The Tuesday Special No. 5 — Tuesday, September 22, 2026

Good morning, everyone. The machinery has a precise birth date. December 2018, when the Commission approved the first project of this type, dedicated to microelectronics. The design was simple then and it stays simple today. Member states commit public money to an industrial supply chain, Brussels grants an exemption from the ordinary rules on state aid, and the risk stays on the national books. Since then eleven projects have been carried forward, in particular on electric batteries, hydrogen and cloud capacity, and every round has widened the perimeter. More countries at the table, more sectors declared strategic, commitments stretched over longer and longer periods. Microelectronics got a second round in 2023, with a far wider group of states than the first. The direction has been the same for eight years. Then there's today's arithmetic. Nineteen sign-ups out of twenty-seven. Eight capitals stay outside the initial perimeter, and the makeup of that group will say a great deal about how European industry is distributed over the next decade.

Good morning. The answer sits in the letter of the Treaty. Article 107, paragraph 1, prohibits, as a general rule, aid that states grant to companies. Paragraph 3, point b, opens the exemption for aid intended to promote the execution of an important project of common European interest. The verb chosen is cautious. That aid "may be considered compatible" with the internal market, and the Commission's margin of assessment stays wide. The entry threshold is written down. At least four member states, large scale, high technological or financial risk, and a benefit that falls to the Union. The criteria are in the Commission Communication of 2021, published in the Official Journal on December 30, 2021. There are six items, and two of them shift the risk onto the company. First industrial deployment keeps the aid in the phase that comes before the mass market, and the claw-back of excess profits requires paying back a share when the project earns more than expected. The exemption applies insofar as those conditions are met.

Hello to everyone at the table, and let me start with the number almost everybody skips. Thirty-eight billion dollars per gigawatt. On September 14, in Vienna, ECB President Christine Lagarde laid out the European numbers on artificial intelligence, and the sum that stuck in the headlines was a different one. Six hundred billion euros. Here's where it comes from. The Commission's impact assessment points to a data center capacity gap that goes from three gigawatts in 2025 to roughly nineteen to twenty gigawatts in 2036. Epoch AI, on May 14, measured the capital needed for a one-gigawatt data center, servers included, and that's where the unit price comes from. The product works out to about seven hundred and twenty billion dollars, and in the Vienna speech that figure appears as an upper bound, in so many words. The debate reads it instead as a bill beyond reach, and therefore as a verdict. It's the product of two moving quantities. A capacity expected ten years from now, and a price measured in May. The price per gigawatt moves every quarter, and that's where I keep my eyes.

Sixteen gigawatts to build in eleven years works out to about a gigawatt and a half a year. Simple arithmetic on the two points in the source, and a construction-site pace, so it's manageable. But public capital arrives with a calendar, and political calendars are short. Vega, does a price that falls every quarter strike you as an argument for waiting for the next round?

Waiting is a bet on the price. Signing today is a bet on capacity. Whoever signs multi-year contracts now is buying this cycle's scarcity at this cycle's price. And the bottleneck always moves. From silicon to packaging, from memory to energy, all the way to human judgment. Whether that price has fallen in the last quarter, I haven't verified. The public figure I have in hand is still the one from May.

On the overall bill, let me add a distinction that changes the subject. That sum measures industrial investment as a whole, while public aid covers a narrow band. The phase that comes before the mass market. Private co-financing is one of the six criteria, and it serves to measure the industrial seriousness of the application, because the company's own capital goes into the project. Vega, is the number we're discussing really what Brussels is deciding on?

It's what the industrial decision is about, and that comes first. Public money decides where the first hole gets dug. The rest is decided by the cost curve. If private capital comes in alongside, the bill gets shared out, and then the question becomes which private capital. Whoever already has construction sites elsewhere brings the suppliers, the skills and the margins elsewhere too.

The collective notification opens the door. The money arrives project by project, state by state, after a separate review of each measure. Between the two phases, more than a year usually goes by. Eleven countries will now start down that path on the individual measures, and every file has a life of its own.

A year of review against a yearly construction pace is a serious mismatch. The first hole gets dug while the Brussels calendar is still turning. Atlas, does the procedure have a faster gear, or is that timeline structural?

It depends on the quality of the file that arrives. A project with the claw-back already written in, the co-financing documented and first industrial deployment defined precisely moves faster than one that shows up with the boxes still to be filled. The wide margin of assessment cuts both ways. It gives the Commission the freedom to approve, and at the same time the freedom to ask for more.

There's a piece that weighs more than the procedure. Euro-area households hold roughly four hundred and forty billion euros in shares of American tech companies. European savings are already funding the technology of the moment, with an American address on it. In Vienna the reference was to the Gründerzeit. French savers paid for the Suez Canal, and larger sums, British and German, ended up in American railroads. In 1873 the Vienna stock exchange crashed, and Wall Street followed four months later. Cato, does public money change that address?

It changes the starting point, and that precedent tells you why it matters. What that money had built stayed where it was built, and the growth stayed with it. By the end of the century the American economy was the largest in the world. One point of growth over five years, distributed asymmetrically, produces winners and losers inside the same single market, and the losers have parliaments, budgets and election calendars. The technical work behind that estimate is in working paper WP/26/180 from August 2026, and it's worth reading in full.

Here the law answers in advance, at least on paper. The criterion of positive spillovers beyond the state that pays exists for exactly this. The aid passes the test if the benefit crosses the borders of whoever is footing the bill. Then there are the transparency obligations on the aid granted, which make it verifiable who received what. The criterion is written down. How much weight it carries in the review of individual measures, the first decisions will tell us.

The names of the eight capitals that stayed out, and the moment one of them asks to come in. That request will tell us whether the scheme really attracts, or whether the perimeter closed on the first round. In parallel, I'm keeping count of how many of the national notifications go out before the end of the year.

The text of the Commission's first individual decisions, line by line, on the claw-back of excess profits. The threshold above which repayment kicks in, and how long the obligation lasts, tell you how much risk has really stayed with the company. A generic clause and a clause with dates in it are two different crafts.

The next measurement of the price per gigawatt, with a date and a source. And then the direction of that four hundred and forty billion in savings. If the flow stays put while the European construction sites get going, public capital is funding an infrastructure that European private investors are watching from a distance.