- The 2027 question is “how much can you deliver in 90 days, and where do the components come from.” EDIP asks it, with the 35% rule, and so do the primes, with platforms that measure deliveries.
- Today in an SME the answer takes weeks: phone calls to the shop floor, a sheet, an estimate. In 2024 Europe declared a capacity three times the real one because nobody had the answer.
- The calculator compares the machines with the single-source component and says which one binds: almost always the supplier.
- Six data points are enough to answer within an hour, and none requires software: they require a decision on who keeps them up to date. The anonymous survey collects the first dataset on the supply chain.
From “how much does it cost” to “how much and when”
For twenty years the question a prime asked a supplier was “how much does it cost.” Since 2022 it has become “how much and when,” and it is becoming “how much, when, and from where.” Commissioner Kubilius told the European Parliament that Russia produces four million shells a year and the Union two: that number is a capacity question addressed to every plant in the supply chain. EDIP, approved in December 2025, turns it into a rule: in funded projects non-EU components cannot exceed 35% of cost, and to prove it you have to know the origin of every component.
Primes already have the tools to ask the question: Leonardo measures more than 5,000 suppliers on a platform that records confirmations and deliveries. Procurement agencies and ministries will follow, and the 2026 AI and Defense Strategy anticipates it by devoting the first year to data consolidation. The question “how much can you deliver in 90 days” will come by contract, and the answer will have to be a number, with the origin next to it.
Three weeks, three phone calls, and an estimate
Today, in a supply-chain SME, the question arrives by email and a chain starts: sales calls production, production calls purchasing, purchasing calls the critical component’s supplier, and after a few days someone adds up the parts on a sheet and adds a safety margin. The answer takes three weeks, and it is an estimate nobody wants to sign.
People do their job. The problem is six data points that live in six different places. The result, at scale, was visible in 2024: the Commission declared a European artillery shell capacity about three times what industry estimated, and an official admitted the real figure was “about a third.” Nobody was lying. The sum did not exist because every addend was in a phone call. Oliver Wyman measured it on sub-tier suppliers: “most supply chain departments do reactive problem-solving”.
Declared and real capacity diverge threefold for a trivial reason: the sum does not exist, because every addend is in a phone call.
Seven numbers, one minute, and the binding constraint
The calculator below is the miniature version of the map. It asks for seven numbers a production manager knows by heart: capacity per shift, shifts, working days, scrap, stock of the critical component, lead time and rate of the single-source supplier. It returns two things: how many parts in 90 days, and which constraint binds. Try the starting values: the machines could make almost fifty thousand parts, the supplier covers a third. The constraint sits outside the factory.
How many parts you can deliver in 90 days, and what stops you
Seven numbers a production manager knows by heart. The result is the lower of machine capacity and single-source component capacity: the second one almost always binds.
Good parts one line makes in a full shift
How many shifts actually run
Net of holidays, planned stops, leave
Parts reworked or thrown away, in percent
Finished parts covered by the component already in house
From today’s order to the first delivery
Finished parts covered per day, at steady state
It is not the machines. It is the single-source supplier: 120 days of lead time, and stock covers about 10 days of production. In between, the line stands still.
Assumptions
- Machine capacity = nominal per shift × shifts × working days × (1 − scrap).
- Component capacity = stock + supply rate × useful days after the first delivery, within the 90.
- One critical component: if you have more than one, the worst one counts. No added shift, no hiring, no second source within 90 days.
- A surface-level estimate, built to tell you which number to ask for first. The real number comes from the supplier sheet and the production log.
The assumptions are written under the tool, because a calculator without assumptions is a promise. The real number comes from the supplier sheet and the production log; the calculator is there to tell you which of the seven numbers to ask for first, and to show in one minute why the number on paper and the real one diverge. The tool has a page of its own, without the article around it, for those who only want the calculation.
