Why Former AWS Chief Selipsky Joined KKR's AI Infrastructure

KKR has appointed Adam Selipsky, former chief executive of Amazon Web Services, as Senior Technology and AI Strategy Advisor. The move targets a specific problem: AI workloads are breaking traditional infrastructure assumptions about power consumption, cooling requirements and network capacity.

The hire reflects the infrastructure reality facing AI companies today. Training large language models requires sustained maximum compute load for extended periods. Inference workloads create traffic spikes that overwhelm network capacity without warning. These technical constraints are driving billions in infrastructure investment.

How AI workloads differ from traditional cloud computing infrastructure

Traditional cloud infrastructure was designed for predictable enterprise applications with steady resource consumption. AI training runs consume power at maximum capacity for weeks or months. A single training cluster can draw more electricity than entire traditional data centres.

Inference presents different challenges. Models need low-latency responses with unpredictable traffic patterns. A viral AI application can generate demand spikes that traditional capacity planning cannot accommodate. Network latency becomes a constraint on model performance.

These requirements translate into different physical infrastructure. Data centres serving AI applications need higher power densities, enhanced cooling systems, and network connections with greater bandwidth. Power grids face strain from facilities operating at maximum capacity around the clock.

“AI is fundamentally rewiring the internet’s infrastructure – exposing new bottlenecks while creating once-in-a-generation opportunities,” says Waldemar Szlezak, Partner and Global Head of Digital Infrastructure at KKR. “We believe this is a transformational moment for the sector. With AI driving unprecedented demand for compute and power, KKR is advancing a coordinated strategy that unites its data centre, energy and digital infrastructure investments into one platform purpose-built for hyperscalers and AI developers.”

Adam’s AWS experience spans AI infrastructure scaling challenges

Adam’s two decades at AWS bracket the platform’s evolution from basic cloud services to AI infrastructure provider. His first tenure from 2005 to 2016 covered building marketing, sales and support during AWS’s growth into the world's largest cloud platform. He returned as CEO in 2021 as machine learning adoption accelerated infrastructure demands.

Between AWS stints, Selipsky served as president and CEO of Tableau Software, managing the data analytics company through its $15.7bn Salesforce acquisition. This experience included transitioning Tableau's operations to cloud infrastructure – relevant for AI companies moving from research to production deployment.

At AWS, Adam managed capacity planning as customers shifted from traditional workloads to machine learning applications. Training jobs that run for weeks create different resource allocation challenges than web applications with predictable traffic patterns. His tenure included managing Amazon’s sustainability efforts as power consumption became a constraint on AI model development.

Adam currently serves on the board of directors of Circle Internet and previously served on the US Government AI Safety and Security Board, providing regulatory perspective on AI infrastructure development.

KKR’s US$179bn infrastructure portfolio targets AI deployment bottlenecks

KKR has committed US$42bn of equity into digital infrastructure across 23 investments, plus US$20bn in power and renewables. The portfolio spans five data centre platforms across the United States, Asia-Pacific and Europe with more than 155 facilities and a 15GW development pipeline.

The firm operates 12 fibre platforms reaching nearly 30 million homes and maintains over 130,000 wireless infrastructure sites. This portfolio addresses different aspects of AI infrastructure: data centres provide compute capacity, power generation ensures reliable electricity supply, and fibre networks handle data transfer for distributed training.

Power availability determines AI model development and deployment

Infrastructure capacity increasingly drives AI development decisions. Training large models requires access to clusters with hundreds or thousands of GPUs, sustained power supply, and network connections capable of handling parameter updates across distributed systems.

Companies developing AI applications face deployment constraints based on power grid capacity and data centre availability in specific regions. Model training locations depend on where infrastructure providers can guarantee power and cooling capacity for extended periods.

Edge AI deployment for real-time applications requires different infrastructure. Low-latency inference may need compute resources distributed across multiple locations rather than centralised data centres. This creates demand for network infrastructure that can coordinate distributed AI workloads.

“KKR is advancing a coordinated strategy that unites its data center, energy, and digital infrastructure investments into one platform purpose-built for hyperscalers and AI developers,” Szlezak says. “Adam’s appointment marks a pivotal step in that journey and we're thrilled to welcome him as we retool the world’s infrastructure for the AI era.”

Infrastructure investment reflects AI compute requirements

Traditional boundaries between power generation, data centre operations and network connectivity blur as AI applications treat these as integrated systems. Model training requires coordinated resource allocation across all three categories.

KKR’s hiring strategy reflects this infrastructure convergence. Adam’s background combines hyperscale operations experience with understanding of how AI workloads differ from traditional cloud applications. His role will focus on coordinating KKR’s assets to serve AI developers and hyperscale operators.

The appointment signals infrastructure providers recognising AI as a distinct market segment with specific technical requirements. Traditional cloud infrastructure planning assumptions no longer apply to workloads that consume maximum resources for extended periods.

“KKR has taken a visionary approach – integrating power, data centres and connectivity to meet the demands of hyperscalers and AI developers alike,” Adam says. “I’m excited to collaborate with Waldemar and the team to build on that foundation and position KKR as the world's leading AI infrastructure investor.”