Data center headlines ricochet between AI demand, the power crunch, and supply-chain bottlenecks.

It’s the perfect storm, creating friction at every stage of data center development and deployment. But this framing misses the point. Time and certainty, not energy, are now the scarcest resources for the industry.

As hyperscalers rush to plan and build 100GW+ campuses, interconnection backlogs surpassed 2,600GW of generation and storage capacity in 2024, while concerns over energy costs and grid reliability continue to rise. In response, behind-the-meter generation has become the default strategy.

By building power onsite, developers can avoid the most time-consuming parts of the process—notably high-voltage grid interconnection, which can take a decade or more from request to operation. It is one of the primary reasons that up to 80 percent of projects in connection queues are ultimately withdrawn.

In the US, federal guidelines and state legislation are increasingly embedding this into law, and even when 'Bring Your Own Power' (BYOP) isn’t explicitly mandated, policy effectively pushes developers in that direction. The 2026 “ratepayer protection pledge” required Big Tech to fund its own grid upgrades, while at the state level, over 60 bills have been introduced across 20+ states since 2025 to similar effect.

“Behind-the-meter” isn’t a silver bullet

While BYOP can reduce delays, it doesn’t remove them. This strategy still depends on new infrastructure, new equipment, new construction, and, critically, new permitting.

Gas turbines, for example, are facing lead times of two to four years, driven by global demand and manufacturing constraints. Skilled labor shortages further extend timelines, affecting everything from manufacturing to installation.

Even if a new gas power plant could be installed tomorrow, it may not be able to operate immediately. New generation assets introduce additional regulatory complexity and often require new permits, and any delay risks triggering re-permitting.

Grid connection offers may lapse if milestones are missed. Environmental permits can be reopened if project specifications change, particularly as most sites are capped under a single air permit. Even minor design adjustments, such as switching turbine models or altering cooling systems, can trigger fresh approvals, especially in regions with strict environmental standards.

Permitting is therefore not the only bottleneck, but it is a visible symptom of a broader issue: every new layer of infrastructure introduces risk. A 2025 survey by the Uptime Institute found that over 60 percent of large-scale projects are delayed by regulatory, permitting, or approval-related challenges.

The pressure on infrastructure is only intensifying. The IEA projects global data center electricity demand will double to roughly 1,000TWh by 2026, equivalent to Japan’s total annual electricity consumption, while data centers could account for up to four percent of global power demand by 2030. At the same time, median US interconnection wait times have stretched to around five years, more than double what they were a decade ago, reinforcing why developers increasingly prioritize “time-to-power” over energy cost alone.

Not all power is created equal

Permitting is far from the only cause of delay, and treating all behind-the-meter strategies as equal obscures a critical distinction. The real divide is between power that adds complexity and power that avoids it. Approaches focused on extracting additional power from existing systems, such as Waste Heat to Power, sidestep many of the constraints associated with new-build generation.

Organic Rankine Cycle (ORC) systems and gas expanders are a clear example. Rather than generating power through additional combustion, they convert waste heat, often from unused exhaust, into electricity. In gas turbine and engine applications, for instance, studies report that ORC‑based waste‑heat recovery can deliver fuel‑economy and overall efficiency improvements up to 30 percent, depending on the configuration and operating conditions.

Crucially, they behave very differently from traditional turbines. They are fuel-free, emission-free, water-free, and do not require new primary generation infrastructure. With a smaller footprint, lower noise profile, and no additional emissions, they also reduce both regulatory burden and potential community resistance.

They are also less exposed to supply chain volatility and can be deployed more quickly, avoiding many of the delays associated with conventional turbine procurement and installation.

While this minimizes risk, whether from component shortages or permitting complexity, it also represents a broader shift in how hyperscalers approach infrastructure: doing more with what already exists, rather than building more from scratch.

The fastest megawatt wins

What is emerging is a broader shift in how the industry defines value. Historically, the focus has been on securing the lowest-cost energy source. Today, the priority is speed, certainty, and minimal friction.

Reducing exposure to supply chain delays, labor constraints, permitting risk, emissions thresholds, and community opposition is now a defining competitive advantage.

Power scarcity is real and here to stay for the foreseeable future, but it is structural and compounded by delays at every stage. In this environment, the most valuable megawatt is not the cheapest; it is the one that can be delivered fastest, with the least resistance. In practical terms, that means looking beyond traditional generation toward solutions that deliver power without adding complexity.