Jason Rathje told CNBC that U.S. investment in open source AI is key because China leads that part of the field. Rathje is president of public sector at webAI and previously ran the Pentagon’s Office of Strategic Capital. His argument covers national security, supply chains and how the country measures progress in AI. That matters for investors because it questions the centralized data center model behind much of the current AI expansion.
Disclosure: webAI builds distributed, on-device AI. Rathje’s case for moving away from centralized data center AI toward individual users benefits his employer’s product category.
Why Rathje Says China Holds the Open Source Lead
In AI, “open source” describes models with published weights. Weights are the numbers a model learns during training. Anyone can download these models, run them on their own hardware and modify them. Closed frontier models are the most advanced systems, and people can use them only through the provider’s own service.
Rathje said “China is really pushing this kind of a more open source approach, where they’re pushing AI into the hands of people and organizations.” He added: “China is leading in that space. It’s cheaper. It’s almost as effective as frontier models.”
The claim that China leads open source AI is Rathje’s. 24/7 Wall St. has not independently verified it.
A National Security Case for Spreading AI Out
Rathje describes centralized AI as a security risk: “The closed frontier data center approach is one that provides acute vulnerabilities to national security. If our AI, our intelligence, is located in one geographical place, now you provide a target.”
His proposed fix: “We think the U.S. needs a strategy that truly democratizes intelligence, that pushes it down to the individual, whether that’s the individual warfighter or the everyday American.”
A Colorado General Assembly report defines edge data centers as smaller facilities operated closer to users, with benefits including reduced network latency and data communication delay, faster content distribution, optimized workloads and improved application performance. Rathje frames this architecture as a national security case rather than a speed case. The same report notes that geographically distributed data centers are the primary means of housing the infrastructure that develops, trains, deploys and operates AI systems, according to Colorado General Assembly.
Supply Chain Gaps Rathje Wants Closed
Rathje’s third point concerns manufacturing resilience: “If we don’t drive forward the supply chain in a way that provides American resilience, we’re going to find ourselves in AI like we are in electric vehicles. We push the scientific frontier, we created an industry, and now there’s a broad” [quote ends]. The electric vehicle comparison is his.
The European Court of Auditors’ special report on the EU’s microchip strategy found that raw materials are controlled by China, Japan, South Korea and the United States, and that microchip production is centred in East Asia, with Taiwan and South Korea manufacturing cutting-edge microchips.
According to the European Court of Auditors, China’s stated chip goal was to reach 70% self-sufficiency by 2025 through its Big Fund vehicles and local government funds. The report does not show whether China met it.
What Investors Should Weigh Before Accepting the Pitch
Rathje’s central claims are difficult to test directly. The supply chain concentration he describes is documented. Whether distributed edge AI is the right response is a judgment call that also benefits his employer’s business model.
A Scoreboard Question Worth Tracking
Underneath his remarks, Rathje is asking whether the U.S. is measuring the AI competition by the right metric. He says that the frontier model race is the wrong scoreboard. In his view, cheaper and widely distributed models offer “a wholly different way to assess the effectiveness of AI.” That question stands on its own, whoever is asking it.