AI Is Growing. Why Can AI Stocks Still Fall?

The technology can succeed while an investment disappoints. The difference lies in who earns the money, when it arrives and what investors paid for it.

You can believe that AI will change how businesses operate and still lose money on an AI investment.

There is no contradiction. A technology’s usefulness, a company’s financial performance and the return earned by its shareholders are connected—but they are not interchangeable.

After reading this article, the reader will understand why those three results can diverge, and be able to compare a hypothetical earnings increase with three valuation outcomes before treating “AI is growing” as an investment conclusion.

Why this matters

Recent market turbulence makes that distinction worth examining. On 22 October, US technology stocks fell. Reporting published on 23 October described investor concern about AI spending. That reporting does not establish a single cause for the price moves. [1]

The useful question is not simply whether AI will keep improving. It is what a particular company must deliver to justify the price investors already pay.

Key points

• Usefulness, company performance and shareholder return are connected, and they are not interchangeable.
• Revenue still has to survive prices, operating costs, the cash tied up in infrastructure, financing and any change in the share count.
• Meta’s 21 October results show revenue growth together with capital spending, negative free cash flow under the company’s stated definition, and a legal charge that changes the reading of operating profit.
• In one hypothetical example, earnings per share rise 20 percent. Depending on the valuation multiple, the share-price return is +20 percent, −4 percent or −28 percent. The figures exclude dividends, fees, taxes and currency effects, and they are not forecasts.
• A falling share price does not show that a company has become cheap. A growing industry does not show that every participant will prosper.

Growth has to travel through several steps

An expanding AI market can create demand for chips, computing capacity, software and implementation services. But demand does not reach shareholders unchanged.

Customers must pay. Revenue must cover operating costs. Infrastructure spending must eventually earn an adequate return. Financing and changes in the share count can also affect what remains for each shareholder.

Competition matters throughout that journey.

A service may attract more users while earning less from each task. A customer may gain substantial productivity while the software provider struggles to charge for the value delivered. A supplier may receive strong orders during construction, while the operator buying its equipment waits years to recover the investment.

The industry can grow while the distribution of its profits changes.

What one company’s accounts can—and cannot—tell us

Meta’s 21 October results illustrate the distinction between business growth and cash demands. [2]

The company reported quarterly revenue of $64.60 billion, up 26% from a year earlier. It also reported operating cash flow of $32.79 billion and capital expenditures, including finance-lease principal payments, of $34.91 billion. Under its stated definition, free cash flow was negative $2.12 billion.

Its full-year capital-expenditure outlook narrowed to $135–145 billion from $130–145 billion. The upper bound remained $145 billion.

Those figures show substantial spending alongside revenue growth. They do not, by themselves, establish that the investment will succeed or fail. The results included a $9.89 billion legal charge. Reported income from operations was $12.99 billion. Meta says that excluding the charge, income from operations would have been $22.88 billion. Net income was $9.86 billion. None of those profit figures is the same as free cash flow.

Accounting profit and cash flow also answer different questions. Equipment spending uses cash when paid; depreciation generally recognises its cost over time. A careful analysis should not treat the entire purchase as a current operating expense and then subtract depreciation again.

The economic question is whether the capacity generates enough cash over its useful life to justify its cost.

The same release says the board has approved a new non-voting Class C common stock, with a stated effective date of 19 January 2027, and a one-time stock dividend of one Class C share for each Class A and Class B share. Meta says it expects the Class A price, after that dividend, to reflect a two-for-one split, and that the dividend would not change voting power. This article does not treat that plan as evidence about whether the AI spending will earn its cost.

A growing business can become a falling investment

Consider a fictional company earning $2 per share annually. Its shares trade at $100.

Investors are therefore paying 50 times annual earnings. That price reflects expectations about the company’s future, not merely its current results.

Now suppose annual earnings per share rise 20%, to $2.40. The investment outcome still depends on the valuation investors accept.

