Ukraine's reconstruction: what the money passes through.

How a pricing annex decided an industry

What we think

Three things this does not claim

Where we are most likely to be wrong

What we would look at, if this is right

Volume is the wrong variable

Roughly sixty-three percent of the money asks nothing about structure

Four constraints belong on that table. The figures are ceilings as at 28 May rather than disbursements. The shares are arithmetic on the ministry's own allocations, not a published statistic. They cover EU money only. And carrying no reform conditions does not make the defence leg unconditional, which matters, because what it does carry is the sharpest evidence here.

The domestic implementing act, adopted 24 June 2026, routes the entire defence-industrial leg through one budget programme with the Ministry of Defence as manager and a single state enterprise as sole recipient.

What the Commission requires of that flow is precise. Procurement proceeds only against a list of products the Commission has positively assessed. The recipient must obtain declarations that the product, the supplier and the producer meet eligibility criteria. Contracts must oblige suppliers to grant the Ministry and the Commission access to documents and to the premises where they are held. The Treasury notifies account openings within one working day and supplies weekly statements. And access runs to the anti-fraud office, the European Public Prosecutor and the Court of Auditors.

That is close supervision by any standard. None of it asks who receives the money. It asks whether the goods are eligible, whether the supplier is permitted, and whether the euro was spent on what it was declared for.

The largest single flow in this reconstruction carries transaction-level conditionality and no structure-level conditionality at all. The instruments that would ask, state aid discipline, public-service separation, insolvency, loan resolution, sit on the eighteen-point-six percent, and §5 traces what became of them.

Two domestic details belong alongside it. Advance payment runs to 100 percent of contract value. And the sole recipient is itself being merged into a new acquisition agency, so the whole leg routes through an entity under restructuring.

Ordinarily a lender needs no conditionality, because the obligation to repay does the work. That mechanism is absent by construction. Principal is repayable exclusively from future Russian reparations, and where those fall short the EU has no claim on any other Ukrainian asset.

Conditionality is therefore not one lever among several. It is the lever. Which places weight on a backlog that is not clearing: nineteen indicators unfulfilled as at July 2026, with roughly €7.35 billion contingent on them. On 30 July the Council mobilised over €8 billion more with twenty-seven new milestones, taking the indicator count from 146 to 173, all due by Q3 2027.

Conditionality is expanding, not easing, and the unfulfilled cluster is the one that governs routing.

Everything above operates at the level of programmes. Legs, tranches, indicators, milestones. That is where the money is designed, and where almost all analysis of this reconstruction is conducted, including the good analysis.

Everything that follows was decided one level down, in profit schedules, eligibility registers, cost-recovery rules and customs lines. Those are not in the Plan and they are not indicators. They operate at instrument altitude; tranche negotiations operate at programme altitude, and the same person is rarely in both.

Hold that, because it is what the sectors are about.

The government's own Programme of Activities for 2026 and 2027, submitted on 18 August, is this upper altitude in its clearest domestic form: four directions, twelve goals, quantitative criteria and 2027 deadlines, and none of the instruments that dispose of the money. It is the Kyiv counterpart to the donor architecture above it, planned at the level that allocates and silent at the level that decides.

The investment pillar of the Facility is the Ukraine Investment Framework: €9.6 billion of capacity, €7.8 billion of guarantee cover plus €1.8 billion of blended finance, aiming to mobilise up to €40 billion. As reported at its seventh Steering Board meeting, €8.5 billion has been allocated against €25.7 billion mobilised, which is 88.5 percent of capacity committed at leverage of about three times on allocation.

It is implemented in indirect management by financial institutions the Commission has pillar-assessed: the EIB Group; the EBRD, IFC and Council of Europe Development Bank; and the European development banks BGK, KfW and AFD. Every one is European or multilateral. No Ukrainian institution appears on the list.

The first Call for Expressions of Interest ran from November 2024 and was open to EU and EEA businesses, drawing 113 investment projects. The second, announced at the 2026 EU–Ukraine Business Summit, is open to European and Ukrainian businesses on a rolling basis.

That sequence is not a preference anyone stated. It is the eligibility rule, and eligibility rules are where routing is actually decided.

The Ukraine Facility Regulation carries two quantitative targets: at least 20 percent of framework funding together with Plan investments toward green investment, and a minimum 15 percent of guarantee capacity to micro, small and medium enterprises.

Fifteen percent of €7.8 billion is about €1.17 billion, 12.2 percent of total capacity, and roughly 8.8 percent of the €40 billion mobilisation target if leverage holds.

That is the third quantified instance of the same asymmetry we have found. In October 2023 a single package of $480 million, IFC $130 million, EBRD $100 million, DFC $250 million, went to MHP, Ukraine's largest poultry and processed-meat producer, against roughly €20 million capitalising the Partial Credit Guarantee Fund in Agriculture launched three months later. The two figures are not like for like: one is direct lending to a single borrower, the other capitalises a facility that leverages into a larger loan book, and the orders of magnitude are the point rather than the ratio. §5 gives the third inside a single domestic instrument, where the same subsidy through the same forty-eight banks still splits by lender.

The entrant channel exists, it is mandated, and it is small by design. No individual decision produced that. Three separate instruments, set by three different bodies, arrived at the same proportion.

Everything above argues from instrument design. Set it against what businesses say they need.

A survey of 504 owners and chief executives, fielded across Ukraine in early 2025 with UNDP and the Ministry of Economy, asked how much additional financing a business needed to execute its three-year strategy. Half answered between thirty thousand and three hundred thousand dollars. Fifteen percent answered under thirty thousand. Eighty-six percent need a million dollars or less.

Set that beside the instruments. A single package of $480 million to one agribusiness. A €1.5 billion acquisition financed with European guarantees and named as the framework's flagship result. A guarantee facility for farms below five hundred hectares capitalised at around €20 million. The instruments and the demand are separated by two to three orders of magnitude.

