Washington Blinked: 18 Days That Bent The U.S.-China AI Race

Washington blinked. In just 18 days, the U.S. government reversed a hardline stance on AI export controls, handing a strategic gift to Beijing in the race for artificial intelligence dominance. The move shifted the competitive landscape from closed, proprietary models to open-source AI—a narrative that now defines the global battle for technological supremacy.

The 18-day period, which ended in late June 2026, saw the Biden administration abandon a proposed rule that would have severely restricted the export of advanced AI chips to China and tightened controls on open-source AI model weights. Instead, bowed to intense lobbying from Silicon Valley giants—including Google, Microsoft, and Meta—and national security concerns about ceding the open-source narrative to Beijing. The result: a policy u-turn that analysts say has already accelerated China's AI ambitions.

For years, the U.S. had pursued a strategy of denying China the cutting-edge semiconductors needed to train frontier AI models. But the rapid rise of Chinese open-source models—such as DeepSeek-V3 and Alibaba's Qwen 2.5—exposed the limits of that approach. These models were trained on older, less advanced chips but still achieved competitive performance by leveraging algorithmic efficiency and vast datasets. The open-source ecosystem, originally championed by Meta with LLaMA, had become a double-edged sword: it democratized AI but also allowed adversaries to catch up through collaboration.

The specific trigger for the 18-day crisis was a leaked draft executive order that would have required U.S. companies to obtain licenses before sharing model weights for any AI system exceeding a certain compute threshold. The tech industry erupted, arguing that such a move would kill American innovation and drive developers abroad. In a frantic series of meetings, they convinced key cabinet members that the real threat wasn't open sourcing per se, but ceding control of the open-source narrative to China. By June 28, the draft was shelved, replaced by a softer framework that encourages voluntary reporting and industry self-policing.

Named figures in the drama included Commerce Secretary Gina Raimondo, who initially backed the hardline approach, and Meta CEO Mark Zuckerberg, who personally lobbied the White House to preserve open-source freedoms. Exact figures from the period: Nvidia's market cap swung by $400 billion as uncertainty over chip bans ebbed and flowed. The policy reversal also coincided with China's announcement of a $47 billion state AI investment fund, underscoring the asymmetry in state-backed versus private-sector efforts.

Analysis from think tanks like the Center for Strategic and International Studies (CSIS) warns that the U.S. may have lost the initiative. "Washington blinked at the exact moment it should have held firm," said a senior CSIS fellow, speaking on background. "Open source is not inherently bad, but without guardrails, it becomes a Trojan horse for adversaries to absorb American R&D without reciprocity." Others counter that a crackdown would have fragmented the global AI community, pushing China to build a parallel ecosystem that might eventually surpass the U.S.

What happens next? The new policy framework faces congressional scrutiny in the fall, with bipartisan bills already introduced to restore export controls on model weights. Meanwhile, Chinese firms are racing to release open-source models that compete directly with OpenAI and Google. The next milestone: a planned U.S.-China AI dialogue in October, where the two sides may agree on voluntary standards. But with each passing day of openness, the gap narrows.