How Silicon Valley Killed Trump's AI Executive Order — And What Comes Next

When the Trump administration issued its AI Executive Order in early 2025, it carried the familiar hallmarks of Washington's relationship with technology policy: ambitious framing, ambiguous implementation, and an immediate collision with the industry it was meant to regulate. What happened next was a masterclass in how Silicon Valley actually shapes American policy — not through lobbying alone, but through the structural reality that the government needs the industry far more than the industry needs the government.

What the Executive Order Actually Said

The EO had three central pillars. First, it directed federal agencies to assess AI risks and develop procurement standards for government use of AI systems. Second, it included provisions echoing previous Biden-era orders around export controls on advanced AI chips — specifically targeting the continued restriction of high-end GPU exports to China and other adversary nations. Third, it called for a national AI safety framework developed in partnership with NIST.

On paper, this was a relatively moderate, continuity-oriented approach. In practice, the implementation became a battleground.

Where Silicon Valley Pushed Back

The flashpoint was the export control provisions. Nvidia — whose H100, H200, and Blackwell chips are the literal infrastructure of global AI — mounted an aggressive lobbying campaign arguing that restricting chip exports was damaging American competitiveness without actually preventing China from developing alternative semiconductor capacity. The argument had merit: TSMC and Samsung's foundry capabilities, combined with Huawei's internal chip development, meant that export controls were slowing but not stopping Chinese AI development — while costing American companies tens of billions in lost revenue.

By Q3 2025, the export control provisions had been significantly softened. A tiered licensing system replaced the blanket restrictions, and several allied nations were given expanded access to advanced AI chips under security agreements. Nvidia's stock hit new all-time highs within weeks of the policy revision.

The AI Safety Framework Debate

The national AI safety framework proved equally contested. Anthropic and OpenAI — both with strong regulatory relationships in Washington — supported a structured safety evaluation process. Meta and Google pushed back on mandatory pre-deployment testing requirements, arguing they would entrench incumbents and disadvantage open-source development. The final framework, released by NIST in late 2025, was voluntary rather than mandatory — a significant victory for the tech industry's 'self-regulation first' position.

What Comes Next for American AI Policy

The 2026 AI policy landscape is characterised by a genuine tension between two imperatives that both sides acknowledge. America needs AI leadership to maintain geopolitical and economic dominance — that requires letting the industry move fast. America also needs AI safety guardrails to prevent catastrophic misuse — that requires slowing the industry down. Neither side has a monopoly on being right. The debate will define the next decade of American technological history.

What is clear is that Silicon Valley retains structural power over AI policy that no executive order can easily override. The government's dependence on private sector AI capability — for defence, intelligence, healthcare, infrastructure — means that the industry will always have leverage. The question for Washington is how to convert that dependency into durable accountability without killing the innovation engine it depends on.