The race to build AI infrastructure is becoming a race for power. While demand for new data center capacity continues to grow, securing enough electricity – and securing it quickly – has emerged as one of the industry’s greatest challenges.
In many markets, grid interconnection timelines stretch for years, forcing operators to explore alternatives ranging from on-site generation and battery storage to flexible load management and microgrids.
The result is an increasingly fragmented landscape. Developers, hyperscalers, and colocation providers are pursuing many of the same goals, but often through bespoke projects assembled from multiple vendors and technologies.
Although innovation has accelerated, there has been little consensus around how these solutions should fit together into a repeatable deployment model.
The Sijbrandij Foundation believes that something needs to change.
This week, the Foundation launched the Data Center Power Coalition, a group of 12 founding technology and energy partners intended to help standardize how power infrastructure is planned and deployed for AI data centers.
The coalition is built around the Foundation’s open Data Center Power Playbook, an open-source framework that outlines a four-phase approach spanning power-first site selection, initial on-site generation, modular clean-energy buildouts, and accelerated grid interconnection supported by embedded load flexibility.
Rather than promoting a single technology, the coalition brings together companies spanning three interconnected areas: on-site power, load flexibility, and grid interconnection. The 12 founding partners include Amperesand, DG Matrix, Emerald AI, Florrent, GridCare, Hammerhead, Hanwha, Hitachi, NeuralWatt, Planted Solar, Skeleton Technologies, and Voltus.
Leading the initiative is Aric Li, who heads the Sijbrandij Foundation’s Data Center Energy Initiative.
Data Center Knowledge spoke with Li about what prompted the coalition’s formation, why the Foundation believes power co-development will become central to AI infrastructure, and how the organization hopes to turn an open framework into an industry standard.
Aric Li, director of the Sijbrandij Foundation’s Data Center Energy Initiative. (Image: Sijbrandij Foundation)
Data Center Knowledge: What led to the creation of the Data Center Power Coalition? Was there a specific challenge or event that catalyzed this initiative?
Aric Li: Power is the single greatest bottleneck to scaling AI compute. However, we realized that the commercially superior way to solve it wasn’t being widely adopted, simply because it was too complex to execute.
Our conversations across both the demand and supply side made this disconnect obvious, especially at SF Climate Week: energy companies are solving parts of the problem in silos, while the AI industry wants streamlined, packaged procurement, not a dozen vendors to stitch together. So, beyond our open Data Center Power Playbook, which we launched to streamline the energy strategy, we built the coalition to streamline that execution.
DCK: This initiative includes some big names. How do you ensure collaboration across such a diverse group of partners? Are there mechanisms in place to align goals and execution strategies?
AL: Collaboration is structured by design, but the coalition is not forcing all 12 partners to work together. We mapped the subcategories necessary to execute the strategy in our playbook and vetted partners to fill each one – the subcategories are complementary (you need a partner to handle each), but partners within the same subcategory may be competitors, which is intentional as we didn’t want to single-source any category.
Our role as the neutral coordinator is to help a demand-side team assemble the right package across these subcategories for their projects – the goal is a coordinated network that can be accessed seamlessly in one place.”
The Data Center Power Playbook provides a replicable four-phase execution template for powering AI compute at scale. (Image: datacenterpower.ai)
DCK: Can you explain the coalition’s model and how it plans to drive the adoption of solar-and-storage-centric on-site power development and embedded load flexibility?
AL: The Sijbrandij Foundation is a philanthropic non-profit, not a commercial venture – our goal is to help build solutions to problems we care deeply about (no revenue model attached).
The model of the coalition itself is power co-development: planning power alongside compute from the start. On-site, solar-and-storage-centric energy campuses built next to compute enable embedded load flexibility, which in turn unlocks accelerated grid interconnection – ultimately lowering energy rates for ratepayers on the local grid.
We see this as the commercially superior path to powering AI data centers, and by proving it through the coalition and reference projects, we expect to shift the industry’s default to powering AI.
DCK: Power is an issue that touches all aspects of the data center industry. How does the coalition plan to address the unique challenges faced by smaller operators or those in regions with less developed energy infrastructure?
AL: The coalition is built to help operators of all sizes. That said, smaller operators benefit in a specific way: they’re the least likely to have deep internal energy and infrastructure teams and thus face the highest barriers.
Our playbook provides a free open energy strategy resource that anyone can use. Our coalition provides a bridge to the vetted providers they’d need to execute on that strategy. We will also support select operators directly on reference projects. Ultimately, the goal is to lower the barriers for everyone on the demand side, regardless of size.
DCK: Load flexibility is highlighted as a key pillar of the initiative. Could you elaborate on how this concept is being implemented and its impact on grid reliability and cost efficiency?
AL: The key insight to understand is that the grid has a capacity challenge, not an energy challenge – the constraint is a handful of peak hours, not total energy. So, if a data center avoids contributing to those peaks, it actually improves grid utilization, meaning the same fixed infrastructure serves a higher load base, which pushes $/kWh down for all ratepayers, thus earning accelerated interconnection (policy standards actively being established).
We get there with on-site resources like battery storage, among other levers that let a facility operate without SLA impacts but flex from a grid perspective – AI compute still gets firm power at the envelope, just not entirely from the grid. And because data centers are over-provisioned inside the fence, orchestrating operational levers also lets you run more compute on the same power envelope.
So, load flexibility does double duty: it improves local grid affordability and resiliency, while getting you more compute per unit of power.
DCK: How does the coalition plan to address the growing energy bottleneck in AI infrastructure beyond the initial 12 partners? Are there plans to expand the network or include smaller, emerging energy players?
AL: The 12 launch partners are industry-leading companies across the subcategories we’ve identified as essential to executing the playbook’s strategy – and we’ve published our vetting criteria openly for transparency.
Through our vetting process, we’re confident in their ability to deliver on reference projects commencing within the next 12 months.
Both the playbook and coalition are living resources we expect to update roughly monthly, given how fast this space moves and as the coalition expands (anyone can subscribe to our list for updates).
We’re already in active discussions with emerging energy-technology players that may not be commercially ready within 12 months but are highly complementary to the strategy, and we expect to feature them in a future coalition update.
DCK: How does the Sijbrandij Foundation envision the future of AI infrastructure in terms of energy strategy? Are there other initiatives in the pipeline to address related challenges?
AL: We see energy velocity – the speed at which firm, scalable power can be secured – as the defining competitive advantage in the AI race, and co-developed power as the fastest path to it.
The trajectory is framework → coalition → reference projects → scaled adoption: publish the open playbook, convene the coalition to execute it, prove it through real reference projects, and scale from there.
Strategy without execution isn’t worth much – that’s exactly why we built the coalition alongside the playbook. Both ultimately need reference projects to prove the model in practice, so that’s our focus from here.