Two caveats. The survey is self-report and measures what owners say binds them. And stating a need is not being creditworthy for it. What it establishes is scale, and on scale it is not close.

The same survey asked which forms of state support a business had used in the previous twelve months, from a list of eighteen. Fifty-three percent answered none. Subsidised credit reached 18.7 percent, international grants 10.7, reservation of staff from conscription 9.5. Everything else was under 6.5, including the tax regime the reconstruction read named as one of three things that built an industry, at 0.8.

That last figure is worth pausing on rather than scoring a point with. Diia.City is written for technology firms and this survey is cross-sector, so a low number is expected. What it shows is what a routing instrument looks like when it works: narrow, specific, and close to invisible to almost everyone. The instruments here are not failing to reach a general population. They were never addressed to one.

The Commission names two flagship private-sector results.

In telecoms, with EU guarantees behind it, the French holding NJJ invested €1.5 billion acquiring and merging two leading Ukrainian operators, Lifecell and Datagroup-Volia, into a single integrated operator, financed alongside the IFC and EBRD. The Commission describes it as the largest foreign direct investment in Ukraine in two decades, bringing improved mobile connectivity to more than ten million people and faster internet to four million households.

In agriculture, a partial first-loss risk cover went to Kernel Group for a 106 megawatt solar park with battery storage. Kernel is the group behind the landmark crop-farming transaction of 2026, the $348 million acquisition of Enselco.

Nothing in either case suggests impropriety, both carry substantial and specific public benefits, and both are the Commission's own selection, chosen to demonstrate impact, illustrative of what the instrument celebrates rather than a representative sample of what it funds. The observation is about the instrument, not the recipients. An instrument that optimises for capital mobilised at scale will find the transactions that move the most capital, and those transactions are acquisitions and mergers. Nobody decided to fund consolidation. The instrument's headline results are consolidation because consolidation is what moves €1.5 billion at once.

Ticket size, eligibility, conditions and channel all select who is served. A fifth class does something else: it decides whether anyone can afford to serve at all.

Margin regulation. A profit schedule adopted in March 2021 paid a defence contractor one percent on what it bought in and thirty percent on what it made itself, with a third rate paying three percent to an authorised entity importing finished goods. In March 2023 that was replaced, for unmanned systems, by a flat twenty-five percent. §4 traces what followed.

What belongs here is the class, and two things about how it behaves.

It prices activities rather than firms. The state did not prohibit importing, and it did not favour assembly by naming assemblers. It priced the activities differently and let the arithmetic sort them. That is why nobody had to decide anything for the structure to change.

And a margin instrument can carry an eligibility instrument inside it. The same resolution that lifted the ceiling required a producer to obtain a valuation from a forensic-expert institution before contracting, and the paragraph after it forbade recovering that cost in the price. The margin made entry possible. The valuation decided who could act on it, and charged them for finding out.

The valuation has a reason and it belongs here rather than in a note. Defence procurement prices needed policing and still do: in the first half of 2026 Ukraine's anti-corruption bodies recorded UAH 726 million saved on unmanned-systems procurement, and a separate case involving a million-dollar bribe over drone contracts. A state that requires an independent valuation before it signs is not inventing friction. It is answering something real. It filters anyway, and both are true at once.

Denmark occupies a role the read should name precisely: it is not only a contributor but the implementing actor through which the EU disburses windfall profits from immobilised Russian assets. That makes it routing infrastructure.

The model on its own terms ran roughly €538 million in 2024 and €1.3 billion in 2025; the wider family of mechanisms ran about $400 million in 2024 and roughly $6 billion in 2025. Against a defence capacity gap of $45 billion, €1.3 billion is about 3 percent and even the family is about 13 percent. The architecture we hold up as the correct one covered a small fraction of the capacity it was designed to activate.

Then it was adopted. On 30 June 2026 the EU disbursed €3.9 billion to Ukraine's State Budget Special Fund, the first instalment of a €6 billion drone package under the €90 billion loan, restricted to Ukrainian drone production, defence-industrial capacity and urgent frontline supplies, with Brussels verifying intended use and delay of subsequent transfers available where it is not observed.

A small country demonstrated a routing architecture, it worked, it stayed marginal in volume for two years, and the largest funder in the system adopted it and multiplied it threefold in a single disbursement. Routing architectures propagate by demonstration rather than by design, and that is the one finding in this section that runs in the read's favour.

Much of this money is European, and the European instinct is that money should carry conditions. It leaves us with a question we cannot answer and would rather put than pretend to resolve.

The conditioned money is the money that does not move.

The defence leg carries no reform conditions and it disburses. The structural leg carries nineteen unfulfilled indicators, roughly seven billion euro contingent on them, and a state aid step overdue since late 2025 with its enabling bill withdrawn.

And the instrument that reorganised the industry carried no condition at all. A profit schedule changed in Kyiv in March 2023, and eighteen months later Ukraine had a defence-industrial base it did not have before. None of the Commission's supervision touched the variable that decided which firms could exist.

There is a homely version of the objection. A mortgage with a schedule specifying how much may be spent on the roof, the terrace and the kitchen. The lender is not in the house. The owner is, and knows which room is failing.

It understates the case. A mortgage lender holds a lien, so line-item control at least protects a claim. The Ukraine Support Loan is repayable only from future Russian reparations, and where those fall short the Union has no claim on any other Ukrainian asset. The lender here waived the security and kept the schedule.

And the objection has a serious answer, which is why the conditions exist. A household's kitchen does not reshape an economy; procurement routing does. A donor that hands over ninety billion euro with no view on market structure is not being respectful, it is being absent.

Underneath sits a confusion worth separating, because need and trust are different animals and this architecture treats them as one.

