AI needs power, and a lot of it. When a single power line failure can destabilize the largest grid in the U.S. for 11 minutes, it's more than a canary in the coal mine—it’s a siren. Today, we weigh the cost of AI's energy hunger against the solutions that might keep the lights on.
Both True · edition
One Power Line, 3 GW, and a Grid in Peril
The lead
STRAIGHT TALK
One Fallen Power Line Exposed AI Data Centers' Growing Grid Strain
A power line failure near Washington, DC, caused data centers to stop drawing over 3 gigawatts of power in 30 seconds, spiking voltage across the PJM grid. The grid, which serves 67 million customers, took 11 minutes to stabilize. Northern Virginia, home to the world's highest concentration of data centers, saw a repeat of a similar 2024 incident, highlighting an escalating infrastructure vulnerability. (Source: TechCrunch) →
The case for
Data centers are the backbone of AI, supporting applications from healthcare to logistics. Their demand for power is a sign of their importance: AI workloads, especially training models, are energy-intensive and computationally demanding. But the ON.Energy system offers a promising way to absorb these disruptions. By deploying battery-backed systems that stabilize power demand, data centers can become grid-friendly. ON.Energy’s technology, which is being installed at four campuses totaling 3 gigawatts, smooths the spikes and dips that currently destabilize grids. This means data centers could continue to support AI growth while minimizing their impact on infrastructure. Moreover, the potential is transformative: by 2040, data centers are projected to account for 24% of the PJM grid’s load. Innovations like these could enable that growth without constant grid crises, advancing both AI and energy resilience.
The cost
The grid's fragility isn’t just a technical issue—it’s a systemic one. Data centers are already 6% of PJM’s load, and their share is expected to quadruple by 2040. This growth demands massive investments in grid infrastructure, from transmission lines to backup power systems. Meanwhile, ON.Energy’s solutions, though effective, aren’t cheap. Retrofitting campuses with battery systems and power converters requires significant capital, and those costs often fall on data center operators—who may pass them to cloud service customers. Even ERCOT’s planned regulation for 'ride through' capability will likely increase costs for data centers operating in Texas. But the biggest cost may be the missed opportunity to decouple AI’s growth from its energy footprint. If AI continues to rely on energy-hungry data centers without broader adoption of renewables or efficiency measures, we risk locking in a high-carbon, high-cost future.
Terms, plainly
- PJM Interconnection
- The largest electrical grid operator in the U.S., managing power for 67 million people from New Jersey to Illinois.
- Gigawatt
- A unit of power equal to one billion watts, used to measure large-scale energy use like that of data centers.
- Ride Through
- A grid regulation requiring large power users to maintain operation during short disruptions instead of disconnecting.
- Uninterruptible Power Supply (UPS)
- A system using batteries and converters to provide stable power during grid fluctuations or outages.
Context
Data centers are not new to grid strain. The 2024 event on PJM’s grid, when 1.5 gigawatts disappeared due to 60 data centers disconnecting, was an earlier warning. Back then, they were 6% of the load; now they’re 3% higher and growing. The U.S. grid is already under pressure from electrification, renewables integration, and aging infrastructure. ERCOT’s proactive regulation shows some utilities are preparing, but PJM’s experience suggests grid managers can’t wait. ON.Energy’s installations are a step forward, but adoption is far from universal. This week’s incident underscores the urgency of creating grid-resilient data centers before these crises become routine.
Both true
AI’s reliance on energy-hungry data centers is both its strength and its Achilles’ heel. The solutions exist—ON.Energy is proof—but they are not keeping pace with the growth of the problem. We’re in a race between innovation and inertia, and the grid is already showing the strain. The promise of AI is immense, but if its infrastructure keeps flickering, the costs could dim that promise. Both true.
Also today
HUMAN IMPACT
AI Analyzes Surgical Technique to Improve Prostate Cancer Care
The movements of a surgeon in a procedure, called 'surgical gestures,' can be used to predict patient recovery. This finding was made by investigators at Cedars-Sinai. (Source: Medical Xpress) →
Why it mattersAI may improve patient outcomes by refining surgical techniques.
TECH & PLANET
America’s Demand for Power Is Soaring—and So Are the Costs of Building Out the Grid
Equipment backlogs, permitting delays, tariffs, and yearslong waits to connect to the grid are increasing costs. The article was written by Jennifer Hiller and published on July 24, 2026. (Source: WSJ) →
Why it mattersGrid challenges aren’t just about AI—they affect the entire economy.
FRONTIER
Rice Receives Nearly $20M NSF Award for AI-Powered Materials Laboratory
Rice University has received a $19.9 million award from the National Science Foundation to lead a four-year project aimed at using artificial intelligence for materials research. (Source: Rice University) →
Why it mattersAI is driving innovation in materials science with major funding support.
My analysis
The bigger picture
AI is reshaping industries and infrastructure simultaneously. From the energy grid to the operating room, its potential is enormous—but so are the demands it places on systems that were never designed for it.
What it means for you
- AI data centers are causing significant strain on aging electrical grids.
- Solutions like ON.Energy’s battery systems can mitigate disruptions but require heavy investment.
- AI is advancing healthcare by analyzing surgical techniques for better outcomes.
- Grid infrastructure issues are a bottleneck for broader electrification and AI scalability.
- Massive funding in AI-powered research, like Rice's materials lab, signals growing institutional commitment.
AI is rewriting the rules—but the systems supporting it need a rewrite too.
Sources
- TechCrunch — https://techcrunch.com/2026/07/25/one-fallen-power-line-exposed-a-growing-ai-data-center-problem-heres-how-to-fix-it/
- Medical Xpress — https://medicalxpress.com/news/2026-07-ai-surgical-technique-prostate-cancer.html
- WSJ — https://www.wsj.com/business/energy-oil/americas-demand-for-power-is-soaringand-so-are-the-costs-of-building-out-the-grid-f57b5adb
- Rice University — https://news.rice.edu/news/2026/accelerating-discovery-rice-receives-nearly-20m-nsf-award-ai-powered-materials-laboratory
