NEW YORK, November 15, 2026 — Shares in companies supplying the equipment, energy systems and construction services behind the artificial-intelligence data-centre boom have suffered a sharp reversal, with a Financial Times analysis reporting that the group has lost about $4.2 trillion in market value since June.
The reported decline has erased nearly half of the $9.4 trillion in market value the companies added during the previous year. The losses have spread beyond chipmakers to industrial and infrastructure businesses that investors had increasingly treated as direct beneficiaries of the rapid expansion of AI computing.
Investors Are Repricing The Suppliers Behind AI
The Financial Times reported on Sunday that the selloff has hit manufacturers, utilities and mining-related businesses connected to data-centre construction. Caterpillar, whose heavy equipment and power-generation products are used in large infrastructure projects, was among the companies facing a substantial valuation decline. The report said Caterpillar’s market value had fallen 44 per cent, while Oklo, Comfort Systems and Corning had each lost about half their value.
The retreat reflects a change in how investors are assessing the AI buildout. During the earlier rally, companies supplying power systems, cooling equipment, construction machinery and materials were rewarded for their exposure to expected data-centre demand. Investors are now asking whether the scale of planned construction, the financing behind it and the eventual revenue from AI services will justify the valuations attached to those suppliers.
Debt-funded spending by major technology companies is one concern. Building a data centre requires large upfront commitments for land, electricity connections, cooling, servers and specialist equipment. The investment can take years to complete, while the revenue expected to repay it depends on customers adopting AI services at sufficient scale and price.
Political resistance is another risk. Local opposition to power-hungry facilities and the prospect of tighter rules in parts of the United States have made project approvals and construction timetables less certain. The FT report said analysts generally viewed a nationwide construction ban as unlikely, but projects could be delayed or redirected towards jurisdictions more willing to accommodate them.
Strong Orders Show That Demand Has Not Disappeared
Recent company results provide a counterpoint to the share-price weakness. Caterpillar reported third-quarter 2026 sales and revenue of $20.7 billion, up 17 per cent from a year earlier. Its Power & Energy segment recorded sales of $8.784 billion, an 18 per cent increase, with the company identifying large reciprocating engines and turbines used in data-centre applications as important contributors.
Siemens has also reported strong momentum. Reuters said the German industrial group’s data-centre orders doubled in its latest quarter, while its backlog included €7 billion in data-centre products. Siemens reported record quarterly orders and raised its medium-term revenue growth outlook, citing demand that included equipment for AI-related facilities.
Power-management company Eaton reported third-quarter sales of $8.9 billion, up 28 per cent year over year. Its results showed accelerating orders and substantial backlog growth in its electrical businesses. Eaton also announced an agreement to acquire COL Group, saying the move would expand manufacturing capacity and power-distribution capabilities for data-centre and utility markets in Europe.
Those results suggest that investment in power, cooling and construction infrastructure remains significant. They do not, however, guarantee that every announced project will proceed on time or that every supplier will convert its order book into profitable sales. Backlogs can be delayed, changed or cancelled, and some customers may place overlapping orders while competing for scarce capacity.
Capacity Delays And Financing Are Becoming Key Tests
The risks are visible in the experience of AI cloud provider Nebius. Reuters reported on November 12 that the company reduced its 2026 revenue forecast after delays in third-party data-centre capacity pushed some expected openings into the first half of 2027. Nebius said demand remained strong and that its own facilities were on schedule or ahead, but the delay showed how infrastructure bottlenecks can affect revenue even when customers are seeking computing capacity.
Wall Street research firms have offered a more measured interpretation of the recent market decline. Reuters reported on November 9 that several brokerages considered the selloff excessive and expected some projects to move further into the future rather than be abandoned. That view does not remove the risks; it suggests the market may be distinguishing between companies with strong customers and financing, and more speculative developers whose plans depend on continued access to debt.
HeadlinesXtra Analysis
The $4.2 trillion figure is a loss in market value across a group of listed companies, not evidence that $4.2 trillion of data-centre projects have been cancelled or that demand for AI has vanished. The distinction matters. Stock prices reflect expectations about future earnings, financing costs and risk, and those expectations can change much faster than construction activity.
The industry is facing a transition from enthusiasm about the size of the AI opportunity to closer scrutiny of who can earn a reliable return from it. Companies with proven orders, strong balance sheets, access to power equipment and the ability to deliver projects may remain well positioned. Businesses that rely on expensive borrowing, uncertain customer commitments or delayed grid connections face a tougher test.
For investors, the next phase is likely to be less about buying every company associated with AI and more about separating durable infrastructure demand from inflated expectations. For technology companies and communities, the central challenge remains practical: turning enormous computing ambitions into facilities that can be powered, financed, approved and operated economically.
Sources And Further Reading
- Financial Times — Data-centre construction stocks lose $4.2tn as AI trade unravels
- Reuters — Siemens raises mid-term outlook as AI-driven data-centre demand boosts orders
- Caterpillar — Third-quarter 2026 results
- Eaton — Record third-quarter 2026 results and order growth
- Reuters — Nebius cuts 2026 revenue forecast after data-centre capacity delays
- Reuters — Wall Street brokerages say AI selloff is overdone
- Helix Digital Infrastructure — Samsung’s $1 billion commitment to AI infrastructure expansion