Due to AI, electricity demand is booming.
In May 2024 reportGoldman Sachs predicted that data centers will eat 8% of the U.S.'s total electricity supply by 2030 (up from 3% in 2022) as cloud service providers expand to satisfy demand for AI infrastructure. If current trends proceed, U.S. utilities will need to take a position around $50 billion in power generation capability to support all the upgraded – and recent – AI-powered data centers.
There might be serious negative externalities. In Kansas, where Meta recently broke ground on a huge recent server complexEnergy supplier Evergy announced that the shutdown of the coal-fired power plant can be delayed by as much as five years. Some experts say that power-hungry data centers – which also large water guzzlers – could contribute to rising utility costs for consumers and disproportionately affect low-income people.
The problem of knowledge center power consumption seems insurmountable. But Jim Gao, Katie Hoffman and Vedavyas Panneershelvam, co-founders of Phaidra, consider it is feasible to retrofit existing facilities to make them more energy efficient.
In fact, they’ve made a business out of it.
Founded in 2019, Phaidra develops AI-powered control systems for data centers, pharmaceutical and business constructing infrastructure. The company's systems collect data from 1000’s of sensors in a facility and make real-time decisions about the right way to cool the equipment inside in an energy-efficient manner.
For many data centers, cooling is some of the energy-intensive components. The cooling system of a median data center consumed about 40% of the middle’s total output.
“The data center industry is within the midst of an arms race to create recent capability wherever land and power can be found,” Gao said in an interview with TechCrunch. “Phaidra's service can provide a more stable cooling system that consumes less energy.”
Gao previously led DeepMind Energy, the team inside Google's DeepMind AI research division answerable for commercializing technologies to handle the challenges related to climate change. During his time at DeepMind, Goa – along with Panneershelvam, then a research engineer at DeepMind – developed an AI system to regulate and optimize the energy consumption of Google's data centers. The topic attracted quite a little bit of attention on the time.
DeepMind made the choice to quietly wind down DeepMind Energy after failing to secure contracts with major industry players reminiscent of UK energy provider National Grid. CNBC's reportingGao left the corporate in August 2019 and Panneershelvam in May 2020 – a couple of months after the departure of DeepMind co-founder Mustafa Suleyman, who was reportedly a driving force behind DeepMind's climate change efforts.
After leaving DeepMind, Gao and Panneershelvam saw a chance to use their experience from the Google data center project to other data centers — and beyond. They recruited Hoffman, who led innovation projects at Trane, a refrigeration systems manufacturer, to start out Phaidra.
For each client, Phaidra develops AI models which can be trained on sensor data to optimize the cooling systems and overall energy management of a facility (reminiscent of a knowledge center). These models are self-improving, Gao claims, consistently learning from their very own experiences managing the ability's infrastructure.
“One of Phaidra's unique approaches to AI is that we mix physical knowledge of how the plant works with learned models of plant dynamics based on sensor data,” said Gao. “The underlying models start with basic representations of ordinary components, however the semantics and hierarchy of the info are individually configured by the actual system.”
Phaidra shouldn’t be the one startup attempting to tackle the challenge of knowledge center energy consumption with AI. Another player on this space was Boston-based Carbon Relay, at the very least until it decided to Rebranding and realignment about DevOps and IT.
Meta and Microsoft are also experimenting with AI-supported data center optimization. But Gao sees Phaidra's principal competition in “the normal way of doing things.”
“Typically, plants hire an out of doors engineering firm or consulting company to research the plant's performance and manually update the backend control programs,” said Gao. “The problem with this approach is that the normal hard-coded control logic forces the plant to operate in the identical way eternally until someone updates the backend programming – which in the economic sector happens every 5 to 10 years.”
One of Phaidra's first customers was not a knowledge center operator, but pharmaceutical company Merck, which used Phaidra's technology to administer a 200-hectare vaccine factory. Today, nevertheless, Phaidra's customer base is heavily focused on the info center sector – a trend that Gao says is being reinforced by the AI hype.
In this context, Phaidra has been named as a finalist on this 12 months's Amazon Sustainability Accelerator, giving the corporate the chance to compete for the prospect to check its technology in Amazon's European offices with a possible investment of as much as 2 million euros (about $2.15 million). Is Phaidra trying to collaborate with Amazon? Gao wouldn't say – however it will surely be in keeping with the startup's long-term growth ambitions.
“We have already launched our first international deployments and expect the upper energy cost regions of the world to account for much of our growth in 2025,” Gao said. “Companies are on the lookout for ways to do more with what they’ve… We are well positioned to execute on our growth plan over the following two years.”
Phaidra makes its money primarily through a SaaS-like annual subscription for its AI. Gao explained: “The fee depends upon the complexity of the plant the AI manages and the worth of energy within the region.”
Seattle-based Phaidra, which employs about 100 people, recently raised $12 million in a funding round led by Index Ventures, bringing Phaidra's total raised to $60.5 million. The recent money will likely be put toward research and development, implementation, customer success and expanded go-to-market efforts, Gao says.
He expects Phaidra to complete the 12 months with a team of 110.
“This was an opportunistic capital raise that allowed Phaidra so as to add Index Ventures to our board and capitalization table,” said Gao. “While Phaidra was not actively looking for additional capital, we’re particularly enthusiastic about Index Venture's scaling capabilities as Phaidra rapidly expands with our industrial clients, particularly in the info center industry.”
