The Gap Between AI’s Exploding Spend And Its Limited Returns

🔍 Read the full analysis: The Gap Between AI’s Exploding Spend And Its Limited Returns on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

As an affiliate, we earn on qualifying purchases.Hide

Apollo chief economist Torsten Slok said an October 24 research note found no visible broad financial payoff from AI outside the technology sector, as companies continue to commit large sums to computing infrastructure. The assessment points to a gap between the measurable cost of building AI capacity and evidence that most non-tech companies are earning more or becoming more productive from it.

Apollo’s note, titled “No Signs of AI Payoff Outside the Tech Sector,” compares the Magnificent Seven—Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla—with the other 493 companies in the S&P 500. Apollo’s charts show inflation-adjusted revenue per worker rising for the seven and remaining broadly flat for the other companies. It also reports margins near 30% for the seven, compared with roughly 10% for the rest. Those figures describe a difference between company groups; they do not by themselves establish that AI caused it.

The note also points to San Francisco Federal Reserve productivity measures that, according to Apollo, show no clear AI-related boost to productivity growth. Separate data from Ramp, a corporate spending platform with more than 85,000 U.S. business customers, show AI-related spending concentrated among a small share of its customers. The top 10% accounted 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. Ramp’s customer data should not be read as a measure of all U.S. businesses.

Meta’s October 22 quarterly filing details the scale of one company’s infrastructure plans. It lists about $355.73 billion in leases not yet started and $454.52 billion in non-cancelable contractual commitments, plus up to $14.72 billion in contingent cloud-capacity purchases. Together, those items total about $825 billion. The commitments include data centers, energy and network infrastructure, cloud capacity and equipment; the source cautions that not all of the spending is for AI.

The Cost of Building AI Capacity

The contrast matters because infrastructure spending is immediate and visible, while returns for companies using AI are harder to identify in broad financial measures. Meta reported $34.91 billion in capital expenditures in the third quarter and narrowed its 2026 spending outlook to $135 billion to $145 billion. Its operating cash flow was $32.79 billion, leaving free cash flow at negative $2.12 billion for the quarter, compared with positive $10.6 billion a year earlier. These figures show pressure on cash generation during a period of heavy investment, but do not isolate AI as the cause.

For investors and executives, the distinction between a promising technology and a profitable deployment is increasingly important. If firms keep paying for computing capacity without corresponding gains in revenue, productivity or margins, spending could weigh on returns and prompt companies to limit purchases. The source also describes financing as a growing pressure point: major infrastructure projects require capital, while lenders and investors must judge whether future demand can support the debt and commitments being taken on.

The weak aggregate picture does not mean no company is benefiting. The source cites studies reporting a 14% increase in issues resolved per hour in customer support, rising to 34% for novice workers, and retail experiments ranging from no sales gain to a 16.3% increase depending on the workflow. Those results suggest outcomes depend on the task and implementation, rather than establishing a uniform return across industries.

What Apollo’s Comparisons Measure

Apollo’s headline conclusion comes from comparisons of company-level financial and productivity trends, rather than a controlled test of AI adoption. Its figures show that the seven largest technology companies have moved differently from the other 493 S&P 500 firms. That is relevant to the debate over AI returns, but the comparison does not show how much of the gap comes from AI rather than other factors, including business mix, labor costs, interest rates or market conditions.

The spending data also need careful boundaries. The $825 billion figure is a total of disclosed future commitments in Meta’s filing, not a single-year bill or a direct estimate of AI-only spending. It includes leases and contracts extending over multiple years, with some obligations running through 2036. Likewise, industry-wide estimates cited in the source—almost $780 billion in 2026 capital spending by five large technology companies and more than $1 trillion annually from 2027 through 2030—are forecasts, not reported commitments.

Evidence from individual business processes can look different from economy-wide data. The source cites an estimate that 83% of nearly 9,000 executives reported no productivity impact from AI so far, with average reported productivity growth of 0.2% a year. These are survey-based findings as presented in the source, not a definitive measurement of every company’s results. Taken together, the data describe a technology with pockets of measurable operational benefit and no clear broad productivity signal yet.

“No Signs of AI Payoff Outside the Tech Sector”

— Torsten Slok, Apollo chief economist, in the October 24 