It’s time to start viewing large compute facilities-driven load growth as a problem that is solvable. Over the past few months, Luminary Strategies, LLC has taken on multiple aspects of load flexibility analysis and regulatory design questions to help our industry come to terms with what is needed to wise up on grid upgrades and grid operations for large data center compute loads.
We're calling this the Flexibility Compact, and these are its components:
Thanks for reading Luminary Strategies, LLC! Subscribe for free to receive new posts and support my work.
The Flexibility Compact - Luminary Strategies Chart - 2025
Developers must design for flexibility and data customers must be willing to provide it in exchange for tangible benefits - the number one being speed to interconnection, and number two being certainty of grid availability. Data centers as grid customers also want to minimize the risk of forced outages and disturbances to loads, as well as de-risk potential for state-level policies which mandate grid providers deprioritize service to these loads or call for early shut-offs or physical disconnects.
Utilities must build programs to reward flexibility and present the opportunity for those rewards early in the interconnection study process.
Grid flexibility comes in three basic forms, and any or all are needed to motivate speed to market and retain public trust and government trust in the viability of GW-scale operations: (i) physical flexibility of assets and order of interconnection (generators, batteries, modeled load versus COD load), (ii) power electronics and auxiliary equipment flexibility (master site controller, individual parasitic loads and station loads), and (iii) software-trained power consumption flexibility across compute and non-compute loads (workload orchestration software).
Utilities should also develop rigorous, standardized, and transparent large load interconnection and commissioning standards. As a rule of thumb, customers should get faster interconnection in exchange for curtailment opportunity and site design reliability.
Regulators must align interconnection facilities cost recovery with reliability outcomes in market structures that do not look alike around the U.S. Serious choices need to be made about who benefits when ratepayers fund an upgrade versus private parties -- noting utilities have the lowest cost of capital (raise debt at a lower cost) compared to private developers.
[Note: Regulated utilities (like Entergy, PGE, or Dominion) raise capital through a combination of equity and debt, with state regulators setting their authorized return on equity (ROE) (typically between 8.5–10.5%). Utilities issue investment-grade bonds (often rated A or better) backed by stable revenues and regulatory oversight, resulting in low interest rates, especially on debt financing. Private energy developers (renewables, data center builders, independent power producers) usually have higher equity return expectations (typically 12–20%), pay more for debt (unless they are very large and well-rated), and rely more on tax equity, which is more expensive than debt. This begs the question - who should pay, can a utility be reimbursed by a customer for accessing capital markets at a discount instead of shifting capex to the private party?]
Grid Flexibility Components for a Large Load Compute Facility - Luminary Strategies Chart - 2025
At this year’s ESIG (Energy Systems Integration Group) Spring Technical Workshop, utility engineers, grid operators, and industry thinkers all echoed a strikingly consistent message on these points.
What’s New and Why It Matters
Over the next 24 months, hyperscale developers in Texas alone expect to build 10 GW of onsite generation across AI data center campuses. These facilities are enormous, often require hundreds of megawatts at a single site, and increasingly operate on a 24/7/365 basis to serve training and inference cycles. This load is:
Concentrated (often in transmission-constrained areas),
Power-electronics based (which challenges stability models), and
Oscillatory at sub-second timeframes (which breaks how we’ve modeled load for decades).
These dynamics mean old planning paradigms won’t work. We can’t just build our way out with poles and wires. We need a flexibility-first approach that turns load into a grid asset from day one.
From Burden to Backbone: Flexible Load as Grid Reliability
Utilities are being asked to serve AI load on timelines that don’t align with interconnection or transmission upgrades. But there’s a solution already being piloted by a few forward-looking organizations that are treating load as flexible and programmable.
Energy Reports, Volume 13, 2025, Pages 3661-3671, ISSN 2352-4847,
AI Load = Infrastructure Build Problem Statement
The AI era isn’t just about more compute. It’s about what kind of infrastructure we want to build. If we do it wrong, we lock in volatile demand, expensive upgrades, and fragile grid conditions. If we do it right, we build smarter — and open up new value for all customers.
The playbook is clear:
Programmable load with on-site dispatch capability,
Co-located storage that buffers the grid,
Grid-integrated site design from the beginning, and
Voluntary programs that reward customers for building flexibility into their operations.
“We can’t handle unexpected losses of large loads today.” But we can start building reliability around those very loads — if we plan for it. - Overhead at ESIG 2025 - Austin, Texas
Across the country, utilities are under pressure to keep rates low. Yet regulators often reduce utilities’ allowed return on equity (ROE), raising their cost of capital and ultimately making innovation riskier. In today’s environment, innovation needs to be fast, focused, and financeable. No one has the luxury of slow.
That’s why a new kind of partnership is forming — between compute developers and energy strategists — to co-design interconnection and operations strategies that:
Lower cost of capital by de-risking siting decisions (who buys what, who supplies what, on what timescale, and at what price)
Avoid stranded investments by embedding flexibility options up front in site design and operations planning, and
Accelerate time-to-market.
Plenty afoot, and may the most flexible foot win.
— A.S.F.
Thanks for reading Luminary Strategies, LLC! Subscribe for free to receive new posts and support my work.
A very interesting view of the coming load challenge. However, off grid connections to natural gas may obviate the need for such an elaborate solution. It depends on whether one views the problem as static or dynamic. I see it as the latter.
A very interesting view of the coming load challenge. However, off grid connections to natural gas may obviate the need for such an elaborate solution. It depends on whether one views the problem as static or dynamic. I see it as the latter.