AI’s Energy Challenge Is Reshaping Private Equity Deals

|
7
|
Financial Times

Carlyle’s planned sale of Copia Power to EQT for $2.6 billion demonstrates how rapidly artificial intelligence is changing the infrastructure investment market. The transaction is expected to generate a return of more than five times Carlyle’s original investment, providing a notable example of a successful private equity portfolio company sale at a time when many sponsors continue to face challenging exit conditions.

Carlyle created Copia Power through its infrastructure business in 2021. The company develops large-scale data centre campuses across the south-western United States by acquiring land, securing grid connections and combining those sites with renewable and conventional power generation. This integrated model addresses one of the most significant constraints facing the technology sector: access to reliable electricity at the scale and speed required by modern data centres.

Copia currently has 2.6 gigawatts of power generation and storage capacity either operating or under construction. Its wider development pipeline includes approximately 20 gigawatts of thermal and renewable generation projects. According to the supplied figures, this amount of capacity would be sufficient to power the equivalent of around 20 million homes.

The company is also developing 9 gigawatts of power capacity intended to support data centre campuses in the south-east and Midwest. Copia’s Harquahala campus in Arizona covers more than 25,000 acres and is designed to use a combination of solar generation, battery storage and thermal power. This mix reflects the growing demand for energy systems that can provide both lower-carbon electricity and continuous power availability.

Power Is Becoming the Critical AI Bottleneck

The expansion of artificial intelligence is increasing pressure on electricity systems across the United States. Data centre developers require large volumes of power for computing equipment, cooling systems, network infrastructure and backup capacity. Yet electricity supply, transmission connections and permitting processes often develop more slowly than demand.

US electricity consumption is forecast to increase by 21% by 2030, according to the consultancy estimate included in the source material. Much of this growth is expected to come from data centres, advanced manufacturing and broader electrification. As a result, access to power is becoming a strategic asset rather than a standard operational requirement.

This shift is influencing the design of new data centre investments. Developers can no longer focus only on land, buildings and computing equipment. They increasingly need to secure generation capacity, grid access, water systems and long-term relationships with utilities before construction can proceed.

Copia’s business model is therefore particularly attractive. It combines land development, power generation and data centre infrastructure within one platform. For investors and technology companies, this structure may reduce development risk and shorten the time required to bring new computing capacity online.

EQT Builds an Integrated Infrastructure Platform

EQT’s acquisition of Copia fits within a broader strategy to provide large technology companies with a more complete infrastructure solution. The Swedish investment group manages more than $300 billion in assets and already has significant exposure to data centres, renewable energy and utility-related services.

Its portfolio includes EdgeConneX, a data centre operator with 90 facilities across 20 countries. EQT has also participated in a $33.4 billion transaction involving power producer AES and has invested in Seven Seas Water Group, which provides water management services relevant to data centre operations.

Across data centres and renewable power, EQT has more than $100 billion invested. Adding Copia could strengthen its ability to connect data centre campuses directly with electricity generation and utility networks. This gives the group exposure to multiple parts of the AI infrastructure value chain, including facilities, energy supply, storage, grid connections and water management.

The acquisition also shows how private capital firms are moving beyond ownership of individual data centre buildings. The emerging model is based on integrated platforms that can solve several infrastructure constraints at once.

AI Infrastructure Drives a Wider Deal Boom

Copia’s sale is part of a broader increase in transactions involving energy and digital infrastructure. A record $240 billion of power sector deals was announced during the first half of the year referenced in the source, representing a 90% increase compared with the same period a year earlier.

Other major investment firms are pursuing similar opportunities. KKR launched a $10 billion data centre initiative intended to address energy and technology constraints through partnerships with companies including Vistra Energy and Nvidia. DigitalBridge also moved to acquire energy investor ArcLight as part of its effort to secure additional power resources.

Carlyle has expanded its own exposure to the connection between energy and data centres. The group, which manages approximately $475 billion, partnered with Diversified Energy on the acquisition of $1.2 billion of oil and gas assets. It also sold power producer Cogentrix for $3 billion in 2024.

The Copia transaction highlights a central reality of the AI economy. Computing capacity cannot expand without corresponding investment in power generation, storage, transmission, land and water. Investors able to combine these assets into scalable platforms may capture a growing share of infrastructure spending.

For private equity, this creates both an investment opportunity and a route to stronger exits. Carlyle’s expected fivefold return suggests that well-positioned power platforms can attract significant strategic demand. For EQT, the $2.6 billion acquisition represents a long-term bet that electricity access will remain one of the most valuable components of global AI infrastructure.

You might also like
Scan the code