The United States is home to 5,426 data centres—more than all other major countries combined. Additionally, the nation is constructing data centres worth $40 billion, marking a 400% increase since 2022, as noted by the analytical publication The Kobeissi Letter.
For the first time in history, the value of data centres under construction in the US will soon surpass that of office buildings. Three years ago, the ratio was 1/7.
“This is the modern gold rush,” The Kobeissi Letter emphasised.
Since the launch of ChatGPT in November 2022, the volume of data centres being built in the US has grown from $12 billion to $40 billion. Moreover, forecasts indicate continued growth in energy consumption. By 2030, electricity usage by data centres will account for approximately 8.1% of the country’s total. In 2023, the figure was 3.9%.
By 2030, US data center energy energy consumption will account for ~8.1% of US power demand.
To put this into perspective, it was 3.9% in 2023.
Morgan Stanley predicts a 36GW power shortfall in the US due to data center buildout over the next 3 years.
We need more power. pic.twitter.com/QrTnJx3FsC
— The Kobeissi Letter (@KobeissiLetter) October 27, 2025
Morgan Stanley has forecasted a 36 GW power shortfall in the US over the next three years.
However, the number of data centres does not tell the whole story, The Kobeissi Letter noted. China has only about 449, yet its energy consumption is comparable—12 GW versus 14 GW in the US. The reason is larger data centres compared to those in America.
Per capita energy consumption in China is already on par with the EU. It has increased by 350% since 2000.
“This is why we keep reiterating: energy will soon become the world’s most valuable resource. Nations are entering an arms race in the field of artificial intelligence, and the world will need alternative energy sources,” The Kobeissi Letter stated.
Since the launch of ChatGPT, electricity prices in the US have risen by 23%. Since 2020, the figure has increased by 40%. This is significantly higher than the overall inflation rate in the country.
One potential solution to the electricity shortage, according to the analytical publication, is nuclear energy. Unlike solar and wind installations, such stations operate around the clock, which is necessary for the continuous operation of AI. It is also one of the most economical energy sources.
However, building nuclear power plants takes a long time. Currently, no major reactors are being constructed in the US, whereas China has 29.
“Without a significantly larger volume of electricity, the artificial intelligence revolution simply cannot happen. AI data centres will consume about 1,600 TWh by 2035—approximately 4.4% of the world’s energy supply,” the analysts concluded.
Back in July, SAP SE CEO Christian Klein stated that Europe does not need a surge of new data centres to compete in the AI field.