Cost Cuts No Longer Drive Returns for Infrastructure Providers

A recent Wall Street Journal article highlighted a familiar pattern: companies announce layoffs and their stocks often move higher.

The conclusion is straightforward—cut costs, improve margins, and investors reward discipline. That conclusion is increasingly wrong, especially for companies in the business of providing the infrastructure of artificial intelligence.

Here’s the reason. In prior cycles, layoffs were a direct lever on profitability. Companies reduced operating expenses without materially changing their cost structure—the balance between labor and capital—and margins improved accordingly.

That relationship no longer holds.

In the current cycle, layoffs are occurring alongside a structural increase in capital expenditures tied to AI infrastructure. Companies are not simply cutting costs. They are reallocating them. The result is a shift in where value is created and where constraints exist within the business.

Market Reaction to Recent Layoffs

As Table 1 shows, layoff announcements from this class of companies over the past six months generated only modest short-term stock reactions, typically within a narrow band of low single-digit-percentage gains or losses. More importantly, the 30-day follow-through is inconsistent and, on average, negative. Layoffs are no longer being rewarded as a clear margin signal, and the market’s initial reaction is increasingly subject to reversal.

Where the Constraint Has Moved

In prior cycles, the constraint in these technology companies was labor. Growth required hiring, but hiring is never perfectly executed. Margins were improved by later separating the wheat from the chaff. Layoffs led directly to higher profitability, and the market’s reaction was rational.

In the current cycle, for these companies especially, the constraint has shifted from labor to capital. The limiting factor is no longer the number of employees, but the availability of compute infrastructure, data center capacity, and the capital required to build it. Companies are cutting labor because capital is being redeployed elsewhere.

Measured at the constraint, which is capital deployment into AI infrastructure, the cost structure is expanding, not contracting. The reduction in labor is offset by increased spending on graphics processing units (GPUs), networking, power, and facilities. The business becomes more capital-intensive, not more efficient.

Capital Reallocation, Not Cost Reduction

The relationship between layoffs and capital allocation becomes clearer when viewed across the same companies. As Table 2 shows, the firms announcing layoffs are simultaneously increasing investment in AI infrastructure, confirming that labor reductions are being used to fund capital deployment rather than reduce overall cost. This shift results in a higher fixed-cost base. That makes continued growth imperative, which explains the weaker and less consistent stock reaction.

A Shift in Market Interpretation

The weakening reaction to layoffs reflects a broader shift in how the market interprets these announcements. As Table 3 shows, layoffs were historically associated with modest positive returns when they functioned as a direct margin lever. During 2022-2023, that effect was amplified by a valuation reset in large-cap technology stocks. In the current period, however, both short-term gains and 30-day performance have compressed.

Importantly, this regime shift does not apply to all AI companies, only those that carry the burden of providing heavy duty AI infrastructure for themselves and others.

Nvidia is the world’s leading provider of AI infrastructure via its GPUs. But Nvidia, unlike say, Intel (NASDAQ: INTC), is fabless. It offloads its infrastructure costs to the manufacturers of its devices, primarily Taiwan Semiconductor (NYSE: TSM).

Then there is the whole class of newish companies that now exist only because of AI, bear few of AI’s infrastructure costs, and reap huge benefits from AI’s replacement of human labor, especially management.

We’ve written elsewhere about Midjourney, an AI image platform serving millions of users, reaping annual revenues circa $200 million, with a staff of no more than 20. Do the long division and revenues come to $10 million to $15 million per employee.

Midjourney is an extreme example of a powerful trend. AI is replacing the middle management hordes formerly required to coordinate corporate efforts. It is massively cutting back on meetings, reports, and layers of supervision.

OpenAI employs massive AI infrastructure, but almost all of it is outsourced to Microsoft Azure. MSFT’s layoffs, as shown above, are largely offset by its huge and growing capital costs—which it maintains to the benefit of customers such as Open AI. Meanwhile, Open AI’s revenue per employee exceeds $1 million.

Other firms are now proactively requiring that before a new employee is hired managers must demonstrate that the job cannot be done by AI.

Sam Altman has suggested that a one-person billion-dollar company is now conceivable—because AI is replacing managers.

The semiconductor industry celebrates that its sales are now pushing a trillion dollars a year. But now, as always, the big money usually goes not to the companies that make semiconductor devices but those that use them most creatively. AI exhibits that truth exponentially. But it works best if you can get someone else to pay for your data centers.

Investor Takeaway

The data does not support the conventional view that layoffs are necessarily bullish. The average reaction is now modest in the short term and negative over a 30-day period.

Increasingly, layoffs are a capital allocation signal.

Investors who always interpret layoffs as a sign of discipline are focusing on the wrong variable. They must ask not only how much cost is being removed, but where that cost is being redeployed. For companies in the AI infrastructure business, the answer is not necessarily encouraging. For companies in the business of using AI creatively the world is a beautiful place.

At the Gilder Technology Report, we sort for technological impacts that make it to the bottom line.