KKR Makes the Case for Durable AI Infrastructure

Institutional capital is doubling down on compute, power, and connectivity as the defining asset class of this decade.

What Happened

KKR published a detailed investment thesis in November 2025 titled "Beyond the Bubble," arguing that AI infrastructure spending will compound well beyond the current cycle of enthusiasm. The firm noted that the four largest hyperscalers are expected to spend more than $350 billion on capital expenditure in 2025 alone, with total sector spending approaching half a trillion dollars. AI related capex now represents roughly five percent of U.S. GDP and is growing at around ten percent per year, placing it on par with the capital intensity of the late 1990s technology boom.

The piece acknowledged the presence of froth in asset prices and predicted a shakeout among weaker business models. But it drew a structural distinction between speculative excess and the underlying demand for physical infrastructure. KKR pointed to data showing that colocation vacancy in North America will remain scarce through at least 2027, suggesting that overbuilding is not the near term risk many observers assume it to be.

Structural Context

What makes this analysis notable is not the optimism itself but the framing of where value concentrates. KKR identified power, land, grid connections, and permits as the core structural moats in data center investment. These are not digital assets. They are physical, location specific, and increasingly scarce. The firm argued that those who control these inputs will earn compounding returns, while those who merely lease them at thin margins will face erosion over time.

This framing aligns with a broader shift visible across institutional capital. Private equity investment in data centers reached $108 billion in 2024, triple the level of the prior year. KKR itself has committed more than $42 billion of equity into digital infrastructure across 23 investments, alongside $20 billion in power and renewables. Its $50 billion strategic partnership with Energy Capital Partners, announced in late 2024, is designed to co locate compute and power generation from the start rather than treating energy as an afterthought.

The first project under that partnership, a 190 MW hyperscale campus in Bosque County, Texas, illustrates the direction of travel. The facility will sit adjacent to a natural gas power plant with a dedicated long term power agreement. The logic is clear: rather than competing for congested grid capacity, the project brings compute directly to the source of generation.

The Enki Perspective

The KKR thesis validates a principle that sits at the center of Project Enki's own infrastructure model. The constraint on AI expansion is not demand, nor capital, nor even the availability of chips. It is access to power. And specifically, it is the mismatch between where power is generated and where grid connected capacity is available. Transmission infrastructure has not kept pace with the acceleration of AI workloads. Permitting timelines for new grid connections stretch into years. Meanwhile, significant volumes of renewable energy are curtailed or stranded because there is no local offtaker.

Enki operates at precisely this intersection. By converting stranded and curtailed energy into scalable AI infrastructure, Enki sidesteps the transmission bottleneck entirely. Where large institutional players like KKR are co locating data centers with dedicated power plants, Enki extends the same logic to underutilized renewable generation, particularly in regions where curtailment is highest and digital infrastructure is most needed.

This is not a competing model. It is a complementary evolution. Institutional capital is validating the thesis that infrastructure must follow energy, not the other way around. Enki applies that thesis at a scale and in locations that institutional megaprojects do not typically reach, unlocking capacity that would otherwise remain idle.

What This Signals

When one of the world's largest alternative asset managers commits tens of billions to the idea that power and land are the defining moats in AI infrastructure, it sends a clear signal about where the sector is heading. The era of treating data centers as digital assets disconnected from energy systems is ending. The next phase of compute expansion will be shaped by those who can deliver integrated power and compute, reduce permitting timelines, and offer repeatable deployment models that institutional capital can underwrite at scale.

KKR's analysis also reinforces the importance of geographic distribution. As AI workloads grow, the concentration of compute in a handful of established markets creates systemic risk. Infrastructure models that can activate new locations by aligning with available generation, rather than waiting for grid expansion, will play a structurally important role in the next wave of deployment.

The direction is set. The question is no longer whether AI infrastructure will scale, but how fast the energy layer can keep up. Models that bring compute to power, rather than power to compute, are positioned to close that gap.

Source: November 2025 https://www.kkr.com/insights/ai-infrastructure