The race to build AI infrastructure is reshaping global power, and the winner won't be determined by energy costs or tax breaks, but by how quickly countries can get data centers operational. A new financial model from the Carnegie Endowment reveals that a single year of delay in bringing an AI data center online costs roughly $500 million in lost value, more than any other factor companies consider when deciding where to build.

This finding upends conventional wisdom in policy circles. For years, governments have competed by offering cheap electricity and tax incentives. But the data tells a different story: companies would rather pay double the power prices to operate a year sooner than wait in permitting queues. The implication is stark. As democracies and authoritarian regimes race to host the computing infrastructure that will power transformative artificial intelligence (AI), the countries that streamline their approval processes will win the competition, not those with the deepest subsidies.

Why Does Speed Matter More Than Energy Costs?

The United States currently dominates the global AI buildout. As of May 2025, nearly three-quarters of the world's advanced AI computing clusters were located on American soil, and U.S. projects move faster than those in most other countries. But that lead is fragile, and the reason has little to do with electricity prices.

The Carnegie model analyzed a wide range of factors affecting data center profitability, including construction costs, equipment expenses, operating costs, tax systems, and timelines. The results were clear: time to power is what matters most. A one-year operational delay costs an illustrative 100-megawatt U.S. data center more than 5 percent of its total lifetime value. By contrast, even doubling electricity prices would cost less than a one-year delay. Moderate tariffs on graphics processing unit (GPU) servers, which are essential chips for AI training, would cost about one-third as much as a delay. Removing typical state tax incentives comes in at roughly 60 percent of the cost of a one-year delay.

The reason is straightforward: AI infrastructure is expensive, and every month a data center sits idle is a month of lost revenue. Companies that can start generating returns sooner will outcompete those stuck in approval limbo, regardless of how cheap their power is.

Which Countries Are Winning and Losing the Infrastructure Race?

The Carnegie model ranked countries on their competitiveness for attracting AI data center investment. The results reveal a clear hierarchy, with speed determining placement. The United States and the United Arab Emirates (UAE) top the rankings, largely because major projects move fastest in those countries. Germany ranks worst, where timelines stretch longest.

But these rankings are fragile. The model shows how quickly positions can shift. If the United States experienced a one-year delay in project timelines, it would drop from first to fifth place, falling behind the UAE, Finland, Canada, and India. Conversely, if India accelerated its approval processes by a year, it would jump from fifth place to second, leapfrogging Finland, Canada, and the United States. Even geopolitical shocks matter: if conflict in the Middle East slowed the UAE by a year and a half, the country would drop from first to fourth.

The stakes extend beyond economic returns. The countries that host the buildout will shape the future of AI itself. They will influence who controls the technology, what values it embeds, and how it is deployed globally. If democracies lead, they have a chance to ensure transformative AI is safe, secure, and reflects liberal values. If authoritarian regimes pull ahead, they gain tools to reshape the world order through repression and military dominance.

How Can Democracies Compete Without Massive Subsidies?

The Carnegie analysis makes clear that governments should not rely on expensive subsidies to attract AI infrastructure investment. Instead, they should focus on removing the obstacles that slow projects down. The report identifies three key policy priorities for countries struggling to attract investment:

• Fast-track review processes: Governments should create expedited approval pathways for data center projects that meet certain criteria, such as using clean energy, investing in grid infrastructure, or making tax commitments. As long as governments can actually shorten approval timelines, they have significant flexibility to negotiate favorable terms with developers.
• Grid flexibility and resilience: Countries should reform interconnection queue processes, allow operators to plan for load flexibility under appropriate circumstances, and encourage production of bottlenecked equipment like transformers and cables that delay grid connections.
• Clean behind-the-meter power: Governments should support distributed solar microgrids and wind farm power purchase agreements as a bridge to full grid connections, allowing projects to start operations while waiting for grid infrastructure to catch up.

These measures cost far less than traditional subsidies while directly addressing what companies care about most: getting online quickly.

What Does This Mean for the US-China Competition?

The AI infrastructure race is fundamentally a geopolitical competition. The United States currently leads, but China is mobilizing to close the gap, while Gulf states are using their energy advantages and capital to attract developers. Many traditional U.S. allies risk being left behind. Across Europe, every major publicly reported AI data center combined appears to contain less computing power than the Amazon-Anthropic mega-cluster at New Carlisle, Indiana. Some U.S. allies, including Australia, Italy, and South Korea, have no major publicly known operational AI chip concentrations at all.

The Carnegie report argues that no single democracy, not even the United States, can build the world's AI infrastructure alone. Instead, democracies need to both compete domestically and cooperate internationally. A broader coalition of democracies pooling geography, power generation, capital markets, and supply chain strengths would be stronger, more resilient, and better positioned to ensure transformative AI is developed and governed on democratic terms.

This year alone, America's biggest technology companies will spend approximately $670 billion, or about 2 percent of U.S. gross domestic product (GDP), building compute clusters. Worldwide, companies and governments will pour nearly $1 trillion into data centers with a single goal: building transformative AI. The speed at which these facilities come online will determine not just economic winners and losers, but the future balance of global power.