Not All Capital Spending Creates Value — The Materials CapEx Trap

You've probably seen the headlines: materials companies are spending more than ever on new plants, new capacity, new infrastructure. The natural instinct is to assume this means growth ahead — that rising capital expenditure is a vote of confidence in future demand.

But capital expenditure is not a prediction. It is a bet. And the difference between Linde's CapEx and Air Products' CapEx this year is the difference between a company building where customers have already committed to buy, and a company building where the market has not yet arrived.

The spending surge is real — but look at where the money goes

The United States is in the middle of a capital spending cycle that dwarfs anything since the 2000s. Private nonresidential fixed investment has climbed past $4.2 trillion annually — roughly 14% of GDP. Announced plans surged to approximately $8.8 trillion in 2025, driven by AI data centers, semiconductor fabs, energy infrastructure, and reshoring. Materials companies sit at the center of it, supplying the industrial gases, specialty chemicals, and infrastructure that every one of these projects needs.

The catch is that not all capital spending creates value. Some of it is building toll roads on roads nobody is driving.

The two paths

Linde, the world's largest industrial gas producer by revenue, announced a $1 billion investment in July 2026 to expand its gas supply complex in Phoenix, Arizona. The complex will feed two new semiconductor fabrication facilities. A few days later, Linde's Taiwan joint venture committed another $800 million for the packaging expansion. Nearly $1.8 billion in new capital — but with long-term supply contracts locking in decades of revenue.

Linde's industrial gas business works on a toll model. The company builds gas-production plants next to its customers' facilities, and those customers sign long-term, take-or-pay contracts to purchase the gas. The customer needs the gas to operate. If the customer does not take it, they still pay. Linde owns the asset, charges for the service, and the cash flow is as close to contractual certainty as it gets in the real economy.

Linde's Q2 2026 results show this model at work. Sales rose 9% to $9.29 billion. Electronics — the semiconductor and data center end markets — grew 18% year-over-year, the fastest-growing segment. Record sales-of-gas project backlog reached $8.1 billion, meaning more contracted revenue is already sitting in the pipeline waiting to start flowing. Free cash flow over the trailing twelve months stands at $5.0 billion. The company generates roughly $10.5 billion in operating cash flow against $5.5 billion in capital spending. The dividend payout ratio sits at 40%, leaving ample room to fund continued investment while growing the payout.

Air Products, the third-largest industrial gas producer, followed a different map.

Over the past 18 months, Air Products has exited at least six major projects. In February 2025, the company cancelled a green hydrogen facility in New York, a sustainable aviation fuel expansion in California, and a carbon monoxide project in Texas — recording a $3.1 billion pre-tax charge. In June 2026, it abandoned the Louisiana Clean Energy Complex, a blue hydrogen plant that would have produced 1,700 metric tons of hydrogen per day, along with a zero-carbon hydrogen facility in Arizona. Another $2.9 billion pre-tax charge.

That is $6 billion in charges from cancelled projects in less than two years. The Louisiana complex alone represented an investment pipeline of up to $9 billion that never reached the ground.

The reason Air Products gives is blunt: the financial returns did not meet the company's criteria. For hydrogen-for-mobility projects, the market developed slower than expected. For the New York facility, regulatory changes made the power supply ineligible for the Clean Hydrogen Production Tax Credit. For Louisiana, there was simply no commercial demand — a buyer deal with fertilizer maker Yara that could have saved the project fell through.

What the financial statements tell you

The numbers make the contrast harder to ignore.

Air Products spent $4.9 billion on capital expenditures over the trailing twelve months — a massive burn for a company generating $4.6 billion in operating cash flow. Free cash flow is negative: -$301 million. The company's total debt stands at $23.9 billion against $16.6 billion in equity. The dividend payout ratio has climbed to 76%, which means nearly everything the operating business earns is already promised to shareholders.

Compare that to Linde: $5.5 billion in capital spending against $10.5 billion in operating cash flow. Free cash flow of $5.0 billion. Debt-to-equity of 0.69 versus Air Products' 1.06. And that 40% payout ratio means Linde funds its dividend on less than half its cash flow, with the rest available for reinvestment or debt reduction.

The market sees the difference. Linde trades at a market capitalization of $224 billion. Air Products sits at $67.9 billion — less than a third of Linde's size, despite being a peer in the same industry with similar technology and geographic reach.

The underlying economics

Industrial gas is one of the most durable businesses in the real economy. Semiconductor fabs need ultra-high-purity nitrogen, oxygen, and argon to operate 24 hours a day, 365 days a year. You cannot build a chip without them. The gas is not optional. That gives the supplier pricing power — the ability to raise prices without losing customers — because the alternative to paying Linde for gas is shutting down a $20 billion fabrication plant.

This is the filter that matters: can the company raise prices through inflation and through cycles without losing demand? Linde passes. The gas is mission-critical, the contracts are long-term and indexed, and the customer cannot substitute or bypass the supplier.

The question is not whether industrial gases are a good business. The question is whether the capital spending behind them is creating assets that earn, or writing checks for stranded capacity.

The timing test

Here is where the leading indicators help separate conviction from exposure.

Semiconductor capital spending remains robust. The five largest hyperscalers committed close to $700 billion in combined CapEx for 2026 alone, and semiconductor fabs are the physical anchor of that buildout. Linde's $1.8 billion commitment in Phoenix and Taiwan is small relative to the fabs it feeds, but it guarantees a share of the cash flow for decades. The demand is not a forecast — it is a contract.

Green hydrogen is another story. The U.S. market is characterized by a fundamental chicken-and-egg problem: suppliers will not build without demand, and customers will not invest without supply. Federal subsidies that were supposed to bridge the gap were rescinded or restricted. Hydrogen-for-mobility has developed slower than almost anyone projected. Air Products is not the only victim — ExxonMobil paused a large blue hydrogen project in Texas in December 2025 — but the pattern is clear. Capital spent before the market arrives becomes capital that cannot be recovered.

What this means for your judgment

When you read about a materials company increasing its capital spending, the first question is not "how much?" The first question is "what is it spending on, and who guarantees they will pay for it?"

Linde's spending is concentrated in semiconductor and electronics gas supply — businesses where the customer has signed a long-term contract before Linde breaks ground. The capital creates an asset that earns from day one. The free cash flow covers the dividend four times over. The balance sheet supports continued investment without stretching.

Air Products' spending was concentrated in green energy infrastructure — hydrogen plants and ammonia facilities — where the customer base had not yet materialized and the economics depended on subsidies that may or may not exist. The capital created write-downs. Free cash flow is negative. The payout ratio is near the ceiling.

Neither company is inherently a bad business. Industrial gases are mission-critical in both cases. But the quality of capital allocation — the discipline to spend only where the revenue is certain — is what separates a durable dividend grower from a company that is trying to catch a market that has not arrived.

The lesson is not that rising CapEx is bad. The lesson is that rising CapEx is only a signal when you can see what it is building, who will pay for it, and whether the free cash flow will cover both the spending and the dividend. Most headline lists of "biggest CapEx increases" stop at the first question. The other two are the ones that matter.