The six data points to answer within an hour, and where they live today
To answer within an hour instead of three weeks you need six data points, and none of the six requires software to exist. It requires a decision: who keeps it up to date, and where. In most of the companies we have met the left-hand column is the real situation.
| Where it lives today | Where it should live | |
|---|---|---|
| 01Capacity per shift, per line | In the shop-floor manager’s head | One row per line, updated when tooling changes |
| 02Shifts and calendar for the next 90 days | In the attendance sheet | In the same sheet as capacity, with stops and leave |
| 03Scrap rate per product | Rebuilt at month end, on a good day | From the non-conformity log, by cause |
| 04Stock of critical components | In the warehouse, counted by eye | In master data, in finished-part equivalents |
| 05Lead time and rate of single-source suppliers | In emails with the supplier | In the supplier map, one row per component |
| 06Country of origin of every component | On delivery notes, when present | In the same map: EDIP will ask for it first |
When the six data points sit in the same place, the one where the company’s know-how and rules live, the answer to the customer becomes a reading instead of a reconstruction, and that place becomes the node the piece on the invisible supply chain talks about: the status, how much and when and from where, exposed to the buyer without exposing the content. The method to build that place without a byte leaving is in the second piece; the cycle by which the data updates is in the fourth.
Nobody has this data. We are collecting it, anonymously
How many supply-chain SMEs can answer within an hour? How many know the country of origin of their components? How many have a single-source critical component and know it? No such data exists, and without data the debate on AI and defense stays made of opinions. That is why we are opening an anonymous ten-question survey: three minutes, no name, no company. With forty answers it is a note; with a hundred it is a chapter the Ministry reads.
The aggregate answers go into the The invisible Arsenal program, together with the Arsenal scale, the twelve-question self-assessment on the factory’s relationship with data. Districts, clusters, and associations that distribute the survey to their companies receive their group’s results.
The 90 days, in practice
Why will a customer ask how much I can produce in 90 days?
Because the sector’s question has moved from “how much does it cost” to “how much and when.” EDIP funds projects with the 35% rule on non-EU components, which requires knowing the origin of every component, and primes already measure supplier deliveries in real time on their platforms. The 2026 AI and Defense Strategy devotes its first year to data consolidation: the question will come by contract.
How does the 90-day calculator work?
It compares machine capacity (nominal per shift, shifts, working days, scrap) with the capacity of the single-source critical component (stock, lead time, supply rate after the first delivery) and returns the lower of the two, the binding constraint, and the ratio between declared and real capacity. Everything runs in the browser, with the assumptions written underneath.
Which data do you need to answer within an hour?
Six: capacity per shift and per line, shifts and calendar for the next 90 days, scrap rate per product, stock of critical components, lead time and rate of single-source suppliers, country of origin of every component. None requires software to exist: it requires a decision on who keeps it up to date and where.
Is the survey really anonymous?
Yes. Ten closed questions, no name, no company, no production data. Email is optional, used only to receive the aggregate results, and is not linked to the published answers. The answers are aggregated in the Arsenal Report.
What does EDIP have to do with production capacity?
The European Defence Industry Programme, given final approval in December 2025, funds projects on condition that non-EU components do not exceed 35% of cost. To comply you have to know the origin of every component and therefore have the supply-chain map: the same map you need to answer “how much can you produce.”
Sources
- [1]EU Perspectives, “Kubilius: we ask defence industry to speed up”, April 2026. euperspectives.eu
- [2]Council of the EU, “European Defence Industry Programme: Council gives final approval”, December 8, 2025. www.consilium.europa.eu
- [3]SupplyOn, “Leonardo extends AirSupply to the entire supply chain and kicks off second phase of LEAP”. www.supplyon.com
- [4]Agenda Digitale, “Strategia IA e Difesa 2026: cosa cambia per imprese, filiere e sovranità tecnologica”. www.agendadigitale.eu
- [5]RFE/RL, investigation into European shells for Ukraine. www.rferl.org
- [6]Forbes, Oliver Wyman, “European defense buildup may cause supply chain delays and shortages”, June 25, 2025. www.forbes.com
This page is written by Raffaele Zarrelli, founder of Yempik, with editing done with Claude. It is the fifth piece in the series on AI for the defense factory and supply chain. The calculator is a simplified model with written assumptions; the numbers cited link to their source in the text. No real client data appears in the piece; the supply-chain companies we met are not named.
Answer the survey: three minutes, no data.
Ten anonymous questions on the visibility of your supply chain. With a hundred answers it becomes the first dataset the Ministry can read. If you would rather talk, write to us: we sell nothing on the first call.