Table of Contents
Raj Agrawal, partner and global head of real assets at KKR, noted that building AI infrastructure requires more than capital investment, pointing to the importance of securing land, power, and other resources needed to develop new facilities
In sum – what to know:
Hyperscaler expansion – A KKR partner said the slowest-growing hyperscaler the company has met with expects to double its cumulative capacity over the next two years, while the fastest-growing projects even steeper growth.
Early AI adoption – The executive argued that AI is “just scratching the surface,” indicating he believes there is significant room for future infrastructure expansion.
Power availability – He also said “there is a massive power shortage,” emphasizing that energy supply will be an important factor as hyperscalers continue to scale AI infrastructure.
Hyperscale cloud providers are planning significant expansions of their data center footprints, highlighting the scale of infrastructure needed to support artificial intelligence workloads, according to comments made by KKR executives at the Morgan Stanley U.S. Financials Conference.
Raj Agrawal, partner and global head of real assets at KKR, said conversations with hyperscale customers point to continued growth in infrastructure demand. According to Agrawal, the slowest-growing hyperscaler KKR has met with expects to substantially increase its capacity over the next two years.
“The slowest growing hyperscaler… would tell you that in the next two years, they’re going to double capacity… relative to the aggregate capacity they brought online since inception,” he said.
Agrawal added that the fastest-growing hyperscaler described an even more aggressive expansion trajectory: “The fastest-growing hyperscaler that we’ve met with would tell you since inception through 2025, in 2026, they will double all their capacity they brought online from 2025 through inception. And in 2027, they’ll double again.”
He said those discussions reinforce his belief that AI adoption remains at an early stage. “We are so early in how we are using AI as companies. We are so early in how we’re using AI as individuals. We’re just scratching the surface,” Agrawal said.
He also compared the current stage of AI adoption to the early days of smartphones, suggesting the technology still has significant room to grow.
.Agrawal also recalled a conversation with a hyperscaler chief executive who estimated future AI deployments could require “1,000x more power than we do today,” adding that he did not consider such projections unrealistic.
The scale of those ambitions extends beyond computing hardware. Agrawal said power availability is becoming a major consideration for future infrastructure development. “There is a massive power shortage,” he said, adding that if hyperscalers follow through on the expansion plans they have shared, “we will need all of the above in terms of capacity.”
Agrawal also noted that building AI infrastructure requires more than capital investment, pointing to the importance of securing land, power, and other resources needed to develop new facilities.
KKR has recently launched Helix Digital Infrastructure, a new data center development platform backed by a planned $10 billion investment and led by former AWS CEO Adam Selipsky. The initiative is supported by partners including the Kuwait Investment Authority, Nvidia and power company Vistra, with a focus on building hyperscale data centers for AI workloads.
Helix aims to address one of the industry’s biggest challenges by combining data center development with secured power supply agreements. According to the company, the integrated approach is intended to reduce delays caused by limited power availability, a growing constraint on new data center projects across the United States and other markets.
As part of the partnership, Nvidia will support deployments with its AI factory-aligned infrastructure technologies, while Vistra will serve as the preferred power provider for Helix investments through its 50 GW generation portfolio across 18 U.S. states.
