‘AI’s Sputnik Moment’: Chinese AI Model DeepSeek-R1 Reportedly Surpasses Leading U.S. AI Models

Chinese AI firm DeepSeek released its latest model, R1, on January 20. According to the company, R1 performs favorably compared to leading Western models on key benchmarking tests despite having been allegedly developed at a fraction of the cost and employing a fraction of the computing power. Silicon Valley venture capitalist and White House advisor Marc Andreessen referred to DeepSeek’s announcement as “AI’s Sputnik moment,” comparing the development to the Soviet Union’s launch of the first satellite.

The timing of DeepSeek’s R1 release, coinciding with President Donald Trump’s first week in office, appears to be a calculated move aimed at rattling confidence in America’s AI leadership and raising doubts about the effectiveness of U.S. export controls intended to stymie China’s AI ambitions.

DeepSeek Model Costs Less, Uses Fewer Chips

In contrast to initial versions of ChatGPT, DeepSeek’s performance is based on its use of a reasoning model that cycles through alternative answers before selecting the best option, an approach designed to produce high-quality responses to coding and mathematics queries.

In a technical report, DeepSeek claimed that it used a cluster of roughly 2,000 Nvidia computer chips to train V3, a cousin of R1, far fewer than the tens of thousands used by Western companies to train similar models. DeepSeek also reported that its training costs amounted to only $5.6 million, though independent researchers estimate that this figure may underreport actual spending given the cost of generalized training prior to training a model on a particular task. By contrast, similar Western models can reportedly cost nearly $1 billion to train.

Unlike Western AI models, DeepSeek is also subject to China’s censorship regulations, with the model refusing to answer politically sensitive queries, including requests regarding the Tiananmen Square massacre or Chinese President Xi Jinping’s leadership weaknesses. Like TikTok, DeepSeek collects large amounts of data from its users, including their keystrokes and operating system information. Mirroring other Chinese apps, DeepSeek models are also subject to Beijing’s data harvesting and national security laws.

U.S. Export Controls Seek to Block Chinese Access to Advanced Technology

DeepSeek’s reported achievement comes as Washington has expanded its use of export controls to stymie China’s AI development. Beijing has poured significant resources into developing domestic AI firms capable of competing with their Western counterparts. By wielding its AI models to advance its innovation base, Beijing also hopes to contribute to the country’s military modernization.

In response, the Biden administration introduced a series of increasingly stringent restrictions on the export of leading AI technologies, including computer chips and manufacturing equipment. These controls culminated in the release of the “Framework for AI Diffusion” on January 13, which proposed a cap on the export of advanced computer chips to prevent China from acquiring U.S. components via transactions with third countries or via cloud computing.

Washington Should Step Up Export Controls and Lay Groundwork for AI Success at Home

As a starting point, the Trump administration should finalize the proposed “Framework for AI Diffusion” to further tighten China’s chip access. Congress should advance proposed measures to regulate outbound investment to prevent U.S. firms from contributing to the development of China’s high-tech sector.

At home, the administration should prioritize working with Congress and the private sector to produce more favorable conditions for AI development, fulfilling Trump’s promise that the United States will be “laser-focused on competing to win” on AI. This includes rapidly increasing U.S. energy production and boosting the resilience of the electrical grid to support the development of energy-intensive AI projects.

Jack Burnham is a research analyst at the Foundation for Defense of Democracies (FDD), where Craig Singleton is a senior fellow and senior director of FDD’s China Program. For more analysis from Jack, Craig, and the China Program, please subscribe HERE. Follow Jack on X @JackBurnham802. Follow Craig on X @CraigMSingleton. Follow FDD on X @FDD. FDD is a Washington, DC-based, nonpartisan research institute focused on national security and foreign policy.