Every condition in it is a trust instrument. Weekly statements, premises access, the anti-fraud office and the public prosecutor on one leg. Audit, state aid control, procurement transparency on the other. Each answers can we be sure this money is not stolen. None answers does this money produce the industry anyone wants to exist. You condition on probity when managing trust. You would condition on profit schedules and eligibility registers if you were managing effectiveness, and nobody does.

And need is why the mismatch persists. A borrower who needs the money accepts terms a borrower who does not would negotiate, so the conditions were never tested against the mechanics. That is not a criticism of anyone. It is what happens when the party with the leverage manages one variable and the party with the knowledge manages another.

The same mismatch has a domestic form. The Programme carries quantitative criteria, but sets their deadlines to 2027 with no interim milestones, so at any point before the end of 2027 the executor is within its deadline and the only accountability question available is whether the work is late, which it cannot yet be. The trust-versus-effectiveness split is not only in the donor conditions, it is in the domestic instrument that answers to them.

Ukraine's anti-corruption bodies published their first-half results in August: one hundred and seven notices of suspicion, one hundred and six brought to trial, fifty-five guilty verdicts in force, and a recorded economic impact of UAH 2.06 billion.

Which sharpens the question rather than answering it. Those are instruments built to find what is missing. What finds what was allocated in the open? The lease without auction moved lawfully. The profit schedule was published in 2021. The guarantee register is a public dataset. None of it was concealed, and none of it is what an instrument built to detect concealment is looking for.

So a control system calibrated for theft works exactly as designed and still does not see the thing that determined the outcome. Whether that argues for fewer conditions, different conditions, or the same conditions applied by someone reading different documents, we do not know, and we do not think it is currently being asked.

The instruments are in place. The mechanism inside them is not.

Project-preparation facilities that would take an investment idea to bankable stage are largely absent for the private sector, so selection happens by default rather than by decision. The National Development Institution is eight months old and is redesigning the flagship subsidised-credit instrument now. Eighteen distinct instruments adopted across 2024 to 2026 govern how public investment is documented and assessed, each mapped to the EU provision it approaches, and the recurring finding across them is that formal adoption and functional capacity diverge, particularly at local level. None of the eighteen governs who may bid. Together they govern who is able to apply, which arrives at the same place more quietly. The Government Action Programme states priorities without specifying the mechanism that would deliver them, and we treat announced programmes as stated intent and never as routing, the same discipline we apply to donor announcements, with no exception for domestic ones.

Routing is therefore being determined by instrument design, ticket size, eligibility rules and administrative capacity rather than by anyone's decision. That is not a gap in the evidence. It is the finding, and it is why the decisions are still movable.

The date that governs is Q3 2027, when the remaining Ukraine Plan reforms fall due. What is unsettled is whether the instruments that would discipline routing, state aid control, the public-service separation that makes state-enterprise subsidy visible, insolvency, non-performing loan resolution, arrive inside that window or after it, while 62.9 percent of the money continues to arrive without asking.

Something has to move

Energy is the sector where the bypass was least available. The network cannot be routed around at transmission level, restoration rebuilds what was there, and the incumbent measures worst on productivity. The parent piece treated that as energy's problem, but it is not only energy's problem; it is the constraint that shaped every mobile sector's geography.

The sector that was most incumbent-locked pushed the sector that was least incumbent-locked into the one part of the country where the bypass was physically available. Defence's dispersion is partly a downstream consequence of energy's inability to disperse.

That is uncomfortable in a useful way. It means the successes and the failures in this reconstruction are not independent observations to be tallied against each other. They are one mechanism seen from two ends.

And it suggests a movement worth naming. If industry relocated to where power is most reliable, and that power is still not reliable enough, the next step is for manufacturers to build their own, distributed generation, storage, on-site capacity. The bypass would then be propagating from defence into energy through the back door, not by policy but because a relocated manufacturer needs power and building it is faster than waiting for restoration. This is a hypothesis and we carry it as one. It appears in §10 as a dated test, not here as a finding.

The three properties above were all true in 2022. The dispersion did not happen in 2022.

Seven manufacturers. Three to five thousand units. Components available off the shelf and purchasable online. Nothing about mobility, topology or siting changed in March 2023, and yet the producer count reached roughly five hundred and output passed four million units a year.

What changed was a paragraph in a pricing schedule.

From March 2021, profit inside a defence contract price was set at one percent of what a producer bought in, components, semi-finished goods, special equipment, subcontracted work, and thirty percent of everything else in production cost. A third rate paid three percent of contract value to an authorised intermediary importing finished goods.

Read that as a schedule of activities rather than as a price control:

Assembling in Ukraine from bought-in components paid an effective rate approaching one percent. Importing into Ukraine through an authorised entity paid three percent of contract value. Manufacturing in Ukraine with in-house capability paid up to thirty.

Assembling in Ukraine paid less than importing into it. And a drone is, definitionally, an assembly of bought-in components.

The rule did not cap profiteering. It priced business models, and it paid least for exactly the model the entrants had, while paying most for the model the state estate has by Soviet design. In March 2023 that schedule was replaced, for unmanned systems, with a flat twenty-five percent of production cost regardless of composition. The assembler was permitted, on paper, one percent, and now twenty-five. The integrated producer, thirty, and now twenty-five. What the schedule permits is not what any contract pays. But the schedule moved, and in one direction only.

That is not a deregulation. It is a reallocation from integrated incumbents to assemblers, executed through a pricing formula, and it is the mechanism we came to describe: an instrument that reads as being about prices turns out to be about which industrial structure can exist.

Mobility, topology and siting are necessary conditions. They make a bypass possible and they say where it would go. They do not make it happen. What makes it happen is whether the instrument pays for it, which is why we say routing is determined by instrument design rather than by anyone's decision, and why the same three properties produced nothing at all for twelve months while the schedule stood.

