If the U.S. Cuts Off the Internet, Can Your Country’s Factories Still Operate? The Economist on the U.S.-China Tech War: Building Data Centers Matters More Than Creating AI Giants

While the world debates whether AI will spiral out of control, another urgent security crisis looms: the future global order may become dependent on two AI superpowers—America and China. On July 15, The Economist pointed out that as AI becomes essential infrastructure for national operations, models, computing power, and data centers are turning into new geopolitical bargaining chips. For other nations, rather than spending vast sums trying to build the 'next OpenAI,' the more practical path is to construct local data centers and safeguard critical technologies—this is the only way to accumulate leverage in the U.S.-China rivalry and preserve autonomy and survival space in the AI era.

Will AI Become a Geopolitical Weapon?

Will AI go out of control? And how can we prevent it? As the world passionately discusses AI safety, the U.S. government has already begun accelerating regulatory efforts. Demis Hassabis, CEO of Google DeepMind, recently advocated in an exclusive interview with The Economist for governments and the private sector to jointly establish a regulatory body similar to a 'financial securities commission,' enabling AI giants to compete within a clear and stable framework.

The urgency in setting rules stems from the astonishing pace at which frontier AI capabilities are advancing—already demonstrating high-risk abilities such as hacking critical digital infrastructure and generating biological weapon recipes. Due to the lack of systematic methods to manage these risks, the U.S. government is currently adopting a step-by-step approach toward cutting-edge models like OpenAI’s GPT-5.6 Sol or Anthropic’s Claude Mythos—allowing usage while simultaneously figuring out how to regulate them. While the ideal scenario would be for the two AI superpowers, the U.S. and China, to reach international consensus, time is clearly running short. Hassabis believes that if the U.S. takes the lead in establishing such a system, the rest of the world will likely follow suit.

While Hassabis makes a compelling argument, The Economist’s deeper concern isn’t just technical失控, but the 'weaponization of AI.' As this technology permeates every industry, the U.S. and China may use their AI models and computing power as geopolitical leverage. Just as Trump once unapologetically used military defense to pressure allies, the U.S. could similarly use AI as a tool of coercion in the future. And China, which has previously threatened the international community with rare earth export bans, may reuse this tactic, treating AI as a strategic bargaining chip in the new era.

It’s Okay If You Can’t Catch Up with OpenAI—Your Real Trump Card Is the Data Center

The absolute dominance of the U.S. and China over frontier AI has undoubtedly backed other countries into a corner. The Economist warns that as AI increasingly becomes essential infrastructure—like water and electricity—being cut off from networks or supply would cause catastrophic damage. Arthur Mensch, CEO of Europe’s leading AI model developer Mistral AI, posed a sharp question: 'If the U.S. suddenly cuts off models and data centers tomorrow, can your factories still operate normally?'

In an era where AI is tied to national security, many countries’ first instinct to avoid being strangled by the U.S. and China is to 'spend heavily to build their own OpenAI.' But The Economist pours cold water on this idea: this path is simply unworkable. OpenAI and Anthropic alone have collectively raised over a thousand billion dollars—government budgets are a mere drop in the bucket. Not to mention that the AI race hinges on luring top talent with astronomical salaries and embracing a venture-capital mindset that accepts total losses—this is simply not the government’s strength. Just as France and Germany’s high-profile 2005 joint 'Quaero' project, which vowed to beat Google, ultimately fizzled out and became a relic of history, the same fate likely awaits such national AI ambitions.

Since building your own OpenAI is out of reach, the most pragmatic form of self-defense is actually building 'local data centers.' The Economist points out that countries should at least build server farms on their own soil to process the most sensitive government and corporate data. This is not just an 'insurance policy' against sudden disconnection from overseas computing power, but also a strategic card to hold—should the U.S. and China ever impose a blockade and cut off access to the most advanced frontier models, at the very least, domestic computing power can run open-source models to maintain basic societal operations.

There Is No Absolute Autonomy—Only More Leverage

Unfortunately, most countries worldwide are already far behind the U.S. and China in data center construction speed. According to data from the Carnegie Endowment for International Peace, the biggest obstacles to building data centers are administrative reviews and power supply. Bureaucratic delays not only lock up capital but are also economically devastating—just a nine-month delay in construction can result in profit losses equivalent to 'doubling the electricity cost' over the entire operational period. Moreover, the efficiency of connecting to national power grids is extremely low: in India and the UK, it takes an average of three years to get connected; in Germany and South Korea, it takes up to three and a half years, while in the U.S., it takes only two.

Therefore, The Economist argues that what governments should do is not to throw subsidies around, but to loosen regulations, simplify environmental assessments and reviews, and even allow tech giants to bypass traditional power grids and generate their own electricity. After all, what AI firms crave most now is computing power, and they are more than willing to pay extra to keep systems running as long as power is available.

In the future, a new paradigm may emerge: 'exchanging computing power for technology.' Countries provide land and electricity to help build data centers, in return for U.S. giants introducing equally advanced models. When countries become major clients of American AI labs, the big firms, eager to secure stable revenue, will naturally lobby the U.S. government to oppose trade protectionist policies that harm overseas customers.

The Economist also notes that a few lucky countries hold exclusive leverage. Taiwan dominates global chip manufacturing, the Netherlands has ASML’s advanced lithography machines, and South Korea controls key memory technologies. Yet even these fortunate few cannot build fully independent, self-sufficient AI capabilities. Therefore, all nations must prepare to engage with AI superpowers through pragmatic negotiations. A democratic U.S. is clearly a better partner than an authoritarian China, but even so, holding more leverage at the negotiating table will be crucial in the future.