In the twelve months to 30 July 2026, the twenty largest declared corporate parents took 55.4% of $2.2 trillion in US federal contract value. Across the record, 23,114 supplier names resolve to 15,578 parent organisations — and beneath 5,946 visible primes sit 103,444 subcontractors most risk teams have never listed. Your supplier list is longer than your supplier base.
Every chief risk officer has a supplier list. Very few have a supplier base — because the list counts names, and names are not companies. One of the most consequential facts in supply-chain risk is also one of the least examined: the mapping from the suppliers you contract with to the corporations that actually control them.
We maintain 83,137,620 public contract awards and 31,046,277 tenders — 114.2 million procurement records — normalised from official government sources worldwide into a single engine. Buried inside the US federal portion of that record is a field most people have never queried: the parent organisation each supplier itself declares in its official registration. It appears on 23,081,988 award records. Read at scale, it dismantles the idea that a long supplier list means a diversified one.
Across the US federal record, 7,899,348 awards went to a supplier whose declared parent is a different company. Collapse those: 23,114 distinct supplier names resolve to 15,578 declared parent organisations. On average that sounds mild. At the top of the tree, it is not mild at all.
| Declared parent | Distinct supplier names | Award records |
|---|---|---|
| Waste Management, Inc. | 78 | 324 |
| RTX Corp | 72 | 31,594 |
| Genesis Healthcare, Inc. | 72 | 586 |
| Danaher Corporation | 61 | 27,639 |
| L3Harris Technologies, Inc. | 59 | 11,783 |
| General Dynamics Corp | 57 | 10,003 |
| Republic Services Inc | 56 | 384 |
| TransDigm Group Incorporated | 55 | 4,798 |
| AECOM | 55 | 1,463 |
| Berkshire Hathaway Inc. | 54 | 5,956 |
| AMETEK Inc | 52 | 1,270 |
RTX sells to the US government under 72 different supplier names. TransDigm — a company built by acquiring niche aerospace-parts makers — appears under 55. A risk team that lists five of those names as five suppliers believes it has five options. It has one counterparty, five times.
This matters beyond aerospace. Danaher's 61 names span diagnostics, life sciences and environmental instruments. Waste Management and Republic Services together hold 134 supplier names in waste services — a category where a municipality's "competitive" bidder field can be two companies wearing several dozen coats. Genesis Healthcare's 72 names sit in nursing and care facilities. The pattern is structural: roll-up corporations acquire, keep the acquired name trading, and the customer's vendor master quietly loses its meaning.
"A risk team that lists five of those names as five suppliers believes it has five options. It has one counterparty, five times."
Concentration by name-count is one lens. Concentration by money is starker. In the twelve months to 30 July 2026, US federal awards carrying a declared parent totalled $2,205.7 billion in potential contract value (base plus all options — the ceiling basis the US government itself reports) across 1,242,142 awards and 47,589 declared parent names.
| Parent group | Value | Share of total |
|---|---|---|
| Top 5 parents | $797.4bn | 36.2% |
| Top 20 parents | $1,221.9bn | 55.4% |
| Top 100 parents | $1,654.9bn | 75.0% |
| All 47,589 parents | $2,205.7bn | 100% |
Twenty corporate families took more than half of everything. One hundred took three quarters. And because ceiling values can flatter, we re-ran the same census on obligated dollars — money actually committed, not contract ceilings. The answer barely moves: the top 20 parents hold 55.2% of $1,654.7 billion in obligations. The concentration is not an artefact of how you measure value. It is the shape of the market.
When a concentration figure holds within 0.2 percentage points across two independent value bases, it is not a quirk of accounting. Twenty corporate families anchor half the addressable US federal market, however you count the dollars.
One more honest turn of the screw: these figures understate family-level concentration. The declared-parent field is a name string, and names drift. Lockheed Martin appears in the twelve-month top 20 twice — as "Lockheed Martin Corp" ($382.1bn) and "Lockheed Martin Corporation" ($41.5bn). Electric Boat ($115.6bn), Bath Iron Works ($21.4bn) and National Steel and Shipbuilding ($15.5bn) each appear as their own parents; all three are shipyard units of General Dynamics. Merge the variants and the true top-20 share rises above 55.4%. We report the unmerged number because it is the one that requires no judgement — the conservative floor.
Ownership is the illusion above the waterline. Below it sits the tier most risk frameworks acknowledge and almost none can see: the subcontractors your suppliers depend on.
Here is the fact that should change how CROs think about tier-2 visibility: it is already public. US federal law requires primes to disclose subawards, and the record now holds 1,345,211 subaward records in our spine. They reveal 5,946 prime contractors sitting on top of 103,444 distinct subcontractors — seventeen companies below the waterline for every one above it. Europe's TED system adds a further 2,601,161 award records carrying subcontracting disclosures.
The tier-2 network most organisations try to map with supplier questionnaires — slowly, voluntarily, and once a year — already exists as disclosed public data for any supplier that touches government work. Nobody has to ask permission to see it.
The standard tier-2 playbook is a survey: ask your suppliers who their suppliers are, wait a quarter, get partial answers, repeat next year. The public record inverts that. A prime's subcontractor network is disclosed, dated, and named — not for every company, but for the large population that sells to governments, which in categories like defence, infrastructure, health and IT services means most of the companies that matter.
