Power is now the primary constraint on data center development; not land, not capital, not compute. With grid interconnection queues stretching five to six years in key markets and ISOs acknowledging only about 20% of queued generation is actually under construction, the old model of waiting for the utility to catch up simply does not work anymore. Developers who move first on power secure their timelines. Those who wait get pushed out by manufacturing lead times and policy friction they never saw coming.
This post synthesizes the key themes from Enverus Intelligence® Research’s (EIR) recent Time to Power webinar, where Enverus Intelligence® Research (EIR) senior analyst Carson Kearl discusses how the market is evolving, and what smart developers are doing about it.
The Queue Is Not Your Friend
Queue-to-commercial-operation timelines have grown roughly 60% since 2017, now averaging over 2,100 days for projects with a first power year in 2025. Large load interconnection faces the same underlying bottlenecks around studies, transmission build-out realities, and a utility workforce that was not built for this pace of growth.
Only about 10% of capacity sitting in interconnection queues will actually get built. The same dynamic applies to both generation and large load projects.
The implication for developers: do not structure your project timeline around a utility interconnection date. Instead, treat the grid connection as a long-term goal and plan for a bridge power strategy from day one, not after you have already discovered the upgrade scope and cost it entails.
Bridge Power and the BYOG Shift
Hyperscalers have pivoted hard toward Bring Your Own Generation (BYOG). EIR is tracking over 40 GW of announced behind-the-meter (BTM) and co-located generation across the country, and this is not a handful of edge cases. It is a structural shift being adopted by Meta, Amazon, Microsoft, Google, Oracle, xAI, Crusoe, and a growing roster of co-locators.
The logic is straightforward: BTM generation can compress a multi-year interconnection wait into a timeline measured in months. For projects targeting a service date in the next 18 to 36 months, the options break down roughly as follows:
- Fuel cells: Fuel cells are currently the fastest path to first power, with some recent deals coming in well under 24 months, though they carry a cost premium.
- Reciprocating engines: Reciprocating engines offer a strong middle ground. There is significant fleet capacity available in the market right now that is not committed to permanent infrastructure, and aggregating enough of it can get a project online quickly at a more attractive heat rate than fuel cells.
- Aero-derivatives: Aero-derivative gas turbines sit between the two on both speed and economics, and carry the added benefit of long-term contract structures (often 10 to 15 years) that insulate suppliers from the eventual grid connection.
Natural gas has been the dominant fuel source for BTM agreements, accounting for more than 10 GW of the total announced capacity. Nuclear, fuel cells, and undisclosed sources make up the remainder.
Why Small Beats Big for Off-Grid Reliability
One underappreciated dynamic in the BTM conversation is the overbuild ratio and how much generation capacity you need to build relative to your peak load in order to ensure reliability.
For a large combined-cycle gas turbine (CCGT) setup serving a 1 GW data center, you are likely looking at a 50% overbuild: you need to have a full extra unit on standby in case one goes offline. With an aggregation of reciprocating engines, that redundancy buffer can drop to as low as 10%, because the probability of 10% of a fleet of hundreds of engines failing simultaneously is far lower than the probability of a single CCGT requiring emergency downtime.
Lower overbuild ratios mean less stranded capital. The tradeoff is higher maintenance costs per unit and somewhat lower thermal efficiency, but for developers who need speed and certainty, the math often works in favor of smaller, distributed generation.
This is also why OFS (oilfield services) companies have emerged as a meaningful new supply source. Liberty Energy, Solaris, VoltaGrid, and others have leveraged their existing relationships with turbine manufacturers and their experience operating distributed generation in demanding environments to pivot into the data center power market. EIR estimates OFS players could meet roughly 25% of BTM data center load growth over the coming years.
Infrastructure Arbitrage: Choosing Where to Build
If bypass strategies are about controlling time to power at a chosen site, infrastructure arbitrage is about choosing sites where time to power is structurally better.
