Want to get a data center online quickly? Give it some flex.

2026-06-16 · Source: MIT Technology Review · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Emerging Technologies & Innovation · Depth: Intermediate, long

Summary

Flexible data centers are emerging as a critical solution to the growing power demands of AI and the limitations of existing electricity grids. Companies like Emerald AI, in partnership with Nvidia and Digital Realty, are developing "power-flexible AI factories" that can dynamically adjust their power consumption. Emerald AI's Conductor software, tested in simulations like a 2020 Euro tournament energy spike and real-world trials with 256 Nvidia A100 GPUs, demonstrated a 25% power reduction for three hours while maintaining performance. This approach addresses the significant bottleneck of grid interconnection, which can take PJM, a major US grid operator, eight years to bring new generation online. Studies, including a 2025 Duke University report, suggest the US grid could offer an additional 76 gigawatts of capacity if data centers reduce usage just 0.25% of the time. Real-world implementations by GridCare, Google, and Voltus further validate the concept, with GridCare enabling an 80-megawatt capacity increase for Aligned Data Centers in Hillsboro, Oregon, by installing a 31-megawatt battery. This flexibility helps mitigate public opposition to new data centers and supports grid stability, especially with projected US electricity demand increasing 25% by 2030.

Key takeaway

For AI Architects and MLOps Engineers planning new data center deployments, embracing power flexibility is crucial. You can significantly accelerate grid interconnection and reduce infrastructure costs by integrating solutions like Emerald AI's Conductor or leveraging VPPs. Consider prioritizing workloads and implementing on-site storage to manage power draw during peak grid stress, ensuring operational continuity while meeting sustainability goals and avoiding regulatory hurdles. This approach helps you get online faster.

Key insights

Flexible data centers dynamically adjust power draw to integrate with existing grids, accelerating deployment and enhancing reliability.

Principles

• Grids are overbuilt; capacity exists if demand is flexible.
• Flexibility reduces need for new power plants.
• Prioritize critical AI workloads during power limits.

Method

Software like Conductor analyzes AI workloads and grid conditions to dynamically throttle power to specific processors, respecting usage limits while preserving critical performance.

In practice

• Install on-site backup power or storage.
• Fund Virtual Power Plants (VPPs) for demand response.
• Implement workload prioritization for power reduction.

Topics

• Flexible Data Centers
• Grid Modernization
• AI Infrastructure
• Demand Response
• Virtual Power Plants
• Energy Management

Best for: CTO, VP of Engineering/Data, Executive, AI Architect, MLOps Engineer, Director of AI/ML

Related on AIssential

• AI Data Centers Need a Bring Your Own Power Mandate — AI data centers must generate their own power to ensure grid reliability and sustainable scaling.
• Why AI Is Redefining the Future of Commercial Power Infrastructure — AI workloads demand a paradigm shift in power distribution from static to dynamic, intelligent infrastructure.
• Energy Officials Pressured to Expand Grid as AI Demand Surges - Broadband Breakfast — AI and data center growth necessitate rapid grid expansion and diverse energy sources, including nuclear power.
• Google strikes first-of-its-kind deal with a virtual power plant — Virtual power plants offer a novel approach to secure energy capacity for large-scale operations without new infrastructure.
• The AI Boom Is Taking Over the U.S. Power Grid, and This Map Shows Where It’s Happening — The AI boom is rapidly expanding data center infrastructure, placing immense strain on the U.S. power grid and energy resources.
• Pure Data Centres and AVK deploy Europe’s first large-scale microgrid — On-site microgrids provide critical dispatchable power, enabling data center development in grid-constrained regions.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by MIT Technology Review.