Reindustrialization is happening in the U.S. However, the recent pushback against AI data centers indicates we might not have fully understood the consequences of our past decisions about offshoring and the dependencies created by those actions.
Regardless of whether we build AI data centers in America, it’s clear AI isn’t going anywhere. We have a choice: influence the direction of AI technology through developing local data centers or let other countries take the lead, particularly those eager to dominate in this area. For instance, China has pledged close to $300 billion in data center investments through 2030, not counting private investments. India is enticing foreign cloud companies with a two-decade tax break, while South Korea is offering regulatory advantages. Yet, many voices in the U.S. are calling for a pause.
The implications for national security regarding who manages AI data centers extend beyond just competition with China. If we don’t create these facilities domestically, we could jeopardize everything that data centers hold: the data itself, computing capabilities, and the intellectual property that powers innovation, upholds privacy, and maintains our sovereignty. Furthermore, we would lose our ability to dictate security standards for this essential infrastructure.
If AI data centers serve as the powerhouse of AI innovation, then data is their lifeblood. A vast amount of data is necessary to make AI effective, and that’s something that has to flow through these centers. This data needs to be carefully managed, especially considering privacy aspects, and should be treated as a national asset.
Adversaries will seize any opportunity, viewing the data flowing through these centers as a means to advance their own ambitions in AI. We’ve seen China engage in espionage aimed at advancing its own AI initiatives repeatedly. They might collect data for spying or to enhance their AI models, utilizing methods like adversarial distillation. Additionally, adversarial attacks could disrupt or degrade the data we rely on for AI performance without us even being aware. Either way, the consequences—ranging from large businesses to private individuals—could be severe, exposing sensitive information. On top of that, consider the computing aspect. The U.S. currently hosts a whopping 75% of the world’s high-performance computing, which presents a crucial competitive edge.
Strict export controls on advanced computing technology can help us increase America’s AI lead and collaborate with allies to influence the future of AI. The efforts to boost domestic semiconductor production, especially through the CHIPS and Science Act, are a positive step in solidifying our advantage with new U.S.-based capabilities. But the question arises, where will these chips go once produced? Many will likely end up in AI data centers.
If we shift AI data centers overseas, those chips will follow suit, along with potentially the companies that supply them. This move could negatively affect U.S. resilience and the economy, which has been preparing for a comeback in semiconductor manufacturing for the past few years.
Currently, China’s semiconductor technology lags a few generations behind the cutting-edge solutions due to U.S. export controls on advanced chips and equipment. But if we allow more chips to be produced overseas where we can’t enforce our security standards, it could make it easier for adversaries like China to access these resources, inadvertently boosting their own AI progress.
Depending on foreign AI data centers could hinder us significantly in a crisis, as other nations—not the United States—would determine where to allocate computing power. We can’t guarantee that we would receive any of that capacity, and what we do get is likely to be slow since it won’t be sourced locally.
The current demand for computing power is already quite strained and will likely grow more intense as the call for AI tools increases. This scarcity will affect various sectors, including our military, which may soon face its own computing bottleneck, along with innovators; researchers, students, and startups all depend on sufficient resources to thrive. Keeping AI data centers within American borders ensures that we can prioritize necessary access during emergencies while also nurturing our innovation-driven economy and skilled workforce.
In the mix of data and computing power is AI intellectual property, which consists of the models and their parameters that have helped keep the U.S. at the forefront. The Chinese Communist Party (CCP) has shown an interest in stealing IP and that includes in the realm of AI. They’ve already employed various techniques, including adversarial distillation, to compete against America’s proprietary AI models. Sending AI data centers abroad could essentially be gifting our intellectual property to entities that will leverage it for their own gain, whether for market dominance or military supremacy. Fortunately, the U.S. currently leads in housing AI data centers. But with about 60% of announced projects for this year yet to break ground, we can’t take this lead for granted. As organizations and policymakers look to reevaluate resource allocation narrative and its benefits to local communities, three immediate actions could bolster public confidence and our national security interests.
First, the U.S. should utilize federal land for rapid AI data center development. Back in July 2025, an Executive Order was signed by President Trump to open federal land and simplify the permitting process for commercial AI data centers. This step could help frame AI as a collective benefit rather than a local issue, shifting responsibility onto the federal government to collaborate with data center developers in scaling AI within reasonable means.
Secondly, the Commerce Department ought to utilize the AI Exports Program to facilitate secure global AI data center development alongside our allies. Other nations need AI capabilities just like the U.S., and we should be keen to support their progress. Establishing standards for both cyber and physical security surrounding AI data centers remains critical for ensuring safe and appropriate AI use. The AI Exports Program aims to share American technology with allies and can serve as the means to enforce standard operations for secure AI data centers overseas. By setting clear expectations through this framework, companies can proceed with confidence and other countries will understand how to leverage U.S. technology for their own benefits, reinforcing its status as a more secure alternative compared to Chinese options.
Lastly, we need to allow space for improvements in AI data centers. While speeding up development is essential, we should ask ourselves if the current pace is necessary for staying ahead in AI. Not every proposed data center will come to fruition, but projections suggest we could see up to $7 trillion in global investments in AI data centers by 2030. This surge in computing could lead to new opportunities for innovative AI methods that might not require large data centers or enhance AI models to function efficiently without as much power. We must grant developers breathing room to achieve these advancements and create flexibility in our approach.
Ultimately, America stands at a crossroads; it can either lead or fall behind in the AI revolution, a physical embodiment of its security agenda. The path forward involves fully supporting the development of AI data centers on its home soil.