One layer sits upstream of all four properties and it is the least visible.

Capital does not route to a project. It routes to a bankable project, one with engineering studies done, independent validation obtained, and documentation a credit committee can act on. That preparation is paid for before any financing decision is taken, by whoever is proposing it, under uncertainty about whether financing follows.

As the Economy Minister frames it, Ukraine has never lacked investment ideas and has few that are prepared to bankable stage, and the project-preparation facilities that would close the gap largely do not exist for the private sector. That is his stated position rather than a measurement, and we carry it at that weight. What it describes, if accurate, is selection by balance sheet occurring before any instrument has been designed and before anyone has made a decision that could be reviewed.

A survey of five hundred owners and chief executives in early 2025 puts a number on what is being prepared. Half of them needed between thirty thousand and three hundred thousand dollars to execute their strategy. Preparation costs for a project of that size are not trivial against the project, and they fall on the proposer, before anyone has said yes.

The pipeline selects for balance sheet before the routing question is even asked. That is the earliest of the three stages of one mechanism, bankability first, then eligibility, then ticket size, and it is the stage with no document to point at.

Four properties, one upstream layer, and a finding that ties them.

Something must move, the asset or the route. Where it moves is set by three structural conditions and one military one. Whether it moves at all is set by whether the instrument pays for the activity. And before any of that, the proposal has to be financeable enough to be considered.

None of it requires anyone to decide who receives the money, which is why nobody did.

The decision still open is what happens when martial law ends. The schedule that paid one percent on bought-in content was not repealed in July 2023. Its first subparagraph was suspended for the duration of the martial-law regime, by name, in a single clause, while the subparagraph paying three percent on imports was left in force and still is. The suspended rule is unamended since March 2021 and carries no expiry of its own.

On the day martial law ends, it returns, unless something replaces it first. The flat twenty-five percent that made the entrant industry viable sits inside a temporary regime; the rule that made assembly unprofitable sits outside it, waiting.

What did not change

What did change

The objection, at its strongest

A reviewer's first move is to say none of this is routing. Drones proliferated because drones are cheap. The capital threshold for assembling a first-person-view aircraft from purchased components is a few hundred dollars and a bench, and when a technology becomes that accessible under that much demand, producers appear and the procurement rules follow the dispersion rather than causing it.

The objection fails on sequence, and the sequence is not close.

The technology was equally available in 2022, the components equally purchasable, the demand if anything more acute. Seven producers, three to five thousand units. Nothing technical changed in March 2023.

What the rule actually said

The rule that changed is easy to miss, because it does not look like an industrial policy. It looks like a line in a pricing annex.

From March 2021, a year before the invasion, profit inside a defence contract price was set at one percent of what a producer bought in, meaning components, semi-finished goods, special equipment and work subcontracted to other firms. Everything else in production cost carried thirty percent. A third rate paid three percent of contract value to an authorised intermediary importing finished goods.

Read it as a schedule of activities rather than as a cap on profiteering. Assembling in Ukraine from bought-in components earned an effective rate approaching one percent. Importing into Ukraine through an authorised entity earned three percent of contract value. Manufacturing in Ukraine with in-house capability earned up to thirty.

A first-person-view drone is, definitionally, an assembly of bought-in components, so its cost base sits almost entirely in the category the schedule paid one percent on. The same schedule paid thirty percent to a producer that machined, tooled and assembled in-house, which is what the inherited Soviet plants do by construction.

The rule did not cap margins. It priced business models, and it paid least for exactly the model the entrants had. Assembling in Ukraine paid less than importing into it.

Picture a workshop in Zakarpattia in the winter of 2022. Six people, a bench, a bill of materials that is ninety percent bought in. The army wants everything they can build. On that schedule, the price they are permitted to charge returns them about one percent of what the thing cost to make, and they have to finance the run before they see any of it. There is no version of that business that raises capital, hires, or builds a second line.

March 2023 replaced the schedule, for unmanned systems, with the flat twenty-five percent. The assembler was permitted one percent, and now twenty-five. The integrated producer, thirty, and now twenty-five. That was not a general loosening. It was a transfer, and it ran from the incumbent to the entrant.

What this section cannot establish

Demand rose sharply over the same period for reasons that have nothing to do with pricing, an artillery shortage running through 2022 to 2024 and an enemy with numerical superiority. Demand and routing moved together and cannot be cleanly separated. The honest formulation is that demand created the pull and the schedule created the ability to answer it. No private industry scales on a margin approaching one percent however badly the army wants the product, so the pricing change was binding, but it was not sole.

Three further limits belong in the open rather than in a footnote.

Whether the schedule actually bound between 2021 and 2023 is not fully established. The Deputy Minister of Economy has said the profit norm could be applied by state customers in that period, and the modal verb is doing work. It suggests contested practice rather than settled application. If customers routinely disregarded the norm, the mechanism is weaker than we have stated it.

The schedule did not act alone. The advance payment, the six disapplications and manufacturer-led certification each removed a distinct barrier a workshop could not have crossed on margin alone. And beyond the instrument stack the war itself was doing its own work: military demand ran unprecedented against a shortage the state could not procure through, design cycles for unmanned systems compressed to weeks, the frontline pulled producers toward configurations faster than any procurement office could specify them, and in certain production segments the entry threshold was low enough that a bench with a soldering iron and a bill of materials constituted a factory. The schedule opened the door. It did not push anyone through it.

The ceiling is also not what any producer took home. Warranty and defect-remediation are covered from the profit component after the contract signs, so a formal margin of twenty-five percent absorbs its own claims before it reaches the balance sheet. Currency-hedging premia on imported components add another three percent or more, and a workshop-scale producer buying most of its bill of materials abroad wears most of that. The schedule's ceiling and the producer's take-home are different numbers, and this section does not have the second one.

