Alexey Gubarev, Servers.com Co-Founder, on Energy as AI's Hidden Bottleneck

Alexey Gubarev has spent two decades building things most people never think about until they stop working: server infrastructure, grid capacity, national tech ecosystems. As Co-founder of Servers.com and Co-founder and Board Member of both Palta and TechIsland, he has learned to spot the bottlenecks nobody is watching yet, long before they become the crisis everyone is suddenly talking about. In this interview, he talks about what Palta's Flo and Simple apps are solving for at scale, why energy and grid connections are the constraint AI leaders are underestimating, what Cyprus still needs to build its own startups rather than just host relocated ones, and why the founders who endure are the ones who can be told they're wrong.

Alexey Gubarev, Co-founder, Servers.com, Palta, and TechIsland

What are you most focused on building or backing right now, and why does it matter to you?

Most of my attention goes to Palta's portfolio. Flo is the largest women's health app in the world: over 420 million installs and more than 80 million monthly users. Simple coaches people through weight management with AI, and Zing Coach does the same for fitness; both exist to give someone who could never afford a personal trainer or a nutritionist something close to one. Lovi, an AI skincare assistant, is the newest bet. Since we founded Palta in 2016, the teams have grown across the UK, the US, Lithuania, Cyprus and the rest of Europe. At that scale, a decision about data, governance or growth lands on tens of millions of people, not on a balance sheet.

Alongside that, I remain closely involved in TechIsland and its work on Cyprus as a technology hub, and I am building a retro car museum in Bishkek. On paper, these have nothing in common. In practice, they share one instinct: build things that outlast the person who started them. That is what interests me at this stage. Not the next exit, but whether what I built keeps working once I have stopped looking at it every day.

What did building Servers.com teach you about spotting infrastructure bottlenecks before they become obvious?

You stop trusting the numbers everyone else is watching and start watching the ones nobody talks about yet.

The first is utilisation drifting upward on something that should be flat. Not a spike; spikes get noticed. A slow drift does not. The second is lead times: if the wait for a new connection, a new rack, or a new permit gets quietly longer year after year, the bottleneck is announcing itself long before it blocks anything. The third is what your own suppliers complain about in private, because they see the constraint tightening before their customers do.

By the time a bottleneck is obvious to everyone, the window to solve it cheaply has closed. The whole skill is treating small, boring trend lines as a warning system instead of waiting for the crisis they are building toward.

When you evaluate founders at Palta, what tells you a team can adapt even if the original product changes?

The product is not the first thing I evaluate. The market underneath it is. If the real opportunity is not worth hundreds of millions of euros, potentially billions, no amount of adaptability saves the business, because there is nowhere for a pivot to land. Timing tells you almost as much. Launching too early is as fatal as launching after the market has moved on; I have watched strong products fail because nobody was ready for them, and that is a harder mistake to recover from than most founders assume.

Within that, two things separate the founders who can adapt. Do they manage money as if it were their last euro? And can they be told they are wrong without getting defensive? Teams that argue their way out of bad news usually cannot pivot when the market forces them to, because they have spent their energy protecting the original idea instead of the company. The ones who adapt well separate their identity from their first idea early, and you can see it long before the product needs to change.

Which physical constraint in AI infrastructure do you think technology leaders are still underestimating?

Energy, and above all the time it takes to get a grid connection. Everyone is focused on chips and models. Meanwhile, data centres consume around 415 terawatt-hours a year globally, heading for roughly 945 TWh by 2030, growing about four times faster than electricity demand from everything else combined.

The aggregate number is not the problem. The problem is that the demand lands in a handful of places at once, where connection queues now run for years rather than months. In Amsterdam, one of Europe's most important digital hubs, fully permitted campuses sit at a fraction of their capacity because the new substations will not arrive until the 2030s. Leaders who treat this as a procurement problem, solvable with a bigger cheque, are going to be surprised. The companies, and the countries, that win the next decade are the ones that treated storage, permitting, and grid capacity as strategy years before they became a public bottleneck.

