Kohlberg Kravis Roberts & Co. launched Helix Digital Infrastructure on Thursday to finance and deliver next-generation AI infrastructure for hyperscalers.
The move addresses the accelerating demand for AI services by providing integrated, long-duration capital. As hyperscalers require massive computing power to sustain artificial intelligence growth, the scale of investment needed for physical data centers and energy grids has surpassed traditional financing models.
Helix enters the market with more than $10 billion [1] in committed capital. The firm is backed by a consortium of partners including Nvidia, Vistra, and the Kuwait Investment Authority [1]. This capital will be used to build the physical facilities necessary to house the hardware and power systems required for large-scale AI operations [2].
To lead the new venture, KKR appointed Adam Selipsky as CEO. Waldemar Szlezak will serve as KKR's global head of digital infrastructure [1]. The announcement was made during an appearance on CNBC’s ‘Squawk on the Street’ [4].
The company intends to bridge the gap between the rapid development of AI software and the slower pace of physical infrastructure deployment. By securing significant upfront funding, Helix aims to accelerate the construction of facilities that can handle the extreme energy and cooling requirements of modern AI chips [3].
This strategy allows hyperscalers to scale their operations without bearing the full immediate burden of infrastructure construction. The partnership with Nvidia ensures that the physical sites are optimized for the latest hardware, while Vistra provides essential energy expertise for the power-hungry data centers [1].
“Helix enters the market with more than $10 billion in committed capital.”
The launch of Helix signals a shift in the AI race from software optimization to physical capacity. By assembling a coalition of chip designers, energy providers, and sovereign wealth funds, KKR is treating AI data centers as a critical utility. This vertical integration reduces the risk for hyperscalers and suggests that the primary bottleneck for AI growth is no longer just the algorithms, but the availability of power and real estate.
