Белый дом спорит, как реагировать на китайские AI-модели вроде DeepSeek

In July 2026, the dispute within Donald Trump's administration over how to respond to increasingly powerful AI models from China intensified. As Wired reports, the White House still has no unified position on Chinese artificial intelligence.

What the White House dispute is about

The dispute within the Trump administration comes down to a choice of strategy: contain China with restrictions on chip and technology exports, or race to widen the lead of American AI companies as fast as possible. The confrontation became publicly known in July 2026 from a Wired report. The two approaches don't fit together well in the details — strict restrictions on selling Nvidia accelerators also hit the revenue of the American manufacturers themselves, which partly funds their own research and development.

As Wired puts it, the issue is specifically how to respond to increasingly powerful Chinese models, not whether they pose a challenge — the fact of the challenge itself is no longer disputed in Washington.

«The

Trump administration is arguing over how to respond to increasingly powerful Chinese AI models», Wired writes.

Why Chinese models worry the US

Chinese labs are quickly closing the gap with American ones, and some models are being released as open source. In January 2025, the release of the reasoning model DeepSeek R1 triggered a sell-off in tech stocks: in a single day, January 27, 2025, Nvidia lost about $600 billion in market value — one of the largest one-day market-cap drops in the history of the US market. That's precisely why the issue escalated to the level of the White House rather than remaining a topic for industry experts.

Following that, open models from Alibaba's Qwen family and DeepSeek updates showed that competitive AI can be trained more cheaply and distributed for free. For Washington, this means that export barriers alone cannot preserve technological leadership.

Where Nvidia chips fit in

Nvidia accelerators are the central element of the dispute, because the most powerful models are trained on them. By restricting their export, the US slows down Chinese labs, but at the same time closes off a major sales market for its own manufacturers. During 2025, US authorities changed the rules for selling AI chips to China several times — sometimes tightening them, sometimes allowing shipments of scaled-down versions — and this inconsistency reflects that very dispute inside the administration. The more powerful Chinese models become, the sharper the question of whether export controls are helping or merely pushing China toward producing its own chips.

What levers are being discussed

In the public discussion around Chinese AI, several levers are usually named, and the administration is balancing between them. The first is tightening export controls on computing hardware to slow the training of powerful models in China. The second is betting on the US's own open ecosystem, so that American models — not Chinese ones — become the standard for developers worldwide. The third is trade and diplomatic agreements that tie chip access to broader terms.

Key points of the dispute:

• Topic — how the Trump administration should respond to the rise of Chinese AI (Wired, July 2026)
• Alarm trigger — DeepSeek R1 (January 2025) and open Qwen models from Alibaba
• Market crash — Nvidia down about $600 billion on January 27, 2025
• Main lever — export controls on advanced Nvidia accelerators for China
• Dilemma — restrictions hit the revenue of American chipmakers who fund their own R&D

What it means

While there is no consensus in the White House, the rules of the game for global AI remain fluid: the outcome of the dispute that surfaced publicly in July 2026 will determine both export restrictions for Nvidia and the terms of competition between American and Chinese labs. For the market, this means continued uncertainty — across the entire chain, from chipmakers to model developers on both sides of the Pacific.