What was not the capital story

Brave1 launched a month after the pricing change and is usually named as the state's instrument for the drone expansion. Its grants total roughly eight to ten million dollars across some four hundred awards.

Against an industry producing millions of units a year, that is not a capital programme.

The third thing usually named is Diia.City, and that one is real. It is a tax and legal regime for technology firms, and the design and production of unmanned aerial vehicles is on its list of qualifying activities by name, as is the production of technological products for defence use. A drone manufacturer that qualifies pays five percent personal income tax, can hire on a contract form available to nobody else, and gets convertible loans and liquidation preferences written into Ukrainian law so that a foreign venture investor recognises the paperwork. The state guarantees those terms for twenty-five years, and the register opened sixteen days before the invasion.

That is the same kind of instrument as the profit schedule, pointed the other way and built to last.

It is also not available to the workshop. Residency requires at least nine employees or gig specialists and an average monthly remuneration of at least twelve hundred euros, which is somewhere around fifty-five thousand hryvnia and well above what an assembly operation pays. A firm can enter as a startup with fewer people for two years, but the salary floor applies from the first month, and failing to reach nine by the end of the second year claws the tax back.

So the six people at the bench get the pricing derogation, and the company that has already grown to nine engineers on European salaries gets the twenty-five year regime. Nobody designed it that way. A threshold set for sensible reasons, to keep the regime for real technology firms rather than shells, sorts producers by an attribute that tracks size, which is what every other instrument in this piece also does. Brave1 does coordination, demand signalling and a marketplace, and it does them well. The money that built the industry came through contracts priced under a changed schedule, not through grants.

The distinction matters beyond bookkeeping. A grant programme is a subsidy layered on top of an existing routing structure, and the evidence on subsidies is discouraging where incumbents are entrenched. What happened here was the routing structure itself changing, which is why that objection does not reach this case. Ukraine did not subsidise entrants. It stopped paying them one percent.

What this leaves

Test one supports the claim, and more precisely than the parent piece could.

Routing determined structure. The instrument was a paragraph in a procurement schedule that nobody described as industrial policy, adopted before the war for reasons unrelated to it, and when it was switched off for one product class that class produced the only genuine industrial dispersion in this reconstruction.

The switch has not been thrown back, because it was never a switch. The paragraph paying one percent on bought-in content was not repealed in 2023. Its operative sentence was suspended for the duration of martial law, by name, in a single clause, while the sentence paying three percent on imports was left standing and stands today. The suspended rule is unamended since March 2021 and carries no expiry of its own.

One more thing about this case is worth stating, because it is why we can describe it at all.

Normally you cannot see this. A pricing schedule changes, and its effect arrives over a decade, mixed with a currency move, a trade agreement, a technology shift and three governments. By the time an industry has a different shape, the paragraph that started it is unrecoverable, and the story gets told as competitiveness or culture or entrepreneurship.

Here it took eighteen months and four documents. The schedule is dated, the replacement is dated, the registrations are counted by year, and the interval between them is short enough that nothing else plausible fits in it. That is not because Ukraine is unusual. It is because the clock ran fast enough to leave a trace.

The contrast is inside this piece. Ukrainian manufacturing state enterprises gained market share for two decades while their productivity fell relative to the firms losing it, and two decades is long enough for the explanation to become anything you like: management, culture, legacy, the war. The same mechanism was running there. Only one of the two ran fast enough to be seen.

One limit belongs here rather than in a note, because this section is where a reader is most likely to overrate what routing does.

Ukraine can now build about fifty-five billion dollars of defence equipment a year and can pay for roughly ten. That gap was fourteen billion in early 2024 and is forty-five now, thirty-five of it in long and medium-range strike, and it widened over exactly the years the routing was working. Routing determines who receives the money. It does not determine how much money there is. No paragraph in a procurement schedule reaches a shortfall set by parliaments in other countries.

So the decision still open is not about procurement volumes and will not be taken at a summit. It is whether one subparagraph is repealed or replaced before martial law ends, and if it is not, the schedule returns to an industry that exists because it was switched off.

The subparagraph is four lines long. The window for converting it into whatever replaces it is the same window in which everyone capable of doing so is fully occupied with the war it was written for.

Ukraine observes the rooms where its routing is set

Three settings, dated within the year

The finding that has to be explained

Two functions, one set of balance sheets

What survives the comparison is sharper than what it removed. In Germany the small-ticket local lenders are not the institutions holding the sovereign's debt. Two functions, two sets of balance sheets, in a system with a three-pillar structure written into law.

In Ukraine they are the same institutions. State banks carry the entrant-credit load and hold roughly three-quarters of domestic government debt securities, in a sector where those run at about 25 percent of bank assets against a European median of 17. Counting subsidised loans and the eOselia mortgage architecture, 33 percent of Ukrainian bank assets depend on public finances.

And the banks carrying both functions run the thinnest capital in the system. At the end of March the sector held regulatory capital adequacy of 15.8 percent against a ten percent minimum, but state-owned banks sat at about fourteen percent and foreign-owned banks at about twenty-two and a half. The institutions doing two jobs have the smallest cushion for either.

Germany separates the two functions; Ukraine concentrates them, and that is a routing decision. The same balance sheets fund the state and fund the entrants, so sovereign stress and entrant credit are one exposure rather than two. The National Bank names the consequence itself: state banks increasingly vulnerable to fiscal risks, with concentration to be reflected in higher capital requirements from 2027.

There is a fiscal turn of the same loop. The state taxes bank profit at fifty percent, and the banks paying it are the ones holding its debt: sector net profit fell from UAH 65.5bn to 45.0bn over five months while pre-tax profit rose, the difference being tax alone. The National Bank states the consequence itself, that the fiscal burden reduces banks' capacity to accumulate capital and that at current profitability the sector cannot expand assets by more than fifteen to twenty percent a year. That is a stated ceiling on the channel the entrants borrow through.

