The rapid advancement of artificial intelligence is driving spectacular growth in data centers, fueled by record investments and strong concentration in the United States. Coface's analysis reveals, however, that this expansion has not yet led to sustainable economic transformation but currently generates exceptional risks such as overheating and imbalances with global repercussions.
Key Highlights
- $475billion invested in IT equipment for data centers in 2025
- $750billion: value of ongoing or planned projects that could be delayed due to the saturation of US energy infrastructure by 2030
- Approximately one-fifth of US GDP growth: share attributable to the boom in AI-dedicated data centers in Q2 2025
- 130 GW: global data center capacity forecast for 2030, 2.3 times more than in 2024.
A bet of gigantic scale
Since the launch of ChatGPT 3.5 in November 2022, AI has crossed over from laboratories to reach the mass market, triggering an unprecedented wave of investments. For 2025 alone, IT equipment investments for data centers reach $475 billion. In the second quarter of 2025, the contribution of the AI-dedicated data center boom represents approximately one-fifth of US GDP growth. The anticipated global capacity for 2030 stands at 130 GW, or 2.3 times the 2024 volume.
(data for the graph in .xls format)
These developments have propelled the stock valuations of major players: in less than three years, the combined market capitalization of Nvidia, Microsoft, Alphabet, Amazon, and Meta has increased by more than $12 trillion.
While this boom already demonstrates macroeconomic relevance, it is not yet transformative. To generate a broader and more sustainable impact, it must evolve from an investment cycle to a productivity cycle that, beyond improving efficiency, also creates new goods and services on a large scale.
Boom threatened by slowdown
Unlike previous digital waves, driven by exponential advances in semiconductor performance and growing network effects, data centers are constrained by tangible resources: land and water, specialized labor and equipment, as well as power and grid capacities. Each billion dollars invested in AI-related data centers requires $125 million in energy sector investments, allocated at two-thirds for the power grid and one-third for production.
In the United States, which represents more than half of newly installed global capacity, grid connection delays can extend over five years. The concentration of investments in a few American hubs increases the risk of local tensions, as evidenced by the tripling of electricity rates observed in certain regions. The value of ongoing or planned projects that may be delayed due to American energy infrastructure saturation is estimated at $750 billion. Moreover, uncertainty prevails regarding future demand.
(data for the graph in .xls format)
Risks of overcapacity and vulnerabilities in the value chain
Projections regarding capacity needs until 2030 diverge by up to 80% depending on sources. This situation increases the risk of overcapacity that would weaken the entire value chain, from cloud providers to hardware manufacturers and service providers. An overcapacity shock would trigger considerable consequences: colocation providers would be affected first, followed by all sector players, causing revenue decline, margin pressure, and cash flow tensions.
"The rise of data centers in the age of AI is a colossal gamble: it promises major advances but also exposes the economy to unprecedented risks of overheating and imbalance. For this boom to have a lasting impact, it will need to be translated into real productivity gains and the creation of new services on a large scale, beyond mere investment growth. At this stage, AI is fueling economic growth without transforming it in any profound way." summarizes Aurélien Duthoit, Information and Communications Technology Economist.
> Download thefull study here
> Coface maintains continuous monitoring and regularly publishes analyses on developments in this sector as well as 12 other business sectors.
