initial entry into the foundation AI market. The release of Llama 4 in April 2025 left developers unimpressed, and Meta later spent billions of dollars to overhaul its AI unit, installing Scale AI Chief Executive Alexandr Wang as the division's leader. Wang's group has recently been rolling out proprietary models under the Muse branding to try to develop new revenue streams.
What Developers Are Told to Believe
Umesh Sachdev, chief executive of business AI startup Uniphore, said Meta burned bridges with third-party developers when it shifted from open weight to proprietary AI models. "I think it's going to take more than a 3,500 worded article from Zuck to convince developers," Sachdev said of Zuckerberg's accompanying manifesto this week. "The emotion of my developers at Uniphore, they almost feel betrayed."
That word matters. Betrayed. The people building on these systems are being asked to trust a platform that has already moved from one model of access to another, then back again, all while chasing revenue and market control.
Sachdev said he is rooting for domestic companies to succeed because "more competition will drive down token cost, and will drive up innovation, and it's always good for consumers."
Forrester analyst Charlie Dai called Meta's latest move "strategically important because it restores a major U.S. frontier AI vendor to the open ecosystem." He said, "Developers and enterprises will likely welcome Meta's shift back toward open weights because it improves transparency, customization, deployment flexibility, and data sovereignty." Dai added that the company must "prove it can cultivate a durable ecosystem beyond releasing competitive models."
That last line says the quiet part out loud. Release the model, capture the ecosystem, keep the leverage.
CNBC reported that Google devices chief Rick Osterloh said Pixel and iPhone are going in "very different directions" as Google leans into Gemini features it says are unique to Android. Osterloh said Gemini could eventually become the primary way people use their phones, laptops and other devices. He also said the memory shortage is forcing consumer electronics prices higher and that Google expects it will need to raise prices further.
The hardware giants keep moving the goalposts. The software giants keep rewriting the rules. And the people at the bottom get the bill, the lock-in, and the lecture about innovation.