Cyprus has become successful at attracting tech companies, so what must change for it to consistently create them locally?

We built one layer very well and left a gap in the middle of it.

Cyprus has an EU jurisdiction, a workable headquarters regime and English as a working language. That combination worked: the technology sector now contributes €5.9 billion directly to GDP, 16.2 per cent of the total. But almost all of it is relocated operations, companies founded elsewhere that found it made sense to move here. Very little of it was started here.

What is missing is startup infrastructure in the literal sense: early capital that can say yes in weeks, founder-friendly legal instruments, a stock-option regime that does not punish employees. The Research and Innovation Foundation has mobilised €207 million in five years, but it pulled in about 23 cents of private money for every public euro. A healthy system runs at a euro or more. Grants alone produce grant-writers. What works is co-investment, private capital leading and public money matching. We have already proven that model on the island. It needs to be scaled ten times over, not run as a pilot.

When a business hits a growth ceiling, how do you identify the real constraint holding it back?

I run one test: if you doubled the budget tomorrow, would the ceiling move, or would you just be spending faster against the same wall?

Most people assume the answer is more money or more marketing. It rarely is. Sometimes the limit is physical; a data centre's growth eventually depends on power availability, not on ambition. Sometimes it is the team. I have made large investments where the product and the pipeline were strong, and the team running them was weak, and no amount of capital fixed that. If the wall does not move no matter what you throw at it, you are not looking at a growth problem. You are looking at a structural one, and that calls for a different kind of decision than raising another round.

What can founders do early to avoid becoming dependent on infrastructure or policy conditions they cannot control?

Build optionality into the business before you need it, not after.

At Servers.com that meant operating across several regions and jurisdictions from early on: data centres in different countries, on different grids, under different regulators, so that no single outage, regulatory change or local market failure could take down the whole company. That lesson gets very expensive when you learn it after the fact.

The founders who get hurt are usually the ones who quietly let one input become load-bearing. One power grid, one regulatory environment, one piece of physical infrastructure they do not own, and nobody ever stress-tested what happens if it changes. You do not need to predict what will go wrong. You need to make sure that no single dependency, infrastructure, or policy can take the whole thing down with it.

"Build, back and give back" runs through your business and community work, so how do you decide where your involvement can create lasting value?

There is no single formula. Sometimes it is personal taste, sometimes it is a gap that nobody else is filling, and sometimes it is simply responding when something happens.

I have always liked parks, and I thought the island did not have enough of them, so when the chance came, I helped fund one in Parekklisia. Cyprus had no professional-level skatepark despite weather that allows one to be used all year, so we built City Skate Park in Limassol and gave it to the city. Other times it is about showing up. After the 2025 wildfires, TechIsland put out a call, and within a few days more than 60 companies and individuals had responded with close to €2 million for the affected communities. And some things start small and grow on their own: the City Friends Club idea was created during COVID, when my wife, my kids, and I were picking up rubbish around our own neighbourhood. My wife decided to create such a project and run it as founder. It is now one of the largest environmental organisations on the island.

If there is one test, it is this: once you start, you follow through. Most initiatives fail not because the idea was wrong but because the person behind it stopped paying attention.

After more than 20 years in Cyprus and several very different ventures, what lesson about building enduring systems would you most want other entrepreneurs to carry forward?

Hire people who are better than you in their field, and surround yourself with strong, reliable people rather than people you feel comfortable managing.

Most founders instinctively hire for comfort: people who agree, who do not push back. That quietly caps how far the company can grow. An enduring system needs people who can run parts of the business better than you could yourself, and it takes some courage to bring them in.

The other half is speed. The world is not waiting for anyone to feel ready, and inaction costs more than a wrong decision. A wrong decision can be corrected. A delayed one keeps compounding.

Follow Alexey Gubarev on social media and explore his website.