Where Data Tells the Story
© Voronoi 2025. All rights reserved.
Artificial intelligence in 2025 is measured in raw compute and energy demand. The latest figures show that power is not evenly distributed. Some nations hold the majority of processing units, while others lag with modest totals despite heavy investments.
The United States has the largest share by a wide margin. It manages about 39.6 million H100 equivalents and owns more than 5.7 million chips. That alone accounts for nearly half of global compute. Energy support is also extensive, with close to 20,000 megawatts available to feed its systems.
The United Arab Emirates holds second place with more than 23 million equivalents, supported by around 188,000 chips. Saudi Arabia ranks third, with over 7 million equivalents and roughly 54,000 chips. Both states are converting oil income into digital strength.
South Korea comes fourth with about 5.1 million equivalents and just over 20,000 chips. India, despite having almost half a million chips, records only 1.2 million equivalents, placing it sixth. China holds more than 628,000 chips, the second-largest stockpile after the United States, yet its compute stands at only 400,000 equivalents. That mismatch reflects reliance on older processors.
France, with nearly one million chips, reports 2.4 million equivalents and ranks fifth. Germany, though industrially strong, holds just 51,000 equivalents and about 32,000 chips. The United Kingdom shows 120,000 equivalents with 52,000 chips, while Finland records 72,000 equivalents and 82,000 chips.
Across the top ten nations, compute adds up to 79 million equivalents, or about 79 exaflops. This is seventy times the capacity of the fastest known public supercomputer. Running these systems at maximum load would require around 55 gigawatts of electricity, equal to California’s summer peak or the combined use of Spain and the United Kingdom.
Spending on AI infrastructure this year reached about 200 billion dollars. Some governments are pursuing chip stockpiles, others are focusing on workforce training or policies to push adoption. The numbers suggest that compute is becoming a core measure of influence, shaping how nations prepare for the next phase of economic competition.