4 million versus 2. European rearmament is won in the factory, not at the front

In April 2026 the EU Commissioner for Defence, Andrius Kubilius, told the European Parliament something nobody wanted to hear: Russia produces four million artillery shells a year, the Union two. On cruise missiles the ratio is 1,100 to 300. On ballistic missiles, 900 to zero. In the same hearing he spoke of a “shortage of production capacity on both sides of the Atlantic.”

The money problem is solved, at least on paper. Readiness 2030 puts 800 billion euros on the table, 150 of them as SAFE loans, and in 2025 eighteen of the twenty-seven member states crossed two percent of GDP. The real problem is more embarrassing. In 2024 Brussels declared a production capacity of one million shells a year, later revised to 1.7 million; industry estimated between 550 and 600 thousand, and an EU official admitted the real figure was “about a third.” The Commission was not lying. It did not know.

Because Europe’s defense industrial base is made of more than 2,500 small and medium enterprises, which the Commission itself describes as organized “on a national basis, with limited cross-border cooperation” and which the European Defence Agency calls “fragmented”. Each one knows how much it can produce. Nobody knows how much they can produce together.

Every company knows how much it can produce. Nobody knows how much they can produce together. That is the bottleneck, and no check can buy it.

The promise of one million shells to Ukraine by March 2024 closed with about half a million delivered at the deadline; the million arrived in November, according to Josep Borrell. Over the same period Russia, according to Estonian intelligence, went from 400 thousand shells a year in 2021 to seven million in 2025, seventeen and a half times more, at under a hundred thousand rubles a shell. The metal exists on both sides. What is missing on one side only is time.

Looked at closely, the bottleneck is almost never the metal. It is chemistry and contracts. The RFE/RL investigation into Europe’s ammunition supply chain reports that Eurenco makes propellant charges for about 500 thousand shells a year, that Czech Explosia calls doubling output “impossible in a few months,” and that Norway’s Nammo refuses to invest without firm orders. Each link reasons on its own, and rightly so, because nobody shows it the whole picture.

The buyers do not have the picture either. Oliver Wyman, which analyzed more than 600 second- and third-tier suppliers, writes that “most supply chain departments do reactive problem-solving”, while European defense spending is expected to grow 80% by 2030, from 417 to 650-750 billion. Anyone who has worked in a factory knows what that means: you chase the urgent, and today’s urgency is the planning that was missing six months ago.

When the orders arrive, though, industry responds. Since 2022 European ammunition output has tripled: the Financial Times, using satellite imagery, counted seven million square meters of new plants across 150 sites of 37 companies. Rheinmetall built a plant at Unterlüß for about 500 million euros in fifteen months, going from 25 thousand to 350 thousand shells a year by 2027, inside a network aiming at one and a half million. So the problem moves again: onto whoever has no firm orders, and onto whoever cannot say how much they could produce if they had them.

Ukraine is always cited for the big numbers. The Ukrainian government talks about millions of drones produced a year, and those totals remain government claims, not independently verified. The interesting industrial figure is another one. In three years the country went from about seven to about five hundred drone makers, and the redesign cycle of a model, from front-line feedback to the new version in production, shrank to seven days.

The mechanism is well described on War on the Rocks: units send reports to the makers over Signal, within days, and the makers fix things. The cost per target hit fell from about $60,000 to about $1,000. The lesson for a European factory lies in the cycle, in how fast a piece of information from the customer becomes a change in production, far more than in the scale. And the cycle shortens first with data, with who owns it and how it moves, and only then with machines.

In the United States software has already entered production, in the literal sense. L3Harris reports using Palantir’s platform to find component shortages in night-vision goggle production and to read millions of shipping documents looking for the country of origin of every part; the result is a bill of materials rebuilt along the whole supply chain. Anduril is building Arsenal-1 in Ohio, a one-billion-dollar factory with four thousand jobs that it calls “software-defined”: the plant is designed starting from production data.

In Europe the gap is measurable. BearingPoint surveyed 151 executives in aerospace and defense across six countries, Italy included: 60 to 70% expect a high digital impact by 2028, and only 20-30% report advanced maturity. In the German sample 10% have an integrated data architecture, 7% list artificial intelligence among their priorities, and 44% name silos as the main barrier. The Italian breakdown is not published; Italy is in the sample.

