The rapid increase in high-powered GPUs, driven by AI, has created a looming energy and innovation crisis where data centers don’t have enough power to meet demand for compute; Phaidra uses AI and reinforcement learning to help data centers scale compute by optimizing cooling systems and reducing energy costs, which represent up to 50% of operating expenses
Phaidra, an AI-based control system that helps mission-critical facilities like data centers improve energy efficiency and increase compute infrastructure, today announced $12 million in new funding led by Index Ventures, bringing their total capital raised to $60.5 million. The new funding will allow Phaidra to accelerate expansion of its autonomous control systems, which use reinforcement learning to help data centers optimize their power usage while improving reliability. The announcement comes at a time when rapid AI innovation relies on scaling compute. This, combined with a shortage of power and skilled labor needed to support mission-critical facilities like data centers, is contributing to a global crisis where AI innovation is constrained by a lack of compute infrastructure.
Read More: AiThority Interview Series With Scot Marcotte, Chief Technology Officer at Buck
The data center industry, the backbone of AI, is growing rapidly. In the US, which accounts for 40% of the global data center market, demand is forecast to reach 35GW by 2030, up from 17GW in 2022. But data centers are extremely energy-intensive, consuming 10-50 times the energy per floor space of a typical commercial office building. As AI models proliferate, relying on resource-hungry GPUs, existing markets are struggling to meet demand. In Northern Virginia, the world’s largest data center market, only 0.2% of grid capacity is available. In Ireland, data centers are projected to account for 32% of national electricity consumption by 2026. There is simply not enough excess power available in the world to support AI’s current growth trajectory without major improvements in data center infrastructure and efficiency.
“For as much as we talk about AI and growing compute capacity, we are ignoring the heart of the problem—power,” says Martin Mignot, Partner at Index Ventures. “There is no AI without sufficient energy. The more time we spent researching this problem, the more clearly we saw the importance of investing in solutions that address the downstream effects of computational growth. Phaidra is a perfect example of the right team with the right product in the right place, at a unique moment in time.”
Phaidra’s AI control platform turns industrial facilities into intelligent, self-learning systems that combine the precision of manual methods with the scalability and efficiency offered by AI. It seamlessly integrates into existing building management systems to capture real-time data, feed that data into its cloud-based reinforcement learning agent, and autonomously alter the settings on the individual components to achieve optimal performance of the entire system. And unlike traditional control systems, which slowly degrade and require regular manual programming updates, Phaidra automatically learns and gets better over time. Phaidra’s early customers have seen efficiency improvements amounting to millions of dollars in cost savings, which translates to increased power availability for revenue-generating services.
Phaidra’s founding team brings a rare mix of specific engineering know-how and world-class AI research credentials. While working as a data center operator at Google, co-founder and CEO Jim Gao was inspired by the AlphaGo documentary to explore how machine-learning technology could be applied to improve efficiency in Google’s data centers. His enthusiasm caught the attention of Phaidra’s now-CTO Vedavyas Panneershelvam, who had worked as a primary research engineer on AlphaGo. They joined forces at DeepMind, where they developed AI-powered solutions that reduced the amount of energy needed to cool Google’s data centers by 30%. After publishing their findings, Gao and Panneershelvam were approached by industrial control systems engineer Katie Hoffman, who was leading innovation projects at Trane Technologies and later joined as Phaidra’s third co-founder and COO.
Read More: AiThority Interview Series with Dr. Michael Green, Chief AI Officer at Blackwood Seven
“The data center industry was already growing quickly and is hitting an inflection point thanks to turbocharged AI demand,” says Jim Gao, co-founder and CEO of Phaidra. “Every data center operator knows they will need to be more efficient to meet this growing demand. At Phaidra, we’re on a mission to revolutionize how data centers are managed with an AI-powered control system that not only adapts in real time, but also continuously learns and improves. The result is cost savings on a massive scale and more revenue to drive further growth.”
Today, Phaidra’s team of about 100 includes top software engineers and researchers from Google, DeepMind, Meta, and Amazon, along with specialists from leading engineering firms like Trane and Johnson Controls. The new funding will support continued investment in research and development, implementation, and customer success, along with expanded go-to-market efforts as Phaidra continues its mission of helping data centers worldwide optimize their energy usage.
“When you meet the team and talk to their customers, it’s clear that Phaidra is in a unique position to tackle this problem head-on,” says Martin Mignot, Partner at Index Ventures. “No other startup has the domain expertise and AI research chops to actually deliver on the promise of autonomous controls, to say nothing of the impressive performance and results they’ve already shown. We’re excited to see Phaidra take the next step as a company and deliver solutions that are not only critical to data centers’ growth and sustainability, but to the overall growth and sustainability of AI development, the technology industry, and the world at large.”
[To share your insights with us as part of editorial or sponsored content, please write to psen@itechseries.com]