The data center industry is expanding rapidly to keep up with the flywheel growth of AI. While these data centers are necessary AI infrastructure, they store an AI company’s compute, they are expensive to build, seemingly more so to run, and they are a huge energy suck. Startups are looking to make data centers more efficient and sustainable, but it isn’t that simple.
The global data center market is estimated to be worth $301 billion, according to P&S Intelligence, and predicted to more than double into a $622.4 billion market by 2030. Data centers consume about 4% of the total power in the U.S. today, according to the Electric Power Research Institute, and it’s predicted to more than double to 9% by 2030.
Data centers, and the big companies that rely on them, are scrambling for power. Last month Microsoft inked a deal with Constellation Energy to restart its nuclear reactor on Three Mile Island to keep up with demand.
In tandem with the increase of data centers is the growth of the number of startups looking to solve the data center industry’s energy crisis and environmental impact. Startups like Incooling and Submer are looking to tackle the space by cooling down existing data center technology so they produce less heat. Others like Phaidra are using software to help data centers more efficiently manage their cooling.
Some are looking to build an entirely new model. Verrus is building a more “flexible” data center using microgrids. Sage Geosystems is building a way to use hot pressurized water to power data centers instead of natural gas.
Sophie Bakalar, a partner at Collab Fund, which is an investor in Phaidra, told TechCrunch that while there were entrepreneurs looking to build data center tech before the AI boom — data centers play a large role in cloud computing and bitcoin mining as well — she’s noticed a 10x increase in founders looking to build tech for this space over the last year.
“We’ve seen a company that is building data centers in space, it runs the whole gambit,” Bakalar said. “Whenever you have such an obvious problem in supply and demand, it’s natural you will see a lot of entrepreneurs eager to tackle the issue from different angles.”
But although data centers are expanding quickly and will need solutions to be more efficient, that doesn’t mean startups should think it will be easy to get their tech adopted.
Data center challenges
Francis O’Sullivan, a managing director at S2G Ventures, told TechCrunch that the speed with which this space is growing may actually make it harder for startups to find partners willing to test their tech or take a chance on it.
“[Data centers] are enormously expensive assets, multi-billion-dollar facilities. The reality there is they must work,” O’Sullivan said. “Therefore the real meaty data center world is not a forum for experimentation.”
The customer base for this kind of tech is also arguably more concentrated, and with that, likely harder to penetrate, said Kristian Branaes, a partner at climate-focused VC Transition. Branaes added that his firm has spent a lot of time researching and going deep into the data center tech category, but while they’ve found cool companies building novel tech, they haven’t been able to gain enough conviction to invest.
Branaes is worried about how companies will be able to scale. He thinks some of the startups he’s found fall under the classic climate tech conundrum of being cool tech but not necessarily a company that can produce venture-like returns. He said that it’s hard to build a venture-scale company that only sells into a handful of large companies like Microsoft and Apple.
“We have come to [the] view: It is very, very hard to build a large company only selling to AWS and Microsoft and whatever; they are ruthless at procurement,” Branaes said. “They are not in the business of giving away a lot of margins. If you start to make too much money, they want to circumvent that or start doing it internally.”
Powering on
While some investors remain skeptical, many startups in this space are seeing traction. Impending regulations in both Europe and in data center-heavy U.S. states like Virginia mean that even if these large customers aren’t shopping for solutions now, they will likely have to in the future.
Helena Samodurova, the co-founder of Incooling, a Netherlands-based startup looking to cool data centers down, launched her company six years ago, before the current AI hype. While data centers, and the energy they used, was an issue then, the demand for Incooling’s tech has completely changed.
“Back in the day, people didn’t really know about it,” Samodurova said. “In the last six years, that has changed tremendously. As we went through this journey, we really had to educate people on what this was. Fast forward six years later, that’s not the case. We are being sought out.”
Samoduorva said interest has increased from both potential customers and investors, too. She added that the data center industry is more broad than just the Amazons and Googles of the world and that helping improve data center emissions is not just focused on those few large companies.
“You have a bus to go to the station, you have a car to take your family to go out, you have a Ferrari to go racing, everything has four wheels but the mechanics of it is different,” Samodurova said. “We provide cooling solutions or computing solutions to fix whatever bottleneck you are.”
O’Sullivan said that for him, while a lot of data center tech is a bit too nascent to get excited about at the moment, there are other categories of companies to back that help solve some of the same issues data center tech is aiming for. One is: solving the issues involved with getting the actual energy to the data center and making sure that power grids can handle that level of power.
For startups focused on data centers, adoption might just be too early for some of the category’s earliest entrants. Unlike Incooling, many companies have just been founded in the past few years. While the data center tech market may be in its earliest innings, AI, and the data centers needed to power the industry, aren’t going away anytime soon.
“I think the main point to consider is there is a real urgency here,” Bakalar said. “The growth is really outpacing the current infrastructure that we have. We need newer, better, faster ways to achieve the promise we have heard about AI.”