Flexible data centers could ease grid bottlenecks

Emerald AI and partners are testing whether data centers can act as flexible grid resources rather than fixed power loads. The approach could speed interconnection, though skeptics warn it cannot replace new power and transmission.

Emerald AI is preparing to deploy its Conductor software at a new facility in Virginia’s Data Center Alley with Nvidia and Digital Realty, aiming to show that data centers can reduce electricity use when the grid is under stress while keeping priority workloads running. The system was tested in simulations including a UK demand surge during the 2020 Euro tournament, when millions of kettles would have sharply raised electricity demand.

Grid flexibility has become a possible answer to a major constraint on AI infrastructure: power connections can take far longer than data center construction. PJM needs eight years to bring new generation online, according to RMI. A 2025 Duke University report found the US grid could support an additional 76 gigawatts for facilities willing to cut usage just 0.25% of the time, or about 22 hours a year. A separate Princeton-linked study found a 500-megawatt facility able to flex for less than 1% of the year could begin full operation three to five years faster than an inflexible one.

Other companies are pursuing related approaches, including GridCare’s digital grid modeling, Google’s workload shifting, and Voltus’s virtual power plant programs. Critics argue flexibility is an optimization tool, not a substitute for new generation, transmission, and distribution. PJM’s market monitor warned that relying on large data centers to reduce demand without enforceable controls amounts to magical thinking.

72

Impact Score

California lawmakers align AI safety and auditor bills

Sen. Jerry McNerney and Assemblymember Rebecca Bauer-Kahan plan a paired framework for voluntary AI standards and independent verification. The effort would create a state commission and a registry for third-party auditors.

Lexar tests SSD offloading for local AI models

Lexar is developing an AI-focused SSD approach that shifts some local model workloads from DRAM to NAND Flash. Internal tests point to lower memory requirements for running larger models on consumer PCs.

NVIDIA Blackwell leads MLPerf Training 6.0 results

NVIDIA’s Blackwell platform posted the fastest time to train across all seven MLPerf Training 6.0 benchmarks, including new mixture-of-experts workloads. Submissions also showed large-scale runs on GB200 NVL72 and GB300 NVL72 systems.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.