The fastest way to a data centre in Europe runs through a dying steelworks

From Oldřichovice, where I'm from, you can see the chimneys of the Třinec steelworks, Třinecké železárny, almost all the time. I grew up with them on the horizon, and for a long time they meant one thing to me: steel. When people talk about Třinec today, they talk about a chapter that is closing. Cheap Asian steel, expensive energy, European industry under pressure. A dying steelworks.

I read it differently. I'm from here, but steel was never my field. I've spent my whole life building technology companies. From the years I've spent investing in public markets and reading about industries that seem unrelated at first sight, I've kept one habit: looking around the corner at something familiar and asking what it really is once I peel off the first, obvious layer. When I look at Třinec that way, I don't see a dying steelworks.

I see two things stacked on top of each other. On top, a steel business that is structurally barely profitable, one its own management calls unsustainable in the long run. Underneath, energy and grid infrastructure that in Europe in 2026 is scarcer than steel, and under certain conditions worth more than the whole works put together. The thesis goes like this: the steelworks sits on an energy asset that the market values as steel, and that mispricing is something you can build a whole business case on.

I'll lay it out, from what exactly sits on the site, through what the group is realistically worth today, to a concrete plan for pulling out the core and what to do with it next. This is a thesis, the way I would look at Třinec if I had to underwrite it.

What sits in Třinec: a power plant, not just a steelworks

Inside the site runs Energetika Třinec, a subsidiary of the steelworks. It holds licences to generate, distribute and trade electricity, to produce and distribute heat, and to distribute and trade gas.

Specifically, it supplies electricity, steam, hot and warm water, blast and compressed air, and process water. It runs its own water system: waterworks, a supply line from the Těrlice reservoir, an intake from the Olše river, recirculation of cooling water and two treatment plants. And for fuel it largely burns the gases the site produces as a by-product, much of it already in the coke ovens, where coal is turned into coke, the rest in the blast furnace and the converter. That last part matters, and I'll come back to it.

Now the numbers, because they're what turn this into an investment thesis. According to Energetika Třinec's annual report, the turbogenerators in the site's two heating plants have an installed electrical capacity of about 101.5 megawatts, roughly 39.5 MW in unit E2 and 62 MW in unit E3. That is generation right on site. Even more important is the connection to the outside: three 110 kV substations with a total transmission capacity of about 363 megawatts, connected to the higher-level distribution grid.

In plain terms, the Třinec works can both feed large amounts of electricity into the distribution grid and draw from it, up to 363 MW. For scale, that is what a smaller regional city draws. The substation in Třinec stands, is connected and has worked for decades. It is the most expensive and slowest-built thing you need in Europe today for a large electricity load, and in Třinec it's done.

The scarce thing isn't the hall, it's the connection

To see why this could be a very interesting investment opportunity, you need to understand one thing about data centres for artificial intelligence. Most people picture a hall full of servers. The hardware in that hall is the most expensive item in the whole project, but you can buy it. What you can't buy, at least not quickly and on reasonable terms, is a connection to a large and reliable enough source of electricity. That is the real bottleneck, and so the real asset.

A cluster of the graphics cards that AI models are trained and run on doesn't have peaks and lulls like an ordinary office. It draws power all the time, day and night, in tens to hundreds of megawatts. Nebius, a European operator that announces its plans publicly, runs a data centre in Finland that has been expanded to about 75 megawatts and is building a new campus in Lappeenranta aiming for 310 megawatts by 2027. The company says its investment in that campus exceeds a billion dollars.

For scale, the first phase of a data centre that makes any sense to consider in Třinec would be somewhere between 80 and 120 megawatts, roughly a third of the site's total connection. And unlike a city or a steelworks, whose demand fluctuates, this load runs continuously, day and night. Třinec has exactly that kind of connection already built.

The second story is the American-Canadian IREN, formerly Iris Energy. It started out mining bitcoin and now presents itself to shareholders as a company with a pipeline of about 2.9 gigawatts of connected capacity for AI and high-density data centres. The logic it's communicating is simple: the company's value isn't in the halls or the hardware but in the right and the ability to connect gigawatts of cheap, secure power.

And that is exactly the bottleneck in Europe. Connecting a new large load to the grid, for a project of around a hundred megawatts, routinely takes ten to fifteen years in practice today. Transformers and substations have lead times measured in years, because the whole world wants them at once. On top of that comes pressure on water and cooling, since the most powerful servers today are water-cooled right at the rack, and European regulation pushes for low freshwater use. Anyone looking for a place in Europe in 2026 for hundreds of megawatts of compute isn't worried about land. There's plenty of land. They're waiting for power, a substation and water. In Třinec, all three stand built and connected.

