Growing concern over the United States’ slipping AI leadership is surfacing across Silicon Valley, with Databricks co-founder and Laude venture partner Andy Konwinski warning that China’s momentum in AI research has become an “existential threat” to U.S. innovation and democratic institutions. His remarks, delivered at the Cerebral Valley AI Summit, highlight a deeper anxiety that America’s traditional engine of scientific progress—open research—has stalled just as China accelerates.

A Shifting Research Power Balance

According to Konwinski, PhD students at Stanford and Berkeley increasingly encounter groundbreaking research coming from Chinese institutions and corporate labs, outpacing the flow of advanced ideas from American counterparts. This marks a notable reversal from the past decade, when U.S. university labs, supported by Big Tech partnerships, consistently set the global AI agenda.

A key driver of this shift, he argues, is the rapid movement of top academic talent into high-paying roles at major AI labs. The compensation packages—often reaching several million dollars annually—create an intellectual vacuum at universities, where open peer review and interdepartmental collaboration historically generated the breakthroughs that shaped modern AI.

Konwinski is not questioning the capability of U.S. giants such as OpenAI, Meta, or Anthropic, all of which continue to advance state-of-the-art models. His concern is that their output is largely proprietary, narrowing the pool of researchers who can build on these advances and stifling the kind of broad, interdisciplinary experimentation that produced the original Transformer architecture.

China’s Open-Source Strategy as a Competitive Lever

China, in contrast, has embraced open-source AI development as a matter of strategic policy. Models from groups such as DeepSeek and Alibaba’s Qwen are meant not only to demonstrate national capability but to serve as open infrastructure for developers, startups, and academics across the country.

Konwinski argues that this approach accelerates China’s path to the next major architectural breakthrough. By encouraging full transparency—weights, training methods, documentation—China fuels a constant cycle of iteration that compounds quickly within its massive developer community.

This dynamic, he warns, is reshaping the global innovation map. With the U.S. increasingly reliant on closed corporate research ecosystems, the traditional scientific diffusion—“scientists talking to scientists,” as he puts it—has begun to erode.

Why Open Source Matters for U.S. Competitiveness

Konwinski believes the U.S. must reopen its innovation pipeline. That means incentivizing open research, strengthening academic funding, and re-establishing norms of collaboration that reward foundational breakthroughs rather than exclusive model ownership.

Without intervention, he warns, the U.S. risks hollowing out its long-term leadership. “We’re eating our corn seeds,” he said. “Fast forward five years, the big labs are gonna lose too.”

Looking ahead, the U.S. AI ecosystem faces a strategic choice: continue down a path dominated by closed corporate competition, or reinvest in open research ecosystems that historically powered American scientific leadership. The outcome may define not just the business landscape of AI, but the geopolitical balance of technological power in the decade ahead.