The data center industry is racing to build new facilities to meet AI's explosive power demands, but a cheaper and faster alternative is hiding in plain sight: optimizing the thousands of aging data centers already connected to the power grid. By upgrading cooling systems, electrical infrastructure, and adding energy storage to pre-2020 air-cooled facilities, operators can extract significantly more computing power from the megawatts they've already secured, without waiting years for new construction or fighting through permitting battles .

Why Are Legacy Data Centers Sitting on Unused Computing Capacity?

Most older data centers were designed conservatively, oversizing their cooling and electrical systems to handle worst-case scenarios that rarely occur. This means they're operating far below their potential. The industry measures this inefficiency using a metric called Power Usage Effectiveness, or PUE . A PUE of 1.7 means that for every unit of power delivered to computing equipment, 1.7 units flow into the facility overall; the difference goes to cooling, fans, and other overhead. That translates to only about 59% of the facility's grid power actually running AI workloads.

Consider a facility with a 100-megawatt grid connection operating at a PUE of 1.7. It delivers roughly 59 megawatts of computing capacity, leaving 41 megawatts reserved for cooling and electrical systems. By improving the PUE to 1.5 through targeted upgrades, that same facility could deliver 67 megawatts of computing power, unlocking 8 additional megawatts without a single new power line or permit .

Scale that across an entire portfolio and the numbers become striking. A company with one gigawatt of grid connections operating at legacy PUE levels might only deliver 550 to 600 megawatts of computing capacity. Moving those facilities into the mid-1.4s PUE range could unlock 100 to 150 megawatts of additional capacity, all without increasing total grid draw .

What Specific Upgrades Can Unlock This Hidden Capacity?

Retrofitting isn't a single fix; it's a systematic program targeting the biggest power consumers. Cooling typically accounts for the largest share of overhead in air-cooled facilities, so that's where the engineering playbook is most established .

• Cooling System Improvements: Better containment strategies, variable-speed drives on fans and pumps, economizer upgrades, and modernized chilled-water plants can significantly reduce cooling overhead. For the densest server racks, targeted liquid cooling offers additional efficiency gains.
• Thermal Energy Storage: These systems super-cool water during off-peak hours when grid electricity is cheaper, then discharge that cooling during peak demand periods. This approach can shift between 10% and 30% of a facility's peak cooling load to cheaper times, flattening the cooling load profile and keeping chillers operating efficiently .
• Battery and Energy Storage: Adding battery storage to clip peak power demand is straightforward in concept but requires electrical system modifications and utility approval. Thermal storage offers a compelling alternative since it uses water rather than lithium-ion chemistry, avoiding complex fire safety permitting.
• Electrical Infrastructure Upgrades: Replacing aging low-efficiency transformers, consolidating over-provisioned power distribution systems, and tuning redundancy levels can free up additional capacity for computing workloads.
• Workload Flexibility: Shifting delay-tolerant AI workloads to cooler hours or lower-PUE sites across a portfolio lets facilities run closer to their efficient operating point instead of being sized for rare simultaneous peaks.

How to Evaluate Retrofit Economics Against New Construction

The financial case for retrofits is compelling. While upgrades aren't free, they cost significantly less than building a new facility and securing new power interconnections, which increasingly come with financial security obligations. Much of the cost base at legacy facilities is already sunk, including the building shell, site work, grid connection, staff, and security .

A retrofit might add 10% to 15% more billable computing capacity to an existing facility. Since retrofit costs are generally lower-cost capital expenditure than new construction, or can often be structured as service contracts, the returns on newly enabled capacity are much higher. A project earning a 15% unlevered return in the base case may move into the high teens or low-20s with efficiency-driven capacity gains. With typical 50% to 80% debt ratios, levered returns on the right projects can climb from the low-20s range into the 30s .

Speed matters equally. Retrofits can deliver new megawatts of computing capacity in months, not the years required for greenfield builds. This speed advantage is critical in an industry where every month of delayed capacity translates to missed revenue and competitive disadvantage.

Why Isn't the Industry Rushing to Retrofit?

Despite the financial and speed advantages, retrofits aren't getting the attention they deserve. Part of the problem is organizational: the teams that optimize existing facilities are rarely the same teams that deploy capital into new builds, and the latter tend to get more attention from leadership. Part of it is structural; many operators have leases and service agreements designed around current configurations, making retrofit planning more complex. And part of it is simply inertia; the industry has been in build-new mode for so long that harvesting existing capacity doesn't always register as a strategic priority .

Investors weighing where to direct their next round of data center capital face a more nuanced question than simply greenfield versus retrofit. The real question is: "How much stranded capacity are we sitting on, and how fast can we convert it?" The fastest and cheapest new AI capacity is hiding in the unproductive overhead of existing facilities .

As AI demand continues to surge, the industry will need both new builds and efficiency-driven capacity. But sequencing and balance matter. Retrofits deliver new megawatts in months, require materially less capital per incremental megawatt, diminish regulatory exposure by aligning with emerging efficiency mandates like the EU Energy Efficiency Directive, and generate returns that are difficult to match with slower, heavier greenfield investments alone.