And 5-7-9 is subsidised by construction, so the institutions carrying the small-ticket load have their margins compressed by design. The same pressure that weakens resilience elsewhere, applied to banks with no separation of function to fall back on.

A state guarantee directs capital to a named enterprise and does not appear in the budget. The Budget Code is explicit: before the guarantee case arises, transactions related to state-guaranteed debt do not affect the state budget. It is the cheapest routing instrument a state has, because the fiscal cost lands only if the borrower fails.

The register runs to twenty-one years, 142 guarantees, UAH 504.0 billion, and the distribution is not close. One borrower holds 21.3 percent of all guaranteed value. The top three hold 53.5 percent, the top ten hold 85.2. Named entities took UAH 483.0 billion across 136 guarantees, 95.8 percent of everything issued. Portfolio and small-enterprise guarantees took UAH 21.1 billion across six.

For eighteen years the entrant channel did not exist as a category. Zero percent of issuance went to portfolio or SME instruments between 2004 and 2013, and zero again between 2014 and 2021. A manufacturer with forty employees and a viable order book could not have applied, because there was nothing to apply to. The share reached 13.8 percent only in 2022 to 2025.

Two constraints belong on that finding and we carry them here rather than in a note. Concentration is partly structural: the largest borrowers are infrastructure monopolies, the gas incumbent, the railway, the grid operators, and guaranteeing a monopoly is not routing around a competitor. And the register records issuance, not outcome; what the guaranteed capital built is not in it.

What the constraints do not soften is the shape. Manufacturing is nearly absent from twenty-one years of the instrument. No manufacturing enterprise appears in the top ten; the closest is a defence industrial enterprise at 2.4 percent across two guarantees.

If guarantees did not sustain manufacturing state-enterprise share, something else did, and it was diagnosed before the war rather than during it.

An IMF mission worked in Kyiv in early 2015 and reported in March. It found budget subsidies to state enterprises at about 2.5 percent of GDP and indirect subsidies, outside the budget entirely, as high as 6 percent. It found the cross-subsidy quantified inside the tariffs: rail cargo charged above cost to carry passenger services at 40 to 70 percent, costing around UAH 8 billion; industrial electricity users charged above marginal cost to subsidise households; gas priced about 80 percent below cost recovery.

And it found the sector doing, in the report's own words, the social work of employer of last resort. Over a million people, more than 2 percent of GDP in wages, with the railway alone carrying an excess workforce near 100,000, a third of its staff. Those hundred thousand jobs are the reason nobody wants to touch it, and saying so is not a criticism of anyone who has declined to.

That is routing inside the firm and inside the price. It appears in no debt register, which is why an instrument-based test could not find it. A guarantee is visible, dated and attributable to a Cabinet decision. A tariff that charges one shipper above cost so another pays less is the same act of direction with nothing to point at, and the shipper paying it has no way of knowing how much.

The 2015 figures are historical and the levels do not transfer. The structure does, and the same concentration shape recurs across eleven years and two unrelated instruments: in 2014 the top 50 state enterprises took more than 80 percent of sector revenue, and across 2004 to 2025 the top 10 guarantee borrowers take 85.2 percent of guaranteed value.

The date matters more than either number. The diagnosis is from March 2015, seven years before the war. Wartime exigency can explain the pace at which Ukraine has addressed this. It cannot explain the origin.

Nor has it closed. On 1 August 2026 the state indexed rail freight tariffs by thirty percent, and the agricultural council's response names the mechanism: cross-subsidisation remains in place, it functions as a tax, and farmers are not paying for services or their cost but for making it cheaper for someone else, while the grain tariff was, on the railway's own account, already profitable.

State aid control is the instrument by which a state constrains its own discretion to direct support toward chosen undertakings. It is the mechanism that would discipline everything described above.

In Ukraine it is suspended. The Ukraine Plan's own word for the required remedy is unsuspension, step 6.9, in the chapter on Management of Public Assets, requiring the full unsuspension of state aid control by the Antimonopoly Committee, due Q4 2025, responsible ministry Economy, Environment and Agriculture. Two carve-outs are written into the step itself, both for the duration of martial law: recapitalisation of systemically important banks holding state-secrecy clearance for defence-industry financing, and aid schemes for small and medium enterprises.

As at July 2026 the step is unfulfilled and the bill that would have delivered it, No. 14345, was withdrawn.

It does not stand alone. The unfulfilled backlog attached to EU money clusters on precisely the instruments that govern routing, the public-service separation that would make state-enterprise subsidy visible, majority-independent supervisory boards, simplified insolvency for micro and small enterprises, non-performing loan resolution, electricity and gas market liberalisation, regulator independence. Nineteen indicators unfulfilled by the July 2026 assessment, with roughly €7.35 billion contingent on them. The instruments that have not been delivered are the ones that would open entrant channels or discipline incumbent routing.

That characterisation is a monitoring consortium's, and the consortium advocates for reform delivery; the €7.35 billion is its aggregation of amounts associated with unfulfilled indicators, not a Commission determination that the money is forfeit. The step text, the deadline, the responsible ministry and the word unsuspension are from the Plan itself.

The clearest instance is not inferred from an outcome. It was said by the minister responsible.

Oleksandr Kravchenko was appointed Minister of Economy and Environment on 16 July 2026, having previously been Managing Partner of McKinsey Ukraine. In an interview recorded that April and published on 29 July, describing how investment should be attracted:

"We need to literally drag some anchor large investors in by the hand."

The reasoning he gives is that anchor entry creates the conditions under which other businesses expand, and he argues openly for a degree of industrial policy rather than pure market allocation, on the grounds that concessional capital is limited. He names the priority sectors: energy, critical and strategic materials including uranium and rare earths, niche machine-building, and the wartime defence industry, selected for export orientation, international offtake, and willingness to sign long-term contracts. Agriculture, the country's largest export earner, is not among them.

