Data cutoff: October 16, 2026, 8:45 p.m. ET.
The part of AI spending that cannot be postponed by software
One of the most revealing earnings announcements on October 16 came from a company listed in Paris rather than on the Nasdaq. Schneider Electric reported quarterly revenue of €11.460 billion, up 16.9% organically and 17.9% on a reported basis, a company record. It raised its full-year outlook again after saying data-center demand grew at a triple-digit rate during the quarter. The numbers provide a direct test of a question confronting investors during Friday's broad AI selloff: is the physical infrastructure needed to support computing still expanding, even as financing markets become less comfortable with the industry's debt?
The answer from Schneider's order book is that demand remained strong. It is not evidence that every proposed data center will be built, every neocloud will survive, or every stock supplying the sector deserves its current multiple. But it is more substantial than a vague claim about AI interest. Switchgear, electrical distribution, cooling and control equipment must be purchased, configured and installed before a facility can operate. Supplier revenue and order execution give investors a measurable view of one layer of the buildout.
The exact guide: adjusted EBITA, not EBITDA
Schneider now targets 2026 organic adjusted EBITA growth of 18%–21%, up from its prior 14%–19% range. It expects organic revenue growth of 12%–14%, versus 10%–13% previously, together with organic adjusted EBITA margin expansion of 100–120 basis points. The company described an implied adjusted EBITA margin around 19.7%–19.9%, incorporating its current estimates of currency and completed transactions.
The accounting label matters. EBITA excludes amortization but includes depreciation, whereas EBITDA excludes both. Replacing the company's adjusted EBITA target with an EBITDA target changes the metric being forecast. Likewise, the 16.9% quarterly growth rate is organic growth, while the 17.9% rate is reported growth; the two should not be used interchangeably. The revenue figure is in euros, not U.S. dollars.
The group broke out useful operating details. Energy Management, the larger division, grew about 19% organically; Industrial Automation grew about 8%. China and East Asia expanded 26% organically, while North America rose 20%. Products accounted for roughly 48% of quarterly revenue, systems 35%, and software and services 17%. Systems grew faster than products, reflecting the greater role of integrated projects, while product pricing contributed alongside volume.
Why this matters to U.S.-listed AI infrastructure companies
Schneider is not the same kind of investment as Nvidia, Broadcom or a GPU rental company. It sells many components that become necessary regardless of which accelerator model wins a particular workload. A hyperscaler that changes its preferred AI chip still needs safe electrical distribution, cooling, monitoring and backup capacity for any site it operates. That gives the supplier a different form of exposure to computing investment.
It also means there are limits to the comparison. Some Schneider sales serve factories, buildings, utilities and conventional industrial customers, not AI facilities. The company said the data-center business was the main driver, but its growth was also geographically broad. An investor cannot multiply the group revenue figure by a guessed AI share to obtain a reliable data-center dollar market size.
The most defensible read-through is to other electrical equipment, thermal-management and grid-connection businesses: projects that are approved and funded can generate a multiyear equipment cycle, often with a backlog that extends beyond the software product cycle. Yet backlogs still require installation, labor and delivered power before final value is realized.
The distinction between customer orders and customer returns
Schneider's result and CoreWeave's credit selloff can both be true. An equipment supplier may receive orders and recognize revenue while the eventual owner of the completed data center faces high interest costs, refinancing risk and uncertain lifetime returns. A healthy supplier order book does not solve a client's cost-of-capital problem. Conversely, weakness in one heavily indebted lessor does not automatically cancel all hyperscaler construction programs.
The debate therefore shifts from whether AI equipment demand exists to which participant captures the economics. Selling a power distribution system can generate cash on a different schedule and with different financial risks from borrowing to own an entire campus and rent its capacity over ten years. Investors who group both businesses under one undifferentiated 'AI infrastructure' multiple may misunderstand the risk they are buying.
There is a cyclical caveat. If the funding environment stays hostile long enough, a project owner can postpone construction, reduce orders or negotiate harder on price. Schneider's current demand visibility into 2027 and 2028 is not a permanent insulation from customers' capital budgets.
Where the outlook could go wrong
The company itself noted that disruption in the Middle East could increase supply-chain pressure and inflation in the fourth quarter. Raw material costs, tariffs, labor availability, shipping delays and equipment installation bottlenecks all matter. Strong demand does not guarantee unlimited margin expansion, especially when customers demand timely delivery on complex systems. A shift from high-margin products to larger project systems can also change profitability even while consolidated revenue accelerates.
Another investor concern is capital allocation. Schneider recently announced a large agreement to acquire industrial-software company PTC, which has its own purchase price, financing structure and integration requirements. The October 16 earnings release is a new operating event, not a replay of that earlier acquisition headline. Investors should judge the raised guidance separately from assumptions about hypothetical software synergies.
The bear case is that data-center orders eventually decelerate as cloud customers reevaluate AI spending; the bull case is that power constraints make the installed infrastructure more valuable and extend the buildout. Neither thesis should be mistaken for a demonstrated outcome at the October 16 cutoff.
Dates and operating evidence to follow
Schneider's official calendar places its full-year 2026 results on February 18, 2027. Until then, follow management's guidance revisions, price-versus-volume contribution, backlog conversion, product-versus-systems mix and the share of growth coming from different regions. In fourth-quarter commentary, the most useful questions will concern whether data-center demand remains unusually strong and whether electrical equipment inflation offsets productivity benefits.
For U.S. stocks, cross-check those signals against actual hyperscaler capital-spending results, grid interconnection progress and balance-sheet funding. A small improvement in chip sentiment cannot substitute for evidence that sites are becoming operable and customers can earn returns on them.
Conclusion: demand is real, but the distribution of profit is unsettled
Schneider's €11.46 billion quarter and upgraded adjusted EBITA target are concrete evidence of strong demand for physical infrastructure. They are especially informative during an AI-stock selloff because they show one supplier still converting orders into revenue. The facts do not settle how much profit cloud operators, financing vehicles or semiconductor owners ultimately keep.
That is the useful dividing line for investors: strong sector spending and attractive security returns are different propositions. Schneider's execution supports the first. The second requires analysis of price, margins, customer concentration, financing and the durability of the contracts behind each business model.