Why Companies Keep Spending On AI Despite Limited Payoffs

🔍 Read the full analysis: Why Companies Keep Spending On AI Despite Limited Payoffs on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

• Fast, free delivery on millions of items
• Prime Video, Amazon Music and more included
• Member-only deals all year

Start your free Prime trial →Free trial for eligible customers · Cancel anytime

As an affiliate, we earn on qualifying purchases.Hide

Apollo chief economist Torsten Slok says available data show no visible AI payoff across most non-tech companies, even as technology firms and other investors commit enormous sums to computing infrastructure. His October 24 research note describes an aggregate gap between AI spending and returns; it does not establish that every company or use of AI has failed.

Apollo’s charts compare the seven largest technology companies, often called the Magnificent Seven, with the other 493 S&P 500 companies. Revenue per worker and profit margins have risen for the seven, while the corresponding averages for the other companies have remained broadly flat. Apollo also cites productivity measures from the San Francisco Federal Reserve that, according to the note, show no clear AI boost to productivity growth.

The note’s evidence also points to concentrated spending. Data from Ramp, a corporate spending platform with more than 85,000 U.S. business customers, show that the top 10% of firms account for 99.5% of spending on model serving and inference, 99% of neocloud spending, and 95.6% of spending on OpenAI and Anthropic, according to the source material. Those figures describe Ramp’s customers, not all American companies.

Meanwhile, Meta’s October 22 quarterly filing lists about $825 billion in future commitments, including leases and contractual purchases, mostly related to data centers, servers and cloud capacity. The total includes commitments extending over years and is not all AI spending. Meta reported $34.91 billion in capital expenditures in the third quarter and narrowed its 2026 capital-spending outlook to $135 billion to $145 billion.

The gap between AI bills and returns

The evidence matters because companies are making large, long-term infrastructure commitments before broad financial gains are clear in public measures. If expected demand or productivity improvements do not arrive, firms could face pressure on margins, cash flow and borrowing costs. That risk reaches beyond technology companies when cloud providers, chipmakers and data-center operators depend on continued investment.

But an aggregate result is not a verdict on every AI project. A flat average across hundreds of companies can conceal firms that gain and firms that lose, and it cannot by itself show whether AI caused the change. Apollo’s conclusion is best read as a warning that economy-wide returns are not yet evident, rather than proof that AI has no value.

There are examples of narrower gains in the source material. Studies cited there report a 14% increase in issues resolved per hour in customer-support work, rising to 34% for novice workers, and retail sales results ranging from no increase to 16.3%, depending on the workflow. These are task- or experiment-specific results; they do not establish a general return across companies.

What Apollo’s figures capture

Slok’s October 24 note, titled “No Signs of AI Payoff Outside the Tech Sector,” builds its case from company revenue and margin trends, productivity statistics, and data on corporate AI spending. He has also described token spending as a cost that may weigh on margins and raised concerns about the cost of computing relative to the benefit users receive. Those are the economist’s interpretations of the measures, not direct measurements of AI’s causal effect on each business.

The distinction matters: the Magnificent Seven are themselves major technology businesses and AI infrastructure investors, while the other 493 companies span many industries and may adopt AI at different speeds. The source material also cites a survey in which 83% of nearly 9,000 executives said AI had not affected productivity so far, with an estimated average productivity effect of 0.2% a year. The survey’s scope and methodology are not detailed in the supplied material, so those numbers should not be treated as a universal measure.

Meta’s filing illustrates the scale, but also the limits of interpreting spending totals. Its commitments include leases for data centers and energy and network infrastructure, as well as purchases of cloud capacity, servers and networking equipment. Some spending is not AI-related, and commitments scheduled over many years are not the same as cash paid immediately.

“AI “has delivered zero financial payoff to corporate America outside of tech.””

— Apollo chief economist Torsten Slok, in an October 24 research note

What the payoff data cannot show

The available figures do not establish whether AI caused the observed revenue, margin or productivity trends. Flat company averages can hide gains and losses within sectors, and other forces—including labor costs, interest rates and tariffs—may also affect results. The supplied material does not provide enough detail to independently evaluate every underlying dataset or the executive survey.

It is also unclear how quickly current infrastructure spending will translate into revenue or cost savings, and whether those returns will accrue to the companies paying for computing capacity or to their suppliers. Meta’s commitments are not equivalent to immediate spending, and the source says they include some non-AI items. The future spending estimates for major technology firms are forecasts, not contractual commitments.

Evidence to watch in coming quarters

Investors and companies will be watching upcoming earnings reports for evidence that AI-related investment is producing measurable revenue, productivity or cost savings. For Meta, future results and cash-flow disclosures can show how its capital spending and commitments develop, but the filing alone cannot determine whether the investments will earn an adequate return.

Further productivity data and more detailed company reporting may help distinguish broad economic effects from gains in particular workflows. Until then, the central issue remains whether targeted improvements can scale beyond a small group of firms and justify the rising cost of computing capacity.

Key Questions

Does Apollo’s note prove AI has no financial payoff?

No. It reports that aggregate measures show no visible payoff across most non-tech S&P 500 companies. Those comparisons do not prove that every AI use has failed or isolate AI’s effect from other economic factors.

Why are companies still spending heavily on AI?

The supplied material shows continuing investment in data centers, servers and cloud capacity, but does not establish each company’s motives or expected return. The spending may reflect expectations of future demand and the need to build computing capacity; whether those investments pay off remains uncertain.

What does Meta’s $825 billion figure represent?

It is an estimate of future commitments listed in Meta’s filing, including leases and non-cancelable contracts, largely for data centers, cloud capacity, servers and network equipment. It is spread over time and includes some items that are not AI-related.

Are there examples of AI delivering measurable gains?

The source material cites reported gains in specific settings, including customer-support work and some retail experiments. Results vary by task, and those examples do not establish a broad return across the economy.

Source: ThorstenMeyerAI.com