Data centres are the hot area du jour, as the AI boom is pushing the need for more computing power up and to the right. And although investors are no stranger to funding the big buildings and servers — private equity, for example, has long backed such infrastructure — it seems that recently more young data centre companies are pitching themselves to VCs.
In the last four months, Jack Wang, a London-based principal at German VC firm Project A, says he’s received six or seven pitches for startups building companies around physical data centres. “To my surprise, we’ve seen a lot of people pitching full stack data centres — like actual buildings and servers and things like that,” Wang tells Sifted.
It’s also a trend Namratha Kothapalli, a London-based principal at Speedinvest, has noticed in the last year. She estimates she’s seen 50 startups pitching data centres, largely from hotspots like France and Germany. It’s even prompted companies that were once more focused on chip design to brand themselves as data centre startups. They’ve “changed their positioning to ride the wave of the new AI and data centre hype cycle as well,” she says.
But despite the buzz, some VCs are dubious about taking a chance, considering the cost of building data centres and the energy and water needed to run them. Some are also concerned about scalability and competition. And it begs the question: are physical data centres really a VC investment case, or are they better suited for the likes of PE firms and banks to fund?
Chasing the AI boom
The popularity of data centres as a startup trend seems to be growing; data centres made Y Combinator’s list of requests for startups this spring, for example.
One reason for the interest is that compute is hard to get these days, says Wang, and “people are legitimately looking for alternative compute places at reasonable prices.” He adds that multiples of public data centre companies increasing alongside the demand for AI compute has put them more in-line with VC metrics (notwithstanding a recent selloff related to Chinese AI model DeepSeek).
Numerous data centre startups have been bringing in loads of money, including UK-based AI infrastructure startup Ori, which partners with data centres to rent cloud-based access to high performance computing (HPC); it raised $140m this month. UK-based Nscale, which is building data centres in countries including Norway, raised $155m in late 2024. Meanwhile Swedish startup Evroc announced plans last week to build a hyperscale AI data centre in France — and has raised $60m in total, according to PitchBook, from investors including EQT Ventures and Norrsken VC. In the US, larger companies like CoreWeave, an AI compute company which recently opened its first data centres in the UK, have also raised large sums; CoreWeave was reportedly valued at $19bn in a funding round last year. Even French OpenAI challenger Mistral recently announced it’s building its own data centre in the country.
A number of the startups raising right now are pitching data centres with a software component — or middleware, which can help with things like scheduling where in the data centre to run a specific job or optimising downtime. Some are focusing on making data centres more “green” and sustainable.
Wang says he’s seeing companies based out of the Nordics, where energy is cheaper and more renewable — a key part of the pitch since cooling costs are a large portion of a data centre’s expenses. Many of these startups are focusing more on HPC — things like aerospace computation or fluid dynamics computation which aren’t time-sensitive and don’t require real-time data, says Wang.
Most of the founders he is seeing now are in their mid-20s and have worked at a data or tech company. “The type of founder profiles we see, ironically, [are] not your infrastructure kind of guy — it’s actually more of your techies, and I think they just naturally gravitate towards VC.” Some companies are looking to raise VC funding for hiring, but are also trying to secure debt for more of the capital expenditures, says Wang. “They have a debt product as well, but you’ve got to give them equity to unlock the debt.”
Why it might not be a VC case
Several VCs are sceptical that straight data centre startups are the right fit for their cheques.
“Every PE investor I talked to is like, ‘Hey, don’t do this. This is not a VC thing.’ I’m like, ‘I’m inclined to agree with you, but why are people looking at this and why are founders building here?’” Wang says.
That question — whether it’s an area better suited for real estate or infrastructure investors — is one of the reasons why Kothapalli says Speedinvest hasn’t made an investment yet. “Time to market is another one, where it takes a very long time to prove any sort of proof points because the tech and the engineering side of it is still fairly risky at that point,” she says. “In some sense, you’re also competing with the hyperscalers of the world.”
Some investors point to startups like US-based Crusoe Energy, which is building data centres and reportedly working with OpenAI, as a success case. Starting that same business in 2025 would put it more in “PE territory” unless a startup was to say they’ve built a new kind of data centre from the ground-up, “but we haven’t seen that,” says Alexandre Flamant, a partner at Paris-based deeptech investor HCVC. The majority of interesting use cases he’s seeing are in setting the data centres up: think construction and energy.
The argument that some founders are now making is that the middleware they’re developing, combined with the theoretically lower energy costs — based on the location or optimisation of the data centre — would translate into a competitive business, says Wang. Founders are saying that “because of their software and energy capacity, they can drive higher margins into that VC case realm”.
But “everybody” is building this middleware, says Wang, and he doubts this is actually a unique selling proposition.
Another potential problem: scalability. Startups building in low-energy-cost regions like the Nordics may have trouble growing outside the region if they want to keep those costs low — and that might also mean that they can only do high performance computing instead of critical or real-time computing, Wang says.
What VCs are liking instead
Although some of the VCs Sifted spoke with aren’t too sure about full stack data centres, they are looking into or have invested in startups addressing markets related to data centres, like the building or cooling process.
Project A has been looking into startups working on tech to lower the temperature of the servers, which get incredibly hot. Wang says Project A is “really bullish” on this area and is looking for “a few” European startups similar to US-based Phaidra, which uses AI to optimise things like downtime and energy intensity.
Wang says he’s more interested in those use cases as opposed to pitching a brick-and-mortar data centre in cost-efficient location but where scalability will be a problem since it would be hard to move it elsewhere. He also thinks that the shortage of Nvidia chips will be a temporary problem.
Kothapalli says Speedinvest is currently looking “very seriously” at four companies, including one in optimisation (around faster and more efficient data processing and energy usage) and one focused on cooling that includes a hardware element.
Flamant says he’s seen some startups focused on data centres which are working on battery tech to store energy during the day and release it at night. He’s also looked into the electrical part of construction, adding that there’s been a shift toward pre-fabricated construction — building things in a factory and assembling them. “We’ve seen that for data centres on the electrical side,” he says.
But there could be even more startups popping up addressing data centre problems, particularly as Europe aims to have more sovereignty around compute and AI. France recently announced €109bn to invest more in AI and data centres. “If we continue on the path that we’re [on], it seems logical that European governments and companies are going to want to do more compute at home on European-owned infrastructure. If that’s the case, then that should create an equivalent boom that we’re seeing in the US,” says Flamant.
Kothapalli believes “we’re just getting started; there’s going to be more to come. I think by next year, this is probably going to be like, ‘Oh my god, this is so crowded.’”
Still, you might not see VCs announcing loads of data centre startup deals anytime soon: Wang says they currently have more questions than answers.
Read the orginal article: https://sifted.eu/articles/data-centre-startups-vc-debate/
