Who controls the silicon? Nations without AI compute risk falling behind

As governments and corporations race to harness artificial intelligence’s transformative potential, a new global divide is emerging—not between rich and poor, but between countries with access to advanced computing resources and those without. According to an interactive analysis by The New York Times published in June, nations lacking cutting-edge compute capacity are being left behind in the AI revolution.

The report, led by Adam Satariano and Paul Mozur, with graphics by Karl Russell and June Kim, draws from extensive data on server locations, commercial cloud deals, government investments, and academic partnerships. It reveals a stark reality: only a handful of countries are securing the advanced semiconductor infrastructure required to develop and run large-scale AI models—exacerbating existing geopolitical and socioeconomic inequalities.

The United States, China, and several European countries, including Germany and the United Kingdom, dominate the global AI compute landscape. These nations possess both world-leading chip fabrication facilities and major investments in high-performance computing clusters. In contrast, middle- and low-income countries often lack either the financial capacity or technological infrastructure to compete on the same scale.

A visual analysis within the interactive piece shows that as of 2025, about 80 percent of the most powerful AI systems were hosted in just three countries. Meanwhile, researchers in the Global South must often rely on outdated GPUs or rent expensive time on foreign cloud servers—an unsustainable strategy for developing proprietary models or conducting scalable innovation.

This imbalance has already triggered national security concerns. Officials from the U.S. and European Union warned that dominance over AI compute infrastructure gives nations not only technological leverage, but also the ability to shape development norms and influence global data flows. “Whoever controls the silicon controls the future,” one U.S. official told The New York Times.

In response, China has stepped up domestic semiconductor production, launching state-backed efforts to grow its AI chip ecosystem. Similarly, the European Union has committed billions of euros to build and subsidize chip manufacturing as part of its “Digital Decade” strategy—moves intended to reduce reliance on American and Asian supply chains and develop sovereign AI capabilities.

Academics and policy experts stress that the divide extends into education and research. Only elite universities with high-performance clusters can reliably produce cutting-edge AI scholarship. A 2020 study cited in the Times underscores how the "compute divide" in deep learning has become a key factor limiting who can participate in frontier AI research.

In lower-income countries, this limitation fuels brain drain and stalls region-specific innovation. The inability to access compute hampers responses to local challenges—ranging from agricultural inefficiencies to public health crises. As Dr. Amina Yusuf, an Ethiopian AI researcher, told the Times, “Without access to big compute, we can’t validate locally relevant models—we’re stuck importing knowledge, not building it.”

To address the disparity, various international coalitions and nonprofit initiatives have emerged. A pilot program in East Africa, for example, subsidizes cloud access and provides AI training to universities in Kenya, Uganda, and Rwanda. Another initiative, driven by European universities, enables researchers from Southeast Asia and Latin America to collaborate on open-source AI systems through shared, federated cloud networks.

Experts warn that if the gap continues to widen, AI may serve to reinforce digital colonialism rather than democratize technology. In this scenario, decisions around data ethics, governance, and development could be shaped almost exclusively by wealthy nations and multinational corporations.

Among the proposed solutions: urging global cloud providers to offer tiered pricing for underserved regions, boosting regional investments through public-private partnerships, and establishing international governance structures to ensure compute access is allocated more equitably.

As the Times concludes, the future of AI may depend less on who has the best ideas—and more on who controls the machines capable of building them.

Also read: The future of work is AI-powered—these 22 roles show why