This is testimony rather than measurement, recorded before his appointment, and we cite it as the minister's stated position and nothing more. As a statement of position it is the second half of our claim, said aloud by the person who would have to act on it. Where Ukraine chooses the routing, the stated strategy is to route to scale, not as a drift, not as an artefact of instrument design, but as policy with a reason attached.

One coincidence is worth recording without a claim built on it. The ministry responsible for delivering step 6.9, the unsuspension of state aid control, is the ministry Kravchenko heads.

Four instruments, all domestic. A credit programme whose entrant load sits on the same balance sheets that hold three-quarters of the government's securities. A guarantee register that ran eighteen years without a single portfolio guarantee. A subsidy architecture operating inside tariffs where no register can see it, diagnosed in 2015 and unaudited in 2026. And the control mechanism that would discipline all three, suspended, with the bill to restore it withdrawn.

None of that was designed in Brussels. Some of it predates the current war by seven years and the guarantee pattern predates it by eighteen. The donor architecture shapes where most of the money goes and it does not explain any of this, so a piece that stopped at the conditionality would have missed the part Kyiv holds.

And every instrument here has someone on the wrong end of it. A manufacturer who cannot borrow because no portfolio guarantee exists. A shipper paying above cost so a passenger pays less. A small provider outside a package written for institutions. None of them has been refused, and none of them has been told anything at all.

One thing here has a date on it, and it is the only one that does. Step 6.9 came due in Q4 2025 and the bill is withdrawn, while all remaining Ukraine Plan reforms must complete by Q3 2027. Whether state aid control returns before then determines whether any of the routing described in this section becomes subject to a rule, or continues to be determined by instrument design, ticket size and administrative capacity, which is to say, by nobody's decision in particular.

The test this section was supposed to run, and why it does not survive

What does discriminate: the same grain, the same week, two prices

There is a second and narrower diagnostic, and it has the same shape.

Ukraine opened its agricultural land market in July 2021 with ownership caps written specifically to prevent concentration. Since then only 0.7 percent of arable land has been sold, at prices near normative value. The instrument did what it was designed to do.

Meanwhile land rights are increasingly acquired not by buying plots but by acquiring corporate rights in the companies that hold them. Crop-farming mergers and acquisitions ran $167 million across eighteen deals in all of 2025, against more than $525 million across ten-plus deals in the first four months of 2026, with the landmark transaction being a $348 million acquisition by the sector's leading consolidator.

An anti-concentration instrument bypassed by a change of legal route is a routing finding, and the comparator countries do not have it, because they never attempted the intervention.

§5 established that where incumbents are entrenched, subsidising entrants has limited aggregate effect, and named the instruments that fall into that class. They are all here.

The Partial Credit Guarantee Fund for farms below 500 hectares, crop receipts as a collateral instrument for agricultural small and medium enterprises, and grant programmes generally are entrant support layered on top of an existing routing structure. §2 gives the scale asymmetry between them and the direct lending available to a single large agribusiness.

So the agricultural entrant channel is both out-scaled and of a policy class the evidence predicts will underperform. That is a harder thing than we say about entrant instruments anywhere else, and we say it here because this is where the evidence for it sits.

Scale did not protect the largest operator. Kernel, the country's biggest grain exporter and a recipient of first-loss cover from the European investment framework §2 examines, lost roughly 45,000 tonnes of wheat and 9,000 tonnes of sunflower oil and suspended its Chornomorsk terminal operations. The entity every instrument in this piece selected for was stopped by the route.

But the loss still allocates by scale, through a different attribute. With the corridor closed and the harvest arriving, grain has to be held somewhere. Ukraine has roughly 74 million tonnes of storage against a harvest expected near 64 million and a carryover of 14.3, so approximately 78 million tonnes need storage before the corn harvest completes. The Ministry of Agrarian Policy calculates existing capacity could be full by early November.

Whoever holds elevator capacity waits. Whoever does not sells into a nominal bid, or does not sell. Kernel's elevator network took in 4.3 million tonnes last fiscal year, up sixty percent, driven partly by higher intake from third-party farmers, which is what a farm without storage does when it cannot reach a port.

There is a standard financial answer to that problem, which is to borrow against the invoice rather than the building. A survey of five hundred businesses found 1.2 percent using factoring, and export finance at 1.6. The instrument that would let a small producer bridge an unpredictable order book barely exists in Ukraine.

Nobody designed that. The instruments we have traced selected for scale over two decades through profit schedules, guarantee registers, ticket sizes and first-loss cover. A routing failure then allocated its losses by exactly the attribute those instruments had been selecting for.

And it does not stop at this harvest. The All-Ukrainian Agrarian Council states the consequence plainly: if terminals cannot recover quickly, exports collapse, storage overflows, and farmers are left without the working capital to finance the next planting season.

Test eight failed as framed and was rebuilt on something physical.

We lose a test and gain a smaller, sharper claim. Six clean tests plus one narrow diagnostic beats seven tests one of which does not discriminate, and it is the difference between something an agricultural economist nods at and something they take apart in the first paragraph.

What survives is measurable and dated: a premium of roughly twelve percent between two routes for the same grain, with about a fifth of national volume able to reach the better one, and a cost of reaching it that exceeds the premium.

In August the government announced an emergency support package for farmers against the blocking of export routes. We treat announced programmes as stated intent and never as routing until the mechanism is specified, the same discipline we apply to donor announcements, so what it does to who reaches the border is not yet knowable.

What is knowable is the shape of the season. The route that sets the price cannot carry the volume, reaching it costs more than the premium for being there, the state has just made it thirty percent dearer, and storage fills in November. A farm with an elevator waits. A farm without one sells into a bid the desks have stopped writing contracts against, and finances next spring out of whatever it gets.

