The Biggest AI Warning Came from the Company That Did Not Report

The Biggest AI Warning Came from the Company That Did Not Report

The latest earnings provided no simple answer to the question hanging over the artificial-intelligence trade. Microsoft offered the strongest evidence that demand remains real: quarterly revenue rose 18% to $90 billion, operating income reached $40.6 billion and Azure growth accelerated to 43% in constant currency. Meta’s advertising engine also remained powerful, with revenue rising 28% to $60.8 billion, but operating margin fell from 43% to 31% as costs and investment surged. Qualcomm, meanwhile, showed encouraging growth in automotive and edge computing, but continued to be restrained by weaker handset demand. The AI earnings were mixed rather than disastrous.

Yet perhaps the most revealing market signal came from a company that did not report at all.

Caterpillar’s second-quarter results are not due until 4 August. Nevertheless, its shares fell 6.9% after Baird downgraded the company and cut its price target from $1,200 to $900. The concern was not a sudden collapse in Caterpillar’s current business. Its first-quarter revenue had risen 22% to $17.4 billion, while data-centre demand helped produce a record backlog for power-generation and backup equipment. Instead, the downgrade questioned how smoothly that growth can continue once valuation, regulation and the physical constraints of data-centre construction are considered.

Caterpillar has gradually become an AI infrastructure stock. Its engines, turbines, generators and construction equipment are used to build and power the enormous facilities required for cloud computing and AI model training. Power and Energy sales rose 22% in the first quarter, supported by reciprocating generators for data-centre customers. Investors therefore treated Caterpillar as another “picks-and-shovels” beneficiary of the AI boom.

The problem is that demand for computing power is not the same as the ability to deliver it.

A hyperscaler may want another data centre, but that project still requires land, planning permission, grid access, transformers, turbines, water and local political consent. New York’s restrictions on new hyperscale developments were cited as one example of growing resistance, including concerns over electricity costs and the burden placed on local infrastructure. Delays may not destroy demand, but they can postpone orders, extend project timelines and reduce the earnings growth embedded in suppliers’ valuations.

That distinction matters. Microsoft’s results suggest that cloud customers are still demanding more capacity. Meta also increased the lower end of its 2026 capital-expenditure forecast to $130 billion, despite free cash flow falling sharply. The warning from Caterpillar is therefore not necessarily that AI spending has stopped. It may be that the market has confused an abundance of proposed spending with an unlimited ability to convert that spending into operational infrastructure.

The wider market context is equally revealing. Vertiv fell sharply after a relatively small revenue disappointment even though management said demand for AI and general computing continued to intensify. GE Vernova had also fallen despite strong growth in its Power and Electrification businesses. These were not identical events, but together they show how little tolerance remains for anything short of flawless execution across highly rated AI-infrastructure beneficiaries.

This may mark the next stage of the AI trade. The first question was who would manufacture the chips. The second was who would provide memory, networking, cooling and power. The next will be more demanding: which companies can convert extraordinary demand into sustainable returns after accounting for financing costs, depreciation, regulation and execution risk?

The lesson is not to abandon AI. Microsoft’s results make that conclusion difficult to justify. It is to stop treating every participant as though it possesses the same economics and risk.

This article is for information only and does not constitute investment advice or a personal recommendation. Capital is at risk, and the value of investments can fall as well as rise.