He Built Russia's Google, Then Walked Away — Now Nvidia Is Backing His Next Move

One evening I fell into a rabbit hole. I had been reading about quantum computing — the technology everyone says will change the world. I spent days on it. And I came away with one honest conclusion: it is real, but it is far too early. Nobody can tell you which company wins, or when. As an investor, that is a beautiful idea you cannot actually act on yet.But while reading, I came across an article that used two words I had never paid attention to: "sovereign cloud" and "neocloud." The writer argued that the real money in Artificial Intelligence today is not in some far-off future — it is in the boring pipes and engines that run AI right now.That one line pulled me in. So before I tell you which company I found, let me first explain — in the simplest words I can — what a neocloud actually is. Because once you understand it, the rest of the story lands much harder.

First, what is a "neocloud"?

You already know the word "cloud." Companies like Amazon, Google and Microsoft rent out computers over the internet. You don't buy the machine — you just pay to use it. Simple.A neocloud is a newer, narrower version of that idea. It rents out only one thing: the special, very expensive computers needed to run Artificial Intelligence. These are racks of Nvidia chips called GPUs — the engines that train and run AI models.Think of the big clouds as a giant shopping mall selling a hundred different things. A neocloud is the single shop that sells only one product: raw AI horsepower, by the hour.And "sovereign cloud"? Same engine, but built so a country or a big company can keep its AI and its data on its own soil, under its own rules — not on some foreign server far away. As every nation wakes up to AI, this demand is exploding.So picture this kind of company as a landlord of the most wanted real estate on earth right now — buildings full of AI chips. Tenants sign rent agreements for years, sometimes even before the building is finished.

So which company is this?

The company I found is one most people in India have never heard of. Yet the biggest names in technology have quietly placed enormous bets on it. Meta has signed a deal with it worth around 27 billion dollars. Microsoft has committed up to 19 billion dollars. And Nvidia — the most valuable chip company in the world — has put about 2 billion dollars of its own money in.

Three of the most powerful companies on the planet, all backing one name you probably do not know.The company is Nebius. And its founder is the real story.

The founder who built Russia's Google

To understand this company, you must first understand the man. Because you can copy a business model. You cannot copy a founder like this one.

His name is Arkady Volozh. Years ago he built Yandex — the search engine so dominant in his country that the world called it "the Google of Russia." He did not stumble into it. He earned a degree in applied mathematics — the exact training you need to build search and AI from scratch. Most articles mention his education in half a line. For me, it is the whole point. The engine of everything he built sits in that mathematics.Then came 2022. When the war in Ukraine began, Volozh did something rare for a man in his position. He publicly called the invasion wrong. He was sanctioned. He stepped away from the Russian business. He walked away from the empire he had spent decades building.But he did not walk away empty-handed, and he did not walk away alone. He helped move a core group of his engineers out of Russia to safety, kept the international pieces of the business, and rebuilt them in Amsterdam under a new name: Nebius.So understand what you are really looking at. This is not a fragile startup learning on the job. It is a battle-tested engineering team, led by a proven founder, given a clean second life and pointed straight at the biggest opportunity of our time.

One company, many businesses

Here is what surprised me most. Nebius is not one business. The founder runs it like a small group of companies under one roof — exactly the way he built things at Yandex. Let me explain each one in simple terms.

1. The AI cloud — the main engine

This is the landlord business: renting out AI computers by the hour and by the contract. It is the part doing those billions in deals with Meta and Microsoft. This is the cash engine of the group.

2. Toloka — the teacher

AI models cannot learn on their own. They need huge amounts of clean, human-checked data, the way a student needs good textbooks. Toloka prepares exactly that. It has become important enough that even Amazon's Jeff Bezos has invested in it.

3. Avride — the self-driving arm

Avride builds self-driving cars and small delivery robots — picture a little robot quietly bringing food to your door. It already works with big names like Uber and Hyundai. A completely different business, riding on the same AI brains.