Italy, seen through the numbers, has a 2026 defense budget of 32.4 billion euros, about 13.1 of it in investment; the top hundred companies in the sector bill 40.7 billion and ten of them make three quarters of the total. Below them are all the others: in the national aerospace cluster alone, 220 companies out of 276 are SMEs. That is the part of the supply chain we work with, and it is the part no platform sees. Companies we have met, with more than a million euros of revenue per employee, still work almost entirely on paper: specifications, non-conformities, and batch traceability in folders and spreadsheets. Annalisa Alberti of Rheinmetall Italia said it at the Manufacturing Forum in May: “we have to think about the smart factory, and therefore about introducing artificial intelligence. Those who don’t risk taking a step backwards.”

This is the point that concerns us. The software Europe needs works on the bill of materials, on specifications, on non-conformities, on the delivery dates of second-tier suppliers, and on one simple question: how much can we produce next month? In a defense company, software like that must be sovereign: the data stays inside the company’s perimeter, the models run where the company decides, the accumulated know-how is the company’s and stays so when it changes vendor. It is the thesis we wrote about data sovereignty, applied to a sector where it counts twice as much.

The software European defense needs works on bills of materials, specifications, and delivery dates. And it has to stay inside the company’s perimeter.

None of the three requires a single byte to leave the company, and none requires buying a model. They require one afternoon of management time, a place to put things, and a process to start from.

The first is to classify information into four classes, public, internal, restricted, and classified, and decide for each which tools are allowed. It takes an afternoon, with management rather than IT, and from that moment the use of artificial intelligence stops being an individual responsibility. The second is to put the company’s know-how and rules in one place, owned by the company, readable by people and by software: from there any model works with an operations log, and swapping the model tomorrow costs nothing. The third is to start from a document process, specifications or non-conformities, and leave production for later: the risk is low, the time returned is measured in weeks, and the method for choosing the first process is the same as for any other company.

Then there is one thing a ministry should do tomorrow: the map of national production capacity, one node per plant, where each company keeps ownership of its data and the buyer sees the status, how much and when, without seeing the content. Kubilius’s number, four versus two, comes from the absence of that map. The rules for building it without a single byte leaving are the ones in AI governance for SMEs, laid out for a factory in the second piece of this series. How the map should work, a federated layer where every node keeps ownership of its data, is the subject of the third, on the invisible supply chain.

1. [1]EU Perspectives, “Kubilius: we ask defence industry to speed up”, April 2026. euperspectives.eu
2. [2]RFE/RL, investigation into European shells for Ukraine. www.rferl.org
3. [3]Kyiv Independent, “EU has delivered over 980,000 shells to Ukraine”, November 11, 2024. kyivindependent.com
4. [4]Euromaidan Press, Estonian intelligence on Russian shell production, February 11, 2026. euromaidanpress.com
5. [5]The Defense Post, “European defense production tripled”, Financial Times data, August 14, 2025. thedefensepost.com
6. [6]Rheinmetall, opening of the Unterlüß plant, September 5, 2025. www.rheinmetall.com
7. [7]European Commission, Defence SMEs. defence-industry-space.ec.europa.eu
8. [8]EESC and European Defence Agency, “No more national preference”, February 2025. www.eesc.europa.eu
9. [9]Forbes, Oliver Wyman, “European defense buildup may cause supply chain delays and shortages”, June 25, 2025. www.forbes.com
10. [10]KSE Institute and Brave1, first comprehensive study of Ukraine’s drone industry. kse.ua
11. [11]Exponential View, “Ukraine’s seven-day drone advantage”. www.exponentialview.co
12. [12]War on the Rocks, “Inside Ukraine’s battlefield innovation loop”. warontherocks.com
13. [13]L3Harris, “Palantir and L3Harris: reindustrializing defense through AI”, December 16, 2025. www.l3harris.com
14. [14]Anduril, Arsenal-1, Ohio. www.anduril.com
15. [15]BearingPoint, “European defense sector faces a major digital execution gap”. www.bearingpoint.com
16. [16]BNP Paribas Economic Research, Readiness 2030 one year on, March 2, 2026. economic-research.bnpparibas.com
17. [17]European Parliamentary Research Service (EPRS), briefing on European rearmament, 2025. www.europarl.europa.eu
18. [18]Geopolitica.info, Italy’s 2026 defense budget. geopolitica.info
19. [19]Mediobanca, defense Top100 (PDF). euroborsa.it
20. [20]L’Imprenditore, CTNA, “Aerospace: a growing supply chain, 80% of the sector is SMEs”. www.limprenditore.com
21. [21]Startupbusiness, Manufacturing Forum 2026, Annalisa Alberti (Rheinmetall Italia). www.startupbusiness.it