According to Financial Times News, KKR’s global head of digital infrastructure argues that AI infrastructure development mirrors historical technology revolutions like electrification and railways, where initial speculative booms often lead to lasting foundational assets. AI hyperscalers in the US are expected to more than double their data center capital expenditure from 2022 levels by 2025, with AI-related capex accounting for approximately 5% of US GDP and growing about 10% annually. The article draws parallels to the dotcom era, where over $500 billion invested in fiber optic infrastructure during the boom created the modern internet backbone despite many companies going bankrupt. Critical infrastructure advantages will include power procurement, permits, land access, and grid interconnections, with even small power cost differences creating massive financial impacts—a 1 cent per kWh variation for a 50MW facility equals roughly $4.4 million annually.
Sponsored content — provided for informational and promotional purposes.
The Power Imperative: Where Location Becomes Destiny
The most significant differentiator in AI infrastructure isn’t processing power—it’s electricity economics. As AI workloads demand unprecedented energy consumption, data center operators who secured favorable power contracts years ago now hold commanding advantages. Regions with stable, low-cost electricity from hydroelectric, nuclear, or emerging renewable sources are becoming the new gold mines. We’re seeing a fundamental shift where data center real estate value is increasingly determined by megawatt access rather than physical location relative to population centers. The International Energy Agency projects that data centers could double their electricity consumption by 2026, creating intense competition for reliable power sources that many analysts underestimate.
Beyond Bragawatts: The Coming Infrastructure Shakeout
Just as in previous technology revolutions, the AI infrastructure market will experience significant consolidation. Companies boasting about their computational capacity (“bragawatts”) without sustainable economic models will struggle when capital becomes more expensive or demand growth moderates. The winners will be those who control scarce resources beyond just silicon—particularly power procurement rights, water access for cooling, and strategic grid interconnection points. We’re already seeing major cloud providers signing decade-long power purchase agreements and even investing directly in generation capacity, essentially becoming energy companies that happen to run data centers.
Where Smart Capital Is Flowing
Sophisticated investors are increasingly bypassing flashy AI application companies to focus on the picks-and-shovels providers with durable competitive advantages. The most attractive opportunities lie in companies that solve critical bottlenecks: specialized cooling technologies for high-density computing, grid modernization services, and firms with expertise navigating complex permitting processes. According to McKinsey analysis, generative AI could add the equivalent of $2.6 to $4.4 trillion annually across various use cases, but capturing this value requires infrastructure that can deliver performance at sustainable costs. The infrastructure players who survive the coming shakeout will likely enjoy pricing power similar to utilities, serving as essential gatekeepers to AI capabilities.
The Emerging Regulatory Challenge
As AI infrastructure expands, regulatory scrutiny will inevitably follow. We’re already seeing local communities pushing back against data center development due to power consumption concerns and environmental impact. Future infrastructure winners will need sophisticated government relations capabilities and environmental, social, and governance strategies that address legitimate community concerns about water usage, energy consumption, and land use. The companies that proactively engage with regulators and local communities to create sustainable development models will secure the permits and social license needed to scale, while those focused solely on technical specifications may find themselves blocked by political and regulatory barriers.