Healthcare, the route opened, then narrowed by administration

Housing, the routing was right and it hit a wall

What this leaves

The estate grew

Because a drone is a bench and a howitzer is not

What is happening to the estate is not privatisation

Converting a state enterprise into a joint-stock company moves nothing. No plant relocates, no machine is bought. What changes is the legal route into the asset, which creates an entity someone can hold a share of.

Thirty-eight enterprises have been through that conversion. One consequence is already operating: Rheinmetall Ukrainian Defense Industry LLC has run since October 2023, fifty-one percent held by Rheinmetall Landsysteme.

It is the same move traced in three other places in this piece, in a different medium each time. Distributed generation changes where power enters a grid whose wires cannot move. A single purchaser paying providers rather than buildings changes where health money enters an estate that cannot move. Corporatisation changes where capital enters an enterprise that cannot move. The asset stays exactly where it is; the way in is new.

The estate is usually described as about a hundred enterprises, down from the 134 the 2014 conglomerate held. The count is published and it is not the question.

The question is what those enterprises are. How many are producing. How many are cold. How many are held against a strategic requirement nobody has exercised in a decade and may never exercise again.

An estate carried because it is needed and an estate carried because nobody has decided otherwise look identical in a count, and they call for opposite decisions. One is a capability to be resourced. The other is an inheritance to be resolved. Every judgement downstream depends on which it is, and on which enterprise: lease, consolidation, disposal, closure.

Two things are visible from outside, and they may or may not be one thing.

Serial manufacturers now hold contracts large enough that buying a smaller producer is the obvious next move. And producers name unpredictable contracting, not capital, as the constraint that binds them.

The mechanism, if it is running, is ordinary. A firm that cannot see its order book past the current tranche does not build a factory. It builds what it can sell this quarter. When the orders do not come it sells itself to someone larger who can wait. That is how a dispersed industry concentrates without anyone deciding to concentrate it.

There is a standard financial answer to that problem, which is to borrow against the invoice rather than against the building. A survey of five hundred businesses in early 2025 found 1.2 percent of them using factoring. Export finance ran at 1.6. The instrument that would let a small producer bridge an unpredictable order book barely exists in Ukraine, which does not cause the consolidation but does remove the obvious thing that would slow it.

There was a third signal in this section and it is not here now. We had a competition authority declining a defence-sector transaction this year, which would have been the visible confirmation that consolidation pressure had reached the point of an actual deal. It traced to a briefing note citing press reporting, not to a source we hold. Reporting it anyway would have made this passage look stronger and be weaker, so it is out, and the section is thinner for it. Whether the two remaining signals are a pattern turns on transactions below the competition authority's notification threshold, which no public record carries.

The ground is moving underneath the question in any case. Ukraine now builds roughly five times more defence equipment than it can fund, some categories running up to fifty percent beyond what the country can use or finance, and controlled exports of surplus are to be permitted with revenue returned to production. That opens a channel which sorts producers into those with export access and those without, and reaching it turns on export licensing, on end-user requirements set by partner states, on certification against foreign standards, and on having the capital to wait longer for payment. None of that is evenly distributed across a producer base that reached its present size in three years. A firm founded in 2023 with six people and one product has none of it, and the holding, already the largest single producer and already inside a joint venture with a European prime, starts that sort from a different position than a drone firm in its third year.

Four large privatisation and sanctioned-asset auctions are scheduled for October. The practitioner literature describes privatisation in Ukraine as a feeder for subsequent mergers, acquired assets becoming platforms for consolidation. Ukraine runs no separate foreign-investment or national-security screen; a foreign acquisition is assessed through merger control alone.

The estate is not a residue. It is a growing producer of the things that cannot be made on a bench, holding property that is moving to private operators without auction against a list nobody outside can read, inside a holding converting its enterprises into vehicles a foreign partner can take equity in, in a sector where consolidation pressure is building and an export channel is about to open.

None of that is a scandal and none of it is a plan. Every step of it was taken by someone doing their job, a Cabinet simplifying leasing for a war economy, a ministry keeping a register of who may hold defence contracts, a holding company corporatising to attract a partner with capital and technology. What is missing is anyone whose job is the shape all of it makes together.

This read stops here, and the stopping is not modesty. What happens next to the enterprises the money went around, which are resourced, which are converted, which are quietly let go, is not inferable from public sources, and we are not going to pretend we can see it.

The decision still open is who is on the register, and on what basis. It is being made now, enterprise by enterprise, and the people it decides for will find out by whether the lease comes through.

What the two readings add up to

Which is not the criticism it sounds like

What we think happens next, and why we could be wrong about it

Three things this could mean, and we are not sure which

The thing Ukraine has that nobody else does

What is missing is smaller than reform

What this piece cannot tell you

That last one is the most important test in this section, because we lost it. The mechanism is not immovable. It moved in eight months, in the one sector where an organised constituency was reading the instrument rather than the outcome. The problem is not that these rules cannot be changed. It is that reading them takes attention nobody has been asked to give, and the solar association had every incentive to give it. That is the whole difference between what stayed and what moved.

The government's Programme of Activities for 2026 and 2027, submitted on 18 August, is the flagship statement of intent over exactly the window in which the conversion would have to happen, and it does not act on the subparagraph or name the instrument class it belongs to. Its nearest approach is a goal to adapt defence industry to martial-law conditions by 2027, at programme altitude and without the instrument. The test is not failed, martial law has not ended, but the document most able to own the conversion has not.

What would break this outright

§1 names the three findings that would do it, and this section exists to keep them attached to dates rather than to restate them. An instrument that reliably allocates against scale. A structural condition on the defence leg. Dispersion in a sector whose schedule paid against it.

None has appeared. If one does, it belongs here, dated, with what it cost us.

Why we did it

What it is built from