Facebook just turned 21. Google is 26. AWS – the best proxy for the modern data center industry – is 19. While the hyperscale giants have matured, the infrastructure they popularized is starting to show its age.

Globally, thousands of legacy data centers – many built between 2005 and 2015 – now face obsolescence in the face of AI’s rising demands. According to the Uptime Institute, over 60 percent of enterprise data centers are more than seven years old, and many colocation assets are even older.

Meanwhile, AI is still in its early days. OpenAI launched ChatGPT in late 2022, and since then, AI-driven compute demand has surged. Industry forecasts estimate that by 2028, AI workloads could account for more than 20 percent of total data center power demand, up from under 3 percent in 2022.

This has revealed a growing gap between infrastructure supply and AI-capable capacity. Traditional facilities – designed for virtualization and storage, not GPU clusters – struggle to support the power density, cooling, and resiliency requirements of inference and training workloads.

Today, over 5,000 US data centers predate the AI era, and globally, that number surpasses 10,000. Most of these sites are not ‘AI-ready’. Yet the pressure to scale compute continues to accelerate.

In Northern Virginia alone, the world’s densest data center region, three major new proposals were denied in the past month due to power scarcity, zoning hurdles, and community opposition.

The situation is no better in other Tier I markets, where permitting backlogs, transmission constraints, and land costs have created an artificial ceiling on new capacity deployment.

Adding to the challenge, greenfield projects are becoming more speculative and expensive. Construction costs now exceed $12-14 million per MW in some metro areas, and lead times for large substations and switchgear can stretch beyond 18-24 months.

Next-gen solutions such as small modular nuclear reactors, geothermal cooling, or photonic AI chips are promising, but likely five to ten years away from commercialization at scale.

That leaves one viable, undervalued asset class: the global fleet of legacy data centers. Too often dismissed due to outdated PUE (averages still hover around 1.7-2.0 for older sites), aging electrical gear, or modest EBITDA, these facilities offer something the market urgently needs: existing footprint, existing interconnects, and existing entitlements.

With a strategic lens, legacy sites can be repositioned quickly and affordably to serve the booming demand for AI inference, content delivery, and low-latency Edge workloads.

Unlike AI training – which requires hyperscale, multi-megawatt GPU clusters housed in purpose-built facilities – inference benefits from metro proximity, lower latency, and predictable workloads. That makes many legacy sites ideal for adaptive reuse, especially those located near fiber routes, cloud on-ramps, or enterprise campuses.

This is not about putting duct tape on old infrastructure. It’s about targeted reinvestment. By upgrading cooling to support 20-40kW racks, increasing electrical resiliency, and enhancing airflow and containment strategies, a site that currently supports 5.5-6.0MW of IT load can reach 7-8MW.

This translates into a 25-40 percent boost in sellable capacity, with retrofit costs in the $5-6 million per MW range – significantly less than greenfield equivalents, which now average $8-10 million per MW in constrained markets.

The EBITDA impact is significant. A 30 percent increase in IT load at stabilized rates can result in a 50-100 percent increase in EBITDA per site, especially if the upgrade enables higher-margin AI tenants. In addition, brownfield retrofits typically complete 6-12 months faster than greenfield builds, shortening time to revenue.

In other industries like aviation, manufacturing, and healthcare, brownfield reinvestment is standard practice. Airports modernize terminals. Hospitals upgrade surgical suites. Automakers retrofit plants for electric vehicles. It’s time for the digital infrastructure sector to adopt the same mindset.

Thousands of legacy sites are waiting to be unlocked. With the right roadmap and modest capital, these facilities can meet tomorrow’s AI demands today – more affordably, more sustainably, and more quickly than any new build can.

At Enabled Energy, we know legacy doesn’t mean low value. Learn more about how we can turn your brownfield facilities into a thriving NextField fleet.