Air Products wrote off $6.6bn of what it built

Air Products sells oxygen, nitrogen and hydrogen through pipelines, mostly on contracts that run fifteen years or longer. It is a ninety-year-old industrial gas business and, for most of its life, one of the least surprising companies on the New York Stock Exchange.

In the June 2026 quarter it reported an operating loss of $2,097.1 million on sales of $3,161.0 million — a margin of negative 66%.

Nothing went wrong in the quarter. What went wrong had been going wrong for four years, and the accounting caught up with it all at once.

What was spent

Between fiscal 2019 and fiscal 2025, Air Products more than tripled its capital spending. Over the same seven years, revenue rose 35% and then stopped.

Read the first two columns together and the shape of the problem is immediate.

Revenue peaked in fiscal 2022 and has fallen every year since. Capital spending in that same period went from $2.9bn to $7.0bn. By fiscal 2025 the company was spending 58% of its revenue on building, against operating cash flow of $3.3bn — a free cash outflow of $3.8 billion in one year.

This is the pattern we screened the whole market for last month — companies spending more on property than they collect from operations. Air Products is that pattern in an unusually pure form, because it can be traced to specific projects with names.

What it was spent on

The build was a bet on hydrogen. A low-carbon hydrogen and ammonia complex in Louisiana. A green hydrogen plant in Casa Grande, Arizona. A sustainable aviation fuel expansion. A multi-billion dollar green ammonia project in NEOM City, Saudi Arabia.

On 26 June 2026 the board and chief executive concluded a review, and the company filed a Form 8-K under Item 2.06 — Material Impairments, an item most large companies go years without using.

The stated reasons are worth quoting rather than paraphrasing. Louisiana was cancelled because “the expected financial returns from the project would not meet its return criteria.” The others were exited because of “challenging commercial conditions, project-specific economic factors, and slower than expected development in certain markets, largely hydrogen for mobility.”

That last clause is the whole thesis in six words. The plants were built for hydrogen-powered transport demand that did not arrive on schedule.

What it cost

The 8-K announced a pre-tax charge of up to $2.9 billion. The 10-Q filed a month later shows what was actually taken — and shows that this was the second wave, not the first.

FY2026 exits   FY2025 exits      Total
Asset write-downs          $2,210.6      $3,307.8     $5,518.4
Other exit costs              696.8         350.6      1,047.4
───────────   ───────────   ──────────
$2,907.4      $3,658.4     $6,565.8

$6.57 billion. Set against the $28.3 billion of capital expenditure across fiscal 2019 to 2025, the exit charges equal 23% of everything the company spent building over seven years.

That is a ratio, not a causal claim — some of the written-down assets predate the period, and not all of that capex went into the cancelled projects. But it is the right order of magnitude for what happened.

The part that is not a cash cost

Here is where a write-down is widely misread.

The June-quarter charge did not take $2.9 billion out of the bank. The money left in earlier years, as concrete and steel. What the charge does is admit that the asset those payments created is not worth what the balance sheet said.

The 8-K puts a number on the actual cash still to go: not more than $925 million, for terminating contractual commitments and settling obligations, with the company expecting less once negotiations conclude. And in the nine-month cash flow statement, the entire $2,929.4 million appears as a non-cash add-back.

Pre-tax charge, FY2026 exits            $2,907.4M
Estimated remaining cash outflow        ≤  $925.0M
──────────
Already spent, now recognised as lost   ≈$1,982.4M

That gap between the charge and the cash is not unique to this company, and it is the reason a restructuring headline is a poor guide to what a plan costs. Across 29 such filings from the largest US companies in 2026, the non-cash share ranges from nothing to 84% — and a $400 million charge at one company drains less cash than a $142 million charge at another.

The business underneath

Strip the charge out and something surprising appears: the operating business had a good quarter.

Sales rose 5%. Adjusted operating income rose 9% and the adjusted margin improved 110 basis points. Equity affiliates’ income rose 22%.