4. ClickHouse — the speed machine

ClickHouse is a super-fast database — software that lets companies search through mountains of data in seconds. Nebius owns roughly a 28% stake in it, and that stake alone is valued at around 6 billion dollars. That is real money sitting quietly on the books.

5. TripleTen — the school

TripleTen is an online school that trains ordinary people for tech careers through bootcamps. Smaller than the rest, but it quietly feeds skilled talent back into the whole group.

The spin-off secret I learned from Joel Greenblatt

Now here is the idea that made all of this click for me. Years ago I read a book by Joel Greenblatt called "You Can Be a Stock Market Genius." One of its most powerful lessons was about spin-offs.A spin-off is when a big company takes one of its smaller businesses and lets it stand on its own as a separate company. Greenblatt's point was simple: the market often values the parent company without giving any real credit to the hidden business inside it. When that hidden business is finally separated and stands alone, its true value suddenly becomes visible — and early shareholders are rewarded.That lesson is exactly what I see here. Nebius has already spun off Toloka and ClickHouse into separate companies while keeping big stakes in them. Avride looks like it could be next. You are quietly owning several businesses for the price of one. The market is mostly paying for the cloud engine — and treating the rest as a free bonus.

Buying brains: the smart acquisitions

The founder is not only building businesses inside the company. He is also buying small, brilliant teams from outside to climb higher up the value ladder.

The goal is a platform called Token Factory. Instead of just renting raw machines, Token Factory lets customers run finished AI models on Nebius and pay for each piece of output. Less like renting a flat, more like owning a toll booth that earns a little every time someone passes through.

To build it faster, Nebius made two clever purchases. In February 2026 it bought Tavily, an Israel-based startup, for about 275 million dollars — it lets AI tools search the live internet instead of relying only on old training data. Then it spent about 643 million dollars on Eigen AI, a tiny but elite team from MIT in America that makes AI models run faster and cheaper on the same chips.

Two deals, one clear message: this company does not want to stay just an electricity supplier. It wants to sell the finished product too.

The honest part: the risk

I will not sell you a fairy tale. This is a risky company, and the risk is mostly about money.

Building these AI data centres costs a staggering amount — the company plans to spend roughly 20 to 25 billion dollars. A business earning a few hundred million spending tens of billions: that gap has to be filled by raising money. Raise it the wrong way, and a company can sink.

But two things give me comfort. First, Nvidia itself put in about 2 billion dollars. Nvidia makes the very chips Nebius depends on. When your most important supplier becomes your investor, that is a serious vote of confidence.

Second, the borrowing is cheap. Nebius raised about 4.34 billion dollars through convertible notes — a kind of loan that can later turn into shares if the company does well, so the lender accepts a very low interest rate today. And the rates here are genuinely low: just 1.25% and 2.625%. For a company spending tens of billions, borrowing this cheaply is a real advantage. Risky business, but sensibly financed.

The honest risks remain: heavy losses today, and the whole story depends on these giant data centres getting built on time. A long delay would hurt.

The moral of the story

This is the part that hit closest to home for me as an Indian.

For decades, our IT giants built their fortunes by selling human hours — armies of engineers billed by the hour to foreign clients. It worked beautifully. But it is, at heart, a manpower business.

Look at what this one founder is doing instead. He builds many small businesses inside one company. He grows them. He spins them off when they are ready. He buys brilliant teams to move up the chain. He creates new value, again and again, instead of just renting out people's time. That is the lesson our IT companies should not miss. The future does not belong to those who only sell hours. It belongs to those who build businesses.

And the deeper moral, for me as an investor, is this. A balance sheet shows you the size of a business. It does not show you the founder who walked away from an empire, the hidden businesses waiting to be spun off, the climb up the value ladder, or the cheap, clever financing holding it all together. To truly understand a company, you have to look beyond the balance sheet. That, more than any stock, was my real takeaway.

A clear disclosure: this is not a recommendation to buy or sell any stock, and I do hold a position in this company, so I am naturally biased. I wrote this purely for learning — to show how business models evolve and why we must look beyond the balance sheet. Please treat it as a learning exercise only, and always do your own research before making any investment decision.