On May 19, 2026, I had the opportunity to join Assistant Secretary Katie Jereza, U.S. Department of Energy, Office of Electricity, and fellow industry leaders at the National Laboratory of the Rockies for a critical roundtable on data center power and AI infrastructure challenges and resulting opportunities.
Notably, the discussion coincided with the unveiling of a major new capability: the Agora large-load grid integration test bed—a first-of-its-kind national platform designed to help data centers actively participate in grid reliability.
The alignment between the roundtable themes and the purpose of Agora could not have been clearer: we are entering a new phase where AI infrastructure and the grid must evolve in lockstep—and where collaboration is no longer optional, but essential.
AI growth is reshaping the grid conversation
AI is fundamentally accelerating demand for data center capacity, and grid infrastructure is struggling to keep pace. At the roundtable, participants agreed: AI timelines are outpacing traditional grid planning cycles, creating friction between speed-to-market and system reliability.
One of the most important shifts we discussed is the segmentation of AI workloads:
- Training workloads, which can tolerate lower availability in exchange for faster access to power.
- Inference workloads, which remain highly sensitive to latency and downtime.
This distinction opens the door to more adaptive, workload-aware infrastructure models—but only if we can validate them with confidence.
Reliability still defines the boundaries of innovation
Despite the urgency, reliability remains a non-negotiable constraint.
Data Center Operators are understandably cautious about introducing novel technologies into the critical power paths. As a result, innovation often lags behind what is technically possible.
This is precisely where initiatives like Agora become pivotal. The test bed replicates the technical complexity of large-scale data center interconnections, allowing stakeholders to evaluate new approaches—such as flexible load management or advanced power architectures—under real-world conditions.
In other words, it provides what the industry urgently needs: proof before production.
From theory to practice: The role of validation infrastructure
One of the clearest conclusions from the roundtable was that validation is the missing link.
We need environments where we can:
- Evaluate new power and cooling architectures.
- Model and then test interactions between workloads and grid behavior.
- Demonstrate reliability under realistic conditions.
Historically, data centers have been treated primarily as passive consumers of energy. But as Agora highlights, the future is different: data centers can—and must—become active participants in supporting the grid to increase power stability.
For example, flexible operations could allow facilities to reduce demand during grid stress events, helping prevent outages and lowering system-wide costs. They may also inject power into the grid to increase overall reliability.
Flexibility is emerging as a strategic advantage
If reliability defines the guardrails, flexibility is becoming the key enabler of progress.
Roundtable participants emphasized that accepting differentiated service levels—particularly for AI training clusters—can unlock faster interconnection timelines in constrained markets.
Agora directly supports this shift by providing a “plug-and-play” environment where utilities, data center developers, and technology providers can jointly explore these flexible models.
Addressing structural barriers to growth
The discussion also surfaced several systemic challenges that must be addressed:
1. The Modeling Gap
Today, utilities and developers often lack integrated tools to align power system studies with digital workload requirements. Agora’s ability to study both simultaneously is a significant step forward.
2. Speculative Load (“Phantom Demand”)
Uncertain or inflated load requests complicate grid planning and slow down real projects.
3. Rise of Self-Generation
Behind-the-meter generation is increasingly attractive—but its full value lies in how it can support both project economics and grid resilience.
4. Public Trust and Transparency
As data centers grow, concerns around energy use, land, and cost allocation are becoming central to deployment success.
A new mandate: Collaboration at scale
A consistent message from both the roundtable and the Agora launch is that no single stakeholder can solve this challenge alone. Bringing together utilities, data center developers and operators, technology providers, and research and policy makers become essential to foster and ultimately adopt innovation. This will allow the industry to answer a critical question: how do we expand AI infrastructure while protecting affordability and reliability for everyone?
Implications for the industry
Looking ahead, several priorities stand out:
- Standardize workload segmentation to guide infrastructure design
- Scale validation platforms like Agora and others to accelerate innovation
- Improve data-sharing and modeling integration across stakeholders
- Develop new commercial frameworks that reduce speculation while enabling speed
At Schneider Electric, we see a clear opportunity to help bridge the grid–data center interface through integrated solutions that leverage advanced modeling tools like Digital Twins, introduce flexible power architectures, and raise the need for intelligent energy management systems.
Closing thought: Building the system around AI
The insights from this roundtable reinforce a critical shift:
We are no longer just building data centers—we are redesigning how they interact with and support the greater energy system itself.
AI is forcing us to move faster, and it is also pushing us to move smarter—toward a future where data centers are not just energy consumers, but active, trusted partners in grid reliability.
If we succeed, the outcome will be more than increased capacity.
It will be a more resilient, flexible, and transparent energy ecosystem—ready to support the full potential of AI.
Add a comment