Phaidra, a startup using artificial intelligence to make data center operations more energy efficient, this week announced key collaborations with Nvidia, CoreWeave and Applied Digital.
The Seattle company revealed “groundbreaking methodology” that predicts and prevents data center heat spikes when computing workloads surge. Phaidra has been partnering with cloud provider CoreWeave and data center operator Applied Digital to test and deploy the cooling strategy.
As data center operations and deployments boom nationwide, demand for energy and water to run the facilities and cool the electronics is likewise surging. Operators are eager to find better strategies for building and operating such complex sites.
Phaidra is led by alumni from Alphabet’s AI research hub DeepMind, launched in 2019. Its technology uses an array of sensors to measure multiple metrics and analyzes that information. The company has raised a total of $120 million and has roughly 90 employees.
The startup is #78 on GeekWire 200, our list of the top privately held technology companies in Seattle and the Pacific Northwest.
“We envisaged a future where AI agents transform static infrastructure in self-learning, continuously improving infrastructure,” Phaidra CEO Jim Gao said on LinkedIn.
“That future became reality on the world stage,” Gao added, when Nvidia CEO Jensen Huang this week announced the collaboration between Phaidra, the global chip giant, and others.
Data centers typically hum at steady operating conditions, but demand can suddenly ramp up when AI training or other large workloads are dispatched. That cranks up the heat produced, which can cause chips to throttle performance to avoid overheating. To prevent this, data center operators often over-cool facilities, wasting power, water and limiting available compute capacity.
Phaidra’s fix is to use an AI agent that monitors power data as an early-warning signal of an impending operations spike so cooling can kick in proactively — rather than waiting for a temperature rise.
Read the full article here