The largest hyperscalers like Microsoft Azure, Meta, Amazon AWS and Google are betting heavily on PJM and MISO, banking large land portfolios across multiple sites to maximize the odds that at least a few clear the data center bottleneck on an accelerated schedule. Smaller developers are finding opportunity in less-contested ISOs like SPP and WECC.
The key metrics for site quality are not just acreage but buildable acreage and power density, meaning how many megawatts can you actually fit per buildable acre. Developers like Vantage Data Centers and Tract are focused on fewer, higher-quality sites with larger average buildable footprints. CyrusOne and Digital Realty rank among the most efficient at converting land into operational megawatts.
Off-market siting opportunities require layering several data sources: parcel buildability screening, substation withdrawal capacity, transmission congestion context, proximity to natural gas supply and pipeline infrastructure, and competitive load analysis. Enverus’s platform connects all of these in a single workflow, allowing teams to evaluate sites in days rather than months.
Supply Chain Risk Is Real — And Time-Sensitive
Equipment lead times are the quiet crisis in this market. Large-frame turbines from GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries are booked through 2028 with a roughly 50/50 split between firm orders and reservations. The window for developers who want large-frame capacity before 2029 is closing fast.
Chip production capacity provides a useful ceiling on how fast AI data center demand can actually grow: roughly 10 GW globally in 2026 and 13 GW in 2027, excluding China. This acts as a natural governor on the most extreme ISO load forecasts with ERCOT projecting 15% annual load growth and PJM projecting 6%, both significantly above what chip supply constraints and actual hyperscaler capex guidance can support in the near term.
The practical implication: align your power procurement timeline with your actual equipment delivery expectations, not the other way around. A generator contract without a clear manufacturing schedule and factory acceptance test booking is not a plan, it is an aspiration.
What Strong Utility Partnerships Actually Look Like
The AEP-Bloom Energy deal has become something of a template for the market. AEP structured up to 1 GW of Bloom fuel cell capacity through an unregulated subsidiary, enabling private contracts with hyperscalers and shifting capital risk away from retail ratepayers. This structure allows AEP to earn private equity-style returns (12 to 15% ROE) rather than the regulated 9 to 10%, which makes it genuinely attractive for utilities, not just developers.
Utilities in Virginia, Texas, and Georgia are actively studying this model. When evaluating a utility partnership, look for: state policy provisions that enable co-location or technology substitution; unregulated subsidiaries with the flexibility to structure non-standard agreements; and a track record of moving faster than the standard interconnection process.
The Long Game: Clean Baseload
Near-term, natural gas in its various forms such as CCGTs, reciprocating engines, aero-derivatives, will dominate BTM power for data centers. But hyperscalers are not stopping there. Meta, Amazon, Google, and Microsoft have collectively signed more than 10 GW in nuclear PPAs, including deals with Constellation Energy, Vistra, Talen Energy, and a growing list of SMR developers.
The realistic timeline for meaningful new nuclear capacity: 2030 at the earliest for advanced reactor pilots, with the 2030s as the decade when SMRs and large nuclear potentially take a significant share of data center load. Geothermal is also worth watching, particularly in Nevada and the California-Nevada border region, where projects that can compete on time to power at sub-36 months and LCOE below $100/MWh will find willing offtakers.
The developers who will be best positioned for the clean baseload era are those who are operating and cash-flowing today, which means securing bridge power now, not waiting for the perfect long-term solution.
Enverus helps data center developers identify viable, power-ready sites faster by connecting land, grid, and power intelligence in one workflow. Reach out to the Enverus team to learn more about the tools and workflows described in this post.
About Enverus Intelligence® | Research
Enverus Intelligence® | Research, Inc. (EIR) is a subsidiary of Enverus that publishes energy-sector research focused on the oil, natural gas, power and renewable industries. EIR publishes reports including asset and company valuations, resource assessments, technical evaluations, and macro-economic forecasts and helps make intelligent connections for energy industry participants, service companies, and capital providers worldwide. See additional disclosures here.
