Who wins federal industrial gas manufacturing contracts

Of the $210M awarded across these 448 contracts, the five largest recipients hold 45.2% and the largest single one holds 12.1%. Widen it to ten firms and the share is 61.9%. By count rather than dollars, the top five took 43.1% of the awards.

49 of the 98 firms here — 50% — appear exactly once. That is the closest thing this data has to a “room at the bottom” reading: a slice where most winners appear once is one you can enter without displacing an incumbent, and a slice where they don’t is one where relationships are already set.

• Largest single recipient12.1% of dollars
• Top 5 recipients45.2% of dollars
• Top 10 recipients61.9% of dollars
• Everyone else38.1% of dollars

Honest about coverage

These pages count 351,812 federal contract awards pulled from USASpending.gov. Every one is a real, published award to a named winner. They are not all of federal spending, and reading them as if they were would mislead you.

The fetch asks USASpending for awards between $25,000 and $2,000,000 — the band a small business actually competes in. It reads them highest-value first, and it has not yet reached the bottom of that band. The smallest award currently in the dataset is $135,957. So what you are looking at is a mid-size slice: 351,812 awards between $135,957 and $2,000,000. It is not a representative sample of federal contracting, and it is emphatically not a sample of small awards.

What that means for the numbers on these pages

• Counts and shares are exact for this dataset. “The top five recipients hold 24% of the dollars” means 24% of the dollars in these 351,812 awards — not 24% of the sector.
• The medians and size bands describe this slice only. You will not find an “average federal contract value” anywhere on this site, because the distribution here is cut off at the bottom and that number does not exist in this data.
• A firm missing from a page has not necessarily stopped winning work — it may simply not have an award inside this band and these windows. Absence is not evidence here.
• We publish no “still active” or “gone quiet” signal. The data arrives in two fetch windows (2023-01-01 to 2025-07-25 and 2025-07-26 to 2026-07-25) that have reached different depths ($441,837 and $135,957), and the period-of-performance start date can predate either of them, so any recency claim would be measuring our own backfill rather than the market.

The backfill is still running, so the floor moves down over time and these pages are regenerated from it. Figures as of 2026-08-03. Always confirm an award on USASpending.gov before you act on it.