Steel keeps the price down

So let's look at the economics. The steelworks isn't a standalone company, it's the top of a group. At the end of the ownership chain are Slovak owners led by Tomáš Chrenek, who control Moravia Steel through Minerfin and personal holding companies. Under Moravia Steel sits the steelworks, and under that Energetika Třinec plus about twenty other companies: engineering works, wire mills, foundries, screw makers, forges, makers of springs and bearing rings, scrap processing, technical and engineering firms.

Now the numbers that decide the price. According to the consolidated results of the Moravia Steel group, revenue fell from CZK 75 billion in 2022 to just under 63 billion in 2024. Net profit dropped over the same period from 5.6 billion to 543 million, and in 2023 it was only 94 million. Net margin is below one percent and EBITDA fell from ten billion to just under four. Profitability also swings quite a lot from year to year, partly depending on how much the group invests in itself in a given year. But I don't think any investment under the current management, and probably not under the current owners either, would fundamentally turn the numbers around as long as the focus stays on the same industry. The chief executive himself describes the business as unsustainable in the long run in the European steel crisis. About fifteen thousand people work in the group, more than six and a half thousand of them in the steelworks itself.

Translated: it's a giant on revenue that barely makes money. Steel in Europe is under pressure from cheap imports and weak demand, and nobody sensible expects that to turn around structurally. One popular bogeyman should be set aside, though: emission allowances aren't weighing on Třinec today. With free allowances and decarbonisation subsidies, the works is for now more subsidised than penalised, so the real pressure is demand and imports, not the price of carbon. The emissions bill will come, but later, once free allowances are cut off, not now. And this is exactly where the mispricing sits. A few years ago, when steel prices were high, I used to hear sale ambitions around this group in the range of CZK 30 to 50 billion. For a company that today earns under a billion a year, and whose main business its own management calls unsustainable, that number is out of touch with reality. On its books the group has over 40 billion of equity, but capital tied up in steel assets earning under one percent can't be sold at book value.

I ran a quick pass on where the group really stands with my own tool. I take it as a rough estimate, not a valuation: the consolidated value of the whole group comes out somewhere in the range of CZK 10 to 15 billion. The exact number isn't what matters. What matters is that steel sets it, and that the most valuable thing in the group doesn't show up through the steel lens at all. You aren't buying a steelworks' profit. You're buying the energy and the connection underneath it, and the steel is just the reason the price is low.

Keep vs. sell

Looked at as an investor rather than a steelmaker, the group splits into two piles: the core you want to hold, and the rest, which can be sold.

The core is three things. Energetika Třinec, because that's what the whole case is about. The steelworks itself, which has to keep running for now: today it's the energy company's main customer and their networks are intertwined, so it can't be disconnected overnight. The whole next section is about that. And M Steel Projects, the group's project and engineering company, which works well as a vehicle for running the transformation of the site.

The rest is non-core, and there's surprisingly much of it. Strojírny a stavby Třinec, Slévárny Třinec, ŽDB Drátovna, Šroubárna Kyjov, Kovárna VIVA, Moravskoslezský kovošrot, the makers of springs and bearing rings. And then things well beyond what you'd expect at a steelworks: the Barrandov film studios, a golf course. These aren't bad assets. Many of the engineering companies have real customers in automotive, rail and energy and can be sold on their own as exporting brands, some as joint ventures with global players. They just have nothing to do with why a data centre operator would want Třinec. Exactly where the line runs between what hangs under Moravia Steel and what the owners hold alongside it through their holding is a question for due diligence. The principle is the same either way: it's a package for sale, not for holding.

That gives the whole logic of the deal. Buy the group for somewhere around 15 billion. Sell the non-core subsidiaries, by my rough estimate from the same screening for somewhere between 7 and 10 billion back. Keep the energy core plus the steelworks itself until a data centre takes over its place as the main customer. After the disposals, the net price for the energy backbone alone comes out at roughly CZK 5 to 8 billion. For 363 MW of grid connection already built and over a hundred megawatts of own generation, at a time when a new connection elsewhere takes ten to fifteen years, that may be cheap. Above all, it's years ahead.

