On July 23, U.S. President Donald Trump finally released is highly anticipated strategy for artificial intelligence (AI) – “Winning the Race: America’s AI Action Plan.”While a lot of the language in the document was bold, direct, and undiplomatic – even to the extent of being somewhat blunt and pompous – its content and direction were actually rather concrete, and not as unconventional as one might have imagined.
While the plan is clearly sets out to compete with and prevail over China in the AI race, in quite a few ways it actually echoes what Beijing has been doing with its own national strategy, starting with the “First Generation AI Development Plan,” published back in 2017. These similarities include the all-hands approach to China’s development (and now the United States’ securing) of the entire AI supply chain, and the push to accelerate AI adoption by all business and industrial sectors and the rest of society, as well as by various levels of the government. Even China’s typical references to Party Secretary General Xi Jinping’s speeches and thoughts are mirrored by the U.S. AI plan’s repeated mention of the vision and accomplishments of Trump.
Chinese policy planning in recent years emphasizes the development of the four “new production factors” for its AI aspirations: data, algorithms, computing power, and talent. The Trump AI plan, albeit under different terminologies, has set out strategies to develop the first three of those factors, but is apparently avoiding the last one: talent. In practice, developing AI talent touches on hugely divisive topics, whether that’s supporting U.S. universities to educate and train U.S. citizens or relaxing the granting of work visas for foreign talents. These are areas where the president’s tech sector supporters split with his wider MAGA base.
This is not to say, of course, that there are no other differences between the visions of the two countries. Most notably, China’s acknowledges the need to establish relevant laws over AI usage, but regulations remain an abomination to the Trump administration. Still, for the United States to adopt a de facto industrial policy at all, a relatively nascent practice for the country, is nonetheless undeniably a page from the China playbook.
To Be Continued: Infrastructure, Computing Access, and Export Control
The United States’ turn toward the pursuance of industrial policies is itself another sign of continuity from one administration to the next. Former President Joe Biden’s Executive Order 14110 of October 30, 2023 (“Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence”), despite being revoked by Trump in his EO 14179 (“Removing Barriers to American Leadership in Artificial Intelligence”) on January 23, 2025, bears a number of shared industrial policy implications with the Trump AI plan. These include identifying AI as a geopolitically important strategic asset and encouraging investment in federal AI infrastructure and domestic data centers, particularly in supporting the construction of AI infrastructure on federal land. The Biden EO in fact went further in directing agencies to favor procurement of U.S.-made AI tools to secure the country’s supply chain.
Another mildly surprising point of continuity in the Trump AI plan is its support for the National Science Foundation (NSF)’s National AI Research Resource (NAIRR) pilot. Despite Trump’s waves of funding cuts to academia and the NSF itself, the plan reaffirms the need to “ensure access to large-scale computing power for startups and academics.” It declared a partnership with “leading technology companies to increase the research community’s access to world-class private sector computing, models, data, and software resources as part of the NAIRR pilot,” and pledged to “build the foundations for a lean and sustainable NAIRR operations capability that can connect an increasing number of researchers and educators across the country to critical AI resources.” While more details remain to be seen, it seems that NAIRR is being spared from the axe, at least for now.
The Trump AI plan also continues the policy of tightening “export control enforcement for AI compute,” including exploring location verification features on advanced AI chips, and establishing a new body for global chip export control enforcement. This is in spite of the recent decision to allow Nvidia to sell H20 chips to China. As there are speculations that Nvidia’s H20 chips supply is limited anyway, and the company is simply clearing its inventory now and moving on, it is fair to say that although the administration’s approach to export control may be more flexible and transactional at times, the overriding direction – limiting cutting-edge chip sales abroad – may continue.
To be sure, the major differences between the two administrations remain in the approach to risk mitigation, with Trump doing away with Biden’s requirement for red-teaming and reporting for large-scale and high-risk AI models, the support for clean energy, and the emphasis on “democratic values and civil rights.” Instead, Trump emphasized his version of “free speech” and “free enterprise” approach. In this regard, the implications on future development of the frontier AI models may be huge, but the process has already started this year, before the launch of this new AI plan.
Meanwhile, other new initiatives from the Trump AI plan, such as the support for open-source and open-weight AI, are coming in response to Chinese advances in those areas, especially by DeepSeek, and not fundamentally opposed to the approach of the previous administration. Likewise, the plan’s support for creating “full-stack AI export packages” for “allies and partners” would be welcomed by any U.S. policymaker on either side of the aisle. In other words, any other president would probably do the same thing.
Global AI Governance: To Be or Not to Be?
It is also interesting to observe that the Trump AI plan includes a section on “leading in international AI diplomacy and security,” specifically “counter[ing] Chinese influence in international governance bodies. The plan names a number of international bodies – “the United Nations, the Organization for Economic Co-operation and Development (OECD), the G-7, G-20, International Telecommunication Union (ITU), Internet Corporation for Assigned Names and Numbers (ICANN), and others” – as having put up AI governance frameworks and development strategies.
The Trump AI plan criticizes some of these efforts as advocating for “burdensome regulations,” and specifically calls out “Chinese companies” for influencing and attempting to “shape standards for facial recognition and surveillance.” In response, the administration endeavors to “leverage the U.S. position in international diplomatic and standard-setting bodies to vigorously advocate for international AI governance approaches that promote innovation, reflect American values, and counter authoritarian influence.” This represents a divergence from the Trump administration’s otherwise isolationist attitude or position toward some of these international or inter-governmental bodies.
All in all, the new White House AI plan actually continues the industrial policy setting of the previous administration, albeit with most of the expected Trump characteristics. But in a number of critical areas, such as export controls, infrastructure support, and public AI compute support for research, there are more continuity than there are significant policy differences.
