Descriptions:
Matthew Berman makes a pointed case that US open-source AI is heading toward collapse — not because of a talent or technology gap, but because of a fundamental business model failure. The video explains how any US lab that invests heavily in training an open-source model immediately hands its competitors a free product they can serve with higher margins, having skipped the R&D cost entirely.
The contrast with China is central to the argument. DeepSeek and similar Chinese labs are effectively subsidized by the CCP, which has a strategic interest in giving away competitive AI to erode the revenue models of leading US frontier labs. Berman walks through the current landscape: Meta reversed its open-source commitment within roughly a year; OpenAI treats its open-source releases as a goodwill exercise, not a core business; Anthropic has no open-source strategy at all; and Google’s Gemma series targets on-device use rather than enterprise deployment.
The one potential exception Berman identifies is Nvidia, which has committed $26 billion to open-source AI development. Because Nvidia is upstream of all inference serving — its chips power the very competitors that would otherwise undercut its margins — the economics work in a way they don’t for any other US player. The video closes with a national-security framing: US enterprises currently choosing between expensive closed-source models from OpenAI and Anthropic versus nearly-as-capable, far-cheaper Chinese open-source alternatives are making infrastructure decisions that carry long-term strategic dependencies.
📺 Source: Matthew Berman · Published April 29, 2026
🏷️ Format: Opinion Editorial