The numbers in this section are mine, not verified from the books. The exact perimeter of companies, their stakes and real sale prices would have to go through proper due diligence. But the range fits what is publicly visible about the group, and the order of magnitude of the argument doesn't change if a decision moves it a few hundred million up or down.

Fade out the steel, fade in the data centre

This is the moment where most ideas like this die, because they get done in the wrong order. Energetika Třinec and the steelworks are a connected system today. The energy company supplies the works with electricity, steam, heat and compressed air, and the works is by far its largest customer. As a supplementary fuel it also uses the fuel gases produced as a by-product of steelmaking. The main fuel of the heating plants, however, has been coal until now, which gas is only now replacing. If you wind down steelmaking, the energy company mainly loses its largest customer and part of the operating logic it stands on. That's why the steelworks isn't switched off as the first step. It's switched off only once a new source and a new large customer that takes its place are running next to it.

The order that makes sense looks like this. First, sell the non-core, which returns capital to the deal and simplifies the group down to what matters. In parallel, map everything on the site that is physically tied to the steelworks and what can run on its own: the gas, steam, water and power networks. This is boring but decisive work, because it determines what can be separated and what can't.

Then comes the fuel switch. In practice that means building independent sources next to the existing energy plant that can feed the new load without the steelworks: more draw from the grid through the existing 110 kV connections, own gas turbines built in smaller blocks, say ten to twenty megawatts each, and batteries for short outages and peaks. Only once this new setup is built and tested for continuous operation can steelmaking start to wind down. Not before.

It's worth noting that the first step in this direction is already being built in Třinec. Energetika Třinec itself is investing over four billion crowns in a new combined-cycle plant with 62 MW of electrical output, gas turbines with a steam boiler, with ČEZ ESCO as partner and a turbine from Siemens Energy, due to start in the second half of 2030. It proves what is key to the whole thesis: the energy company in Třinec can build a new, independent source next to itself, with a reputable energy partner and in the range of a few billion crowns. The fuel switch here is already a construction site.

One important detail makes the whole thing real: the first phase of the data centre doesn't have to wait for the furnace to go cold. Eighty to a hundred and twenty megawatts could be built on spare capacity, on what is left of the 363 MW after subtracting the steelworks' current draw and the reserve the substation has to keep so that it can carry the load on the remaining transformers if one of them fails. For an operation where a power failure means irreversibly ruined lines, you can't run flat out at the maximum. How much is really free will only come out of a grid study, and it's more likely part of that difference than all of it. The data centre can start ramping up alongside the steel winding down. The fade-out of the steelworks and the fade-in of the data centre run against each other like two sides of the same balance sheet: as steel output falls, and with it the draw and the fuel flows, compute load grows and takes over that capacity. The pace isn't set by ambition but by how fast the replacement sources and the new load cover what the steelworks disconnects.

What has to be true for this not to be speculation

This isn't an announced project or a deal proposal. The owners may not want to sell, and without a connection agreement and an anchor tenant it's still just a thought, not an investment project.

For the thesis to become a project, a few things have to fall into place. I've ordered them by what is most likely to kill the case:

1. Grid connection. Without a binding connection agreement for a new type and volume of load, 363 MW is only capacity on paper. This is where it's decided. The existing substation gives a head start, but it doesn't replace a new agreement, grid studies and most likely reinforcement of the nodes.
2. Anchor customer. Without an AI compute operator signing up to long-term offtake, nobody pays for the fuel switch and the new sources. A data centre without a contractually secured tenant is just an expensive hall.
3. Cost and time of the fuel switch and the steel wind-down. Building replacement sources and safely winding down steelmaking is capex and years of work, not an accounting operation. Underestimate it and you get a cold site.
4. Politics and fifteen thousand people. Winding down the largest Czech steel producer is politically explosive. Who carries the social cost, the retraining and the decarbonisation capex is a question anyone going in has to have an answer to. Without one, the whole idea is academic.
5. Willingness and price. The Slovak owners may not want to sell at all, or only at a number that shifts the whole calculation. The thesis stands or falls on the energy core being cheap precisely because it hangs on loss-making steel.

The first two yeses, grid and tenant, are what tips the whole story. Until they come, the restructuring of Třinec is just an interesting thought. Once they do, it's ready-made AI infrastructure for a fraction of the cost of a greenfield.

It seems a shame to me that the country's largest steelworks has been talked about as a problem for ten years and nobody has asked what it actually stands on. Literally.