KKR Launches Helix Digital Infrastructure With $10B+ to Build AI Data Centers — Ex-AWS CEO Adam Selipsky at the Helm

KKR has closed over $10 billion in commitments for Helix Digital Infrastructure, a new company built around a single bet: the real bottleneck in AI isn’t models or chips, it’s everything holding them up — power, cooling, fiber, and physical data center space. The announcement came April 30, making it one of the largest private-equity infrastructure plays in the current AI cycle.

Adam Selipsky — who spent four years as CEO of Amazon Web Services before departing in 2024 — will lead Helix as chief executive.

What Helix Actually Does

Helix is not a cloud provider and not a colocation landlord. It sits in between, partnering directly with hyperscalers to build purpose-built AI infrastructure at scale. The mandate covers data centers, power generation and transmission, and connectivity — the full physical stack that large-scale model training and inference requires.

The pitch to institutional backers, which include at least one sovereign wealth fund alongside strategic partners, is straightforward: AI compute demand is accelerating faster than infrastructure can be built using existing channels. Hyperscalers need a dedicated partner with the capital and operational focus to move faster than internal procurement cycles allow.

Selipsky knows that pipeline intimately. During his AWS tenure he watched the gap between customer demand and available capacity become a recurring constraint — and he’s now positioned to attack that problem directly from the infrastructure side.

Why Private Equity Is Racing Into AI Infrastructure

KKR is not alone. Blackstone has committed $150 billion to data center infrastructure over the next several years. KKR itself announced a separate $50 billion joint venture with Energy Capital Partners earlier this year targeting power generation specifically for AI compute loads.

The pattern is consistent: the hyperscalers will spend an estimated $700 billion combined on AI infrastructure in 2026, but a significant portion of that capacity will be delivered through third-party infrastructure operators rather than built entirely in-house. That creates a durable, contracted revenue stream for infrastructure funds — predictable, long-term, and with demand that shows no sign of plateauing.

The Infrastructure Bottleneck Is Real

Power availability has emerged as the defining constraint in the current cycle. New hyperscale data centers can be designed and permitted, but getting grid connections at the scale AI training requires takes years under existing regulatory frameworks. Companies like Helix that can secure power agreements, land, and fiber rights in advance — before a hyperscaler signs the contract — have a structural advantage.

The Helix announcement follows a broader recognition that the AI infrastructure buildout is no longer a pure tech play. It’s a real assets problem: long construction timelines, utility negotiations, zoning, and the unglamorous logistics of moving megawatts of power to where the GPUs are.

Selipsky’s background gives Helix credibility with its anchor customers. The open question is execution: building and operating AI-optimized data centers at this scale, with the power density and cooling requirements modern GPU clusters demand, is materially harder than conventional enterprise colocation. Helix is raising the capital to try.