Demand for electricity is growing rapidly because of artificial intelligence.
In May 2024 reportGoldman Sachs predicts that data centers will use 8% of total U.S. power by 2030, up from 3% in 2022, as cloud providers expand to fulfill demand for AI infrastructure. Assuming current trends proceed, U.S. utilities will need to speculate about $50 billion in generating capability to support all the upgrades — and latest — AI-enabled data centers.
There might be serious negative externalities. In Kansas, where Meta recently began construction huge latest server complexEvergy power plant announced that it could delay the retirement of a coal-fired power plant by as much as five years. Some experts say that energy-hungry data centers — which are also large water absorbers — could contribute to rising utility costs for on a regular basis taxpayers, which is able to have a disproportionate impact on low-income earners.
The problem of energy consumption in data centers seems insoluble. But Jim Gao, Katie Hoffman and Vedavyas Panneershelvam, co-founders Phaedraimagine it is possible to retrofit existing facilities to make them more energy efficient.
In fact, they built a business on it.
Phaidra, launched in 2019, creates AI-powered control systems for data centers, in addition to pharmaceutical and industrial building infrastructure. The company’s systems collect data from hundreds of sensors throughout a facility and make real-time decisions about easy methods to cool equipment inside in an energy-efficient manner.
For many data centers, cooling is one of the most energy-intensive components. The average data center cooling system consumes roughly 40% of the center’s total capability.
“The data center industry is in an arms race to build new capacity wherever there is land and energy,” Gao told TechCrunch. “Phaidra’s service can provide a more stable cooling system that uses less energy.”
Gao previously led DeepMind Energy, a team inside Google’s DeepMind AI research division responsible for commercializing technologies to handle climate change. While at DeepMind, Goa—along with Panneershelvam, then a research engineer at DeepMind—developed an AI system to regulate and optimize energy usage in Google’s data centers that gained much attention at the time.
DeepMind has decided to quietly wind down DeepMind Energy after failing to secure deals with major industry players similar to British utility National Grid, in keeping with CNBC reportGao left in August 2019, and Panneershelvam in May 2020 — months after the departure of DeepMind co-founder Mustafa Suleyman, who was reportedly a key driving force behind DeepMind’s efforts to combat climate change.
After leaving DeepMind, Gao and Panneershelvam saw an opportunity to use the lessons learned from Google’s data center project to other data centers—and beyond. They recruited Hoffman, who had led innovation projects at Trane, a cooling company, to launch Phaidra.
Phaidra develops AI models for each client trained on sensor data to optimize a facility’s cooling systems (e.g., a data center) and overall energy management. These models are self-improving, Gao says, always learning from his own experiences managing the facility’s infrastructure.
“One of the unique approaches Phaidra takes to AI is to combine physical knowledge of how a facility works with learned models of plant dynamics based on sensor data,” Gao said. “The underlying models start with basic representations of standard components, but the semantics and data hierarchy are configured in a way that is unique to the real system.”
Phaidra isn’t the only startup that has tried to unravel the data center energy problem with AI. Another vendor in the space was Boston-based Carbon Relay, at least until it decided to rebranding and change of direction for DevOps and IT.
Elsewhere, Meta and Microsoft have also experimented with AI-based data center optimization. But Gao believes Phaidra’s predominant competition is “the traditional way of doing things.”
“Plants typically hire an outside engineering or consulting firm to analyze plant performance and manually update back-end programming,” Gao said. “The problem with this approach is that traditional hard-coded control logic forces the plant to operate in the same way until someone comes in and updates the back-end programming—which happens every five to 10 years in the industrial sector.”
One of Phaidra’s first customers wasn’t a data center operator but the big pharmaceutical company Merck, which deployed Phaidra’s technology to regulate a 500-acre vaccine plant. But Phaidra’s customers are now largely focused on the data center sector—a trend fueled by the AI craze, Gao says.
As a result, Phaidra has been nominated for this 12 months’s Amazon Sustainability Accelerator, which provides it the opportunity to pilot its technology in Amazon’s European operations with a potential investment of as much as €2 million (~$2.15 million). Is Phaidra targeting Amazon? Gao wouldn’t say — but it will surely be in line with the startup’s long-term growth ambitions.
“Our first international deployments are now live, and we expect higher energy-cost regions of the world to drive a significant portion of our growth in 2025,” Gao said. “Businesses are looking for ways to do more with what they have… We are well-positioned to execute our growth plan over the next two years.”
Phaidra makes most of its money by charging its AI an annual SaaS-like subscription. Gao explained: “The fee is a function of the complexity of the object the AI is managing and the price of energy in the local region.”
Seattle-based Phaidra, which employs about 100 people, recently raised $12 million in a funding round led by Index Ventures. This brings Phaidra’s total raised to $60.5 million, with the latest funds going toward R&D, implementation, customer success and expanded go-to-market efforts, Gao says.
Phaidra expects to complete the 12 months with a squad of 110.
“This was an opportunity that allowed Phaidra to bring Index Ventures onto our board and cap table,” Gao said. “While Phaidra was not actively seeking additional capital, we are particularly excited about Index Venture’s scaling expertise as Phaidra rapidly grows with our industrial clients, particularly in the data center space.”