The launch of DeepSeek-R1, a groundbreaking AI model from China, is more than just a technological milestone—it’s a wake-up call for policymakers in the United States. DeepSeek-R1, developed by a Chinese firm widely believed to have substantial government backing, promises potential advancements in fields as diverse as medical diagnostics, weather modeling, and defense simulations. While American companies remain leaders in many aspects of AI development, the rise of models like DeepSeek-R1 highlights how close the competition with foreign rivals is. There’s one area where America is shooting itself in the foot: red tape and process requirements are radically slowing down the construction of infrastructure needed to support cutting-edge AI technologies.
The heart of AI innovation lies in data—the ability to process, analyze, and learn from massive datasets at breakneck speeds. But that requires data centers, and not just any data centers: we’re talking about advanced facilities equipped with high-performance computing (HPC) clusters, state-of-the-art cooling systems, and uninterrupted access to affordable energy. Building and maintaining these centers isn’t just a technological challenge; it’s also a regulatory one. Unfortunately, the United States has made it unnecessarily difficult for companies to build the infrastructure required to keep pace in the AI race.
The Red Tape Problem
Data centers are not like traditional office buildings; they are complex engineering marvels that require significant investments of time, capital, and resources. In the United States, getting approval to build these facilities often involves navigating a maze of local, state, and federal regulations. From zoning restrictions to environmental impact assessments to utility hookups, the regulatory hurdles can delay projects for years.
This isn’t just a hypothetical issue. In 2021, a report found that permitting delays for infrastructure projects from the National Environmental Protection Act (NEPA) alone–just one of the many laws imposing red tape—cost the economy over $229 billion. These delays are particularly pronounced in states where the demand for data centers is highest, such as Virginia, Texas, and California.
Contrast this with China, where government-backed initiatives have streamlined the permitting process for tech infrastructure. Chinese AI developers didn’t have to wait years for local governments to greenlight their facilities; they benefited from centralized planning that prioritized speed and scalability. This isn’t an argument for adopting China’s centralized model wholesale, but it does underline the need for reform. If America’s AI sector is going to remain competitive, it needs a regulatory environment that matches its innovative spirit.
Executive Order 14141: A Step in the Right Direction
One positive development came in 2024 with the issuance of Executive Order 14141, titled “Advancing United States Leadership in Artificial Intelligence Infrastructure.” This directive, aimed at accelerating the deployment of clean energy projects, included provisions to streamline permitting processes for infrastructure critical to AI development, such as renewable energy projects and grid upgrades. By cutting red tape and establishing clearer timelines for approvals, the order marked a significant step toward addressing the bottlenecks that have long plagued U.S. infrastructure development.
Importantly, EO 14141 recognized the interplay between energy policy and AI competitiveness. AI data centers are energy-intensive operations, often requiring as much electricity as a small city. Ensuring these centers have access to reliable, affordable, and sustainable energy isn’t just good policy—it’s a national security imperative. The executive order’s focus on modernizing the grid and promoting energy efficiency aligns directly with the needs of the AI sector.
However, one executive order is not enough. The new administration should use EO 14141 as a foundation for broader reforms. For example, the administration could remove a few of the remaining requirements that interfere with rapid construction of key AI and energy infrastructure, such as onerous labor requirements. In addition, owing to the new administration’s trifecta, it could work with Congress to pass legislation that sets nationwide standards for permitting timelines, reducing the patchwork of state and local regulations that currently slow projects down. Similarly, federal incentives for private investment in AI infrastructure could help bridge the funding gap for companies looking to build next-generation facilities.
Why This Matters
Some might wonder: why focus so much on data centers and energy infrastructure when the U.S. already has leading AI firms like OpenAI, Google DeepMind, Meta AI, xAI, Anthropic, and more? The answer lies in the scale of what’s coming next.
As AI models grew more sophisticated, their computational requirements increased exponentially. OpenAI’s GPT-4, for instance, was already a massive leap forward, requiring hundreds of GPUs and specialized hardware to train. Most later models, including those designed for multimodal capabilities or real-time applications, demanded even greater compute resources. DeepSeek’s new R1 model broke the mold, having been trained in a manner that used significantly less compute than comparable models, but even if these efficiencies in training become the new norm, most experts expect a Jevons Paradox to dominate. In a Jevons Paradox, when an efficiency gain reduces the input requirements for a previously supply-constrained resource into an industrial process, the result is increased overall demand for that resource as new uses are developed and increased resource consumption in turn. If training existing models now requires less compute as AI firms incorporate R1’s training efficiencies, then in all likelihood additional compute will be used to train even better models at the same cost, and some additional compute may be utilized on the inference side to improve user experience. Without the infrastructure to support this growth, American companies risk being outpaced not just by Chinese firms, but by others around the world.
Moreover, the benefits of investing in AI infrastructure extend far beyond the tech industry. Data centers create jobs, drive demand for American energy, and strengthen local economies. They also enhance America’s ability to tackle societal challenges, from improving healthcare outcomes to mitigating weather-related natural disasters to bolstering national defense. Falling behind in AI infrastructure isn’t just an economic risk; it’s a strategic one.
The Road Ahead
DeepSeek-R1 should serve as a reminder that technological leadership is not guaranteed. For the United States to maintain its edge, it must address the structural challenges that are holding its AI sector back. That means cutting red tape, modernizing the grid, and creating an environment where innovation can thrive.
Executive Order 14141 was an encouraging start, but it’s only the beginning. Policymakers at all levels of government need to recognize that AI is not just a buzzword; it’s the backbone of the 21st-century economy. By making it easier for American companies to build the data centers and energy infrastructure they need, we can ensure that the leading AI models are developed here, and that American businesses and consumers are the first to benefit from the new inference capabilities of those new models. This is how the United States can ensure it continues to be the global leader in AI innovation.