The following twelve-month scenarios are entirely hypothetical. They are not forecasts for Meta or any other company.

Each ending price equals earnings per share multiplied by the assumed valuation multiple. Returns exclude dividends, fees, taxes and currency effects.

In every scenario, earnings grow. In two, the share price falls.

This does not mean valuation multiples move independently of the business. Investors may accept a lower multiple because they expect slower subsequent growth, stronger competition, higher financing costs or greater uncertainty.

The example simply shows why “the business is growing” cannot complete an investment argument.

Three different business paths to watch

One possibility is that paid demand catches up with capacity. Customers renew, usage converts into revenue and infrastructure becomes more productive. Evidence would include improving cash generation and sustainable customer economics, rather than spending announcements alone.

Another is that demand develops more slowly than investment. The product may remain useful, but the period before cash recovery lengthens. Financing needs, contractual commitments and the ability to delay further construction become more important.

A third is that adoption accelerates while competition reduces prices. Customers benefit, but providers may need much higher volume to preserve their margins.

These paths can coexist across companies. A chip supplier, cloud operator and application developer do not necessarily face the same economics at the same time.

What to watch next is therefore specific: whether cash generation and customer economics improve, whether financing and construction commitments extend the wait for cash, and whether volume is high enough to offset lower prices. Those indicators can differ by company at the same time.

What this means for a GCC reader

For someone investing through a broad international fund, the relevant exposure may extend beyond an explicitly labelled AI fund. Different funds can hold overlapping large technology companies.

Before adding another investment, examine the actual holdings and how much depends on the same companies or spending cycle. Several fund names do not necessarily represent several independent sources of return.

For a business owner or professional, AI creates a different opportunity: using the technology to deliver a service customers will pay for.

That opportunity also has costs—customer acquisition, implementation, support and continuing maintenance. An attractive investment story about a platform does not automatically make its ecosystem profitable for every service provider.

Readers comparing Gulf AI spending with what a local service business can actually charge may also read The Gulf Is Spending Billions on AI—But Where Is the Opportunity for Local Small Businesses?

The question that improves the decision

A stronger AI investment thesis identifies:

• The paying customer and the value they receive.
• The company’s ability to retain revenue after competition and costs.
• The capital required before cash is generated.
• The growth and profitability already implied by the purchase price.
• The evidence that would weaken the thesis.

A falling price does not establish that a company has become cheap. A rising industry does not establish that every participant will prosper.

AI can become more useful while investors revise what they are willing to pay for its future profits. Understanding that gap makes the discussion more useful than choosing between enthusiasm and a prediction of collapse.

Educational analysis, not an individual security recommendation. Meta’s third-quarter results are dated 21 October 2026. Market reporting published on 23 October describes the 22 October trading session. The valuation table and business paths are CBML illustrations, not return forecasts.

Sources

[1] John Power, “US tech stocks tumble as investors fret over AI spending,” Al Jazeera, published 23 October 2026, on the 22 October 2026 US trading session. Accessed 25 October 2026. Market report. The piece quotes Jay Goldberg of Seaport Research Partners, who said the swings appeared driven more by sentiment than by one fundamental cause, and that people he spoke with cited different causes. Al Jazeera described Meta’s capital-spending update as a rise to $135–145 billion. Meta’s own release says the range narrowed from $130–145 billion. This article follows the company comparison and does not treat the trading session as having a single cause.

[2] Meta Platforms, Inc., “Meta Reports Third Quarter 2026 Results,” press release, Menlo Park, 21 October 2026, for the quarter ended 30 September 2026. Accessed 25 October 2026. No subsequent correction was identified on the investor release on that access date. Figures used here come from that release, including revenue, operating cash flow, capital expenditure including finance-lease principal payments, free cash flow under Meta’s stated definition, the narrowed capital-expenditure outlook, the legal charge, the stated operating-income effect of excluding that charge, net income, and the Class C stock description. Meta’s statements about its share-price performance since its initial public offering were not used.