Concentration diagnosis without a substitution plan is just better-documented anxiety. So we ran the play a CRO would actually need to run, end to end, on a real company.
Take Transaero, Inc. — a New-York-based aerospace parts supplier. In the twelve months to 30 July 2026 the US federal record shows it winning 62 awards worth $15.44 million across 4 government buyers in aircraft parts and auxiliary equipment (NAICS 336413). Suppose it is your critical supplier and you need alternatives — companies demonstrably active in the same category, at a comparable scale, today.
The category is deep: 2,783 distinct suppliers won 26,133 awards in that NAICS code in the same twelve months. Filter deterministically — same category, US federal buyers, twelve-month footprint between 0.5× and 2× of Transaero's — and the record returns 143 candidate substitutes, each with its award count, value, buyer spread, and declared parent attached.
| Candidate supplier | Awards | Value | Buyers | Declared parent |
|---|---|---|---|---|
| Top Flight Aerostructures, Inc. | 50 | $30.7m | 4 | independent |
| Triman Industries Inc | 177 | $29.7m | 7 | independent |
| Meggitt Aircraft Braking Systems | 98 | $29.1m | 4 | Parker-Hannifin Corporation |
| Northrop Grumman Systems Corp | 6 | $28.4m | 1 | Northrop Grumman Corporation |
| Integrated Procurement Technologies | 158 | $27.1m | 8 | independent |
| Malones CNC Machining Inc | 229 | $26.5m | 4 | independent |
| Rotair Aerospace Corp | 29 | $26.0m | 5 | independent |
| Aero-Glen International LLC | 19 | $25.7m | 5 | independent |
| Columbia Helicopters, Inc. | 3 | $25.1m | 2 | independent |
The ownership column is where the illusion gets caught in the act. Of the 143 candidates, 30 — one in five — declare a parent different from their own name. Meggitt Aircraft Braking Systems looks, by name, like a perfect independent alternative; its declared parent is Parker-Hannifin. A risk team "diversifying" from Parker into Meggitt has diversified nothing. Without the ownership layer, one substitute in five in this category is a mirage — and the mirages are invisible at exactly the moment you are moving fast, because supplier failure is when substitution decisions get made in days, not quarters.
Every step of that census — footprint, band, category, ownership screen — is a deterministic query over public records. No survey, no questionnaire, no supplier self-reporting. The same play runs for any category the public record covers, from aircraft brakes to food distribution to cloud services.
Measure your real counterparty count. Resolve your vendor master against declared and mapped ownership, and count parents, not names. If your 400 suppliers resolve to 300 parents, you should know that number the way you know your VaR.
Read concentration at the family level. Category concentration analyses run on supplier names systematically understate exposure — in US federal aerospace and waste categories, as shown above, by design of the corporate structures involved.
Mine the disclosed tier 2. Before commissioning another supplier survey, check how much of your supply base already discloses its subcontracting in public procurement records. For government-facing suppliers, the network is largely on the record.
Pre-build the substitute bench. An ownership-screened substitute census, kept current, converts supplier failure from a crisis into a decision. The worst time to discover your backup shares a parent with your incumbent is the week you need the backup.
Universe: Valan's procurement spine — 83,137,620 public contract awards and 31,046,277 tenders from official government and multilateral sources, normalised into a single DuckDB engine, live as at 30 July 2026. Concentration figures use the US federal record, where 23,081,988 award rows carry the parent organisation the supplier itself declares in official registration data (USAspending recipient parent). We report that field as declared — including its imperfections, such as name variants and self-declared subsidiaries — because it is source-side truth requiring no inference. Where variants exist, our figures are conservative floors.
Twelve-month values are stated on two bases and labelled: potential value (base plus all options — the ceiling basis reported by the source) and obligated value (committed dollars). The top-20 concentration finding holds on both (55.4% and 55.2% respectively). Subcontracting figures count disclosed subaward records (US, 1,345,211) and awards carrying subcontracting disclosures (TED, 2,601,161); TED's subcontract-percentage field is populated on fewer than 4,000 records and contains implausible outliers, so we cite disclosure counts only and no percentage statistics from it.
Supplier identity and substitution matching is deterministic — exact and normalised identifiers and category codes only, never similarity-guessing. Where an award links to a listed supplier or listed ultimate parent, it carries that entity's ticker as of the award date: 2,801,648 awards are ticker-linked, 9,394,136 carry a listed ultimate-parent ticker, and 2,736,905 carry the supplier's or parent's point-in-time market capitalisation as of the award date. Where a link cannot be confirmed against a real dated disclosure window, the field is null — honestly null, never a current value dressed up as history. All figures in this article were computed read-only against the live engine on 30 July 2026.
This is not a claim that concentration is bad, or that roll-ups are villains. Consolidation often makes suppliers stronger. It is a claim about visibility: that the gap between a supplier list and a supplier base is measurable, that the tier below your suppliers is more public than your questionnaires assume, and that a substitute bench can be censused from the record before the phone call that makes it urgent.
The data to close all three gaps has been published for years, by governments, as a matter of law. What was missing was the assembly — one engine, resolved identities, dated records, and the discipline to leave a field null rather than guess. That last part is the point: a risk number you cannot trace is not a risk number. Every figure in this article traces to a query against the public record, run on 30 July 2026, and we will happily show our working.
New research from Valan's 114M-record procurement dataset, delivered as it publishes. No noise.