Selling industrial gas was never the problem. We normally treat adjusted figures with suspicion, because “adjusted” is where inconvenient costs go to be excluded. Here the adjustment is the honest read in one direction and the dishonest one in another: the operating business is genuinely earning a 25% margin, and $6.6bn of shareholder capital is genuinely gone.

Two nine-month periods, the same line

The detail that makes this more than a bad quarter is that it has now happened twice.

FY2026        FY2025
Sales                          $9,435.3      $8,870.4
Business and asset actions      2,929.4       2,952.0
Operating income (loss)          (609.9)       (893.8)
Net income (loss)                 (52.2)       (399.4)

In two consecutive fiscal years, a company with a 25% operating margin in its base business reported an operating loss across three quarters, because the same line item consumed all of it. Fiscal 2025 closed with a full-year net loss of $395 million, against $3,828 million of net income the year before.

What is still on the books

Air Products has not exited hydrogen. It has exited hydrogen in North America.

The surviving bet is NEOM Green Hydrogen Company, a joint venture in Saudi Arabia with ACWA Power and NEOM Company. Air Products holds one third of the voting interests, and consolidates it anyway — it is the primary beneficiary of a variable interest entity because it can unilaterally direct key design and construction decisions.

Both balances grew during the nine months in which the North American projects were being cancelled.

And the offtake structure is the thing to read carefully. Air Products is the exclusive offtaker of NEOM’s green ammonia under a long-term take-if-tendered agreement. If the plant produces and tenders, Air Products buys.

We are not saying that is a bad arrangement. We are saying the balance sheet now assumes a demand curve for green ammonia, in the same way it previously assumed one for hydrogen-for-mobility, and that the first assumption cost $6.6 billion when it did not hold.

Why this matters beyond one company

Every article we have written this month about the AI build-out describes the same manoeuvre: raising capital to build capacity ahead of the demand that would justify it. Alphabet borrowed $25bn for forty years. CoreWeave borrowed at SOFR plus 5.50% against GPU servers. The bet is that the demand shows up.

Air Products made that bet in a different sector, four years earlier, with a longer asset life and a better balance sheet than most of the companies now doing it. The filings show precisely what it looks like on the other side: the cash goes out first, the revenue does not arrive, and the recognition comes years later in a single line that erases three quarters of operating profit.

Nothing here predicts that outcome for anyone else. It documents the shape of it.

What would make this wrong

• The exit charges may not be final. The 8-K says estimated contract cancellation and other project cancellation costs “are subject to further refinement and may ultimately differ materially from actual costs recorded in the Company’s fiscal 2026 third quarter and beyond.” The FY2025 wave was already revised upward by $28.3m in the following quarter.
• The 23% ratio is a comparison, not an accounting identity. Some written-down assets were built before fiscal 2019, and much of the seven years’ capex went into plants that are running normally.
• We have not valued the remaining assets. $25.3bn of net property is on the books. Whether the rest of it earns its carrying value is a question the filings pose and do not answer.
• NEOM is not evidence of anything yet. It is under construction. Consolidating $8.1bn of VIE assets is an accounting consequence of control, not a judgement about the project.
• Adjusted figures are the company’s own. Adjusted operating income excludes exactly the charge this article is about. We quote it because the contrast is the point, not because it is the better number.
• This is one company in industrial gas. The comparison to the AI build-out is an analogy about capital committed ahead of demand. Different assets, different contracts, different customers.

Check it yourself

The 8-K and the 10-Q are linked below and neither is behind a paywall. The revenue, capex, operating cash flow and net property series come from the SEC’s XBRL company facts endpoint for CIK 2969 and can be pulled in one request. Note 4 of the 10-Q contains the $6,565.8m table exactly as reproduced here; Note 3 contains the NEOM balances.

If a figure on this page does not match a filing, tell us and it will be corrected on the article and on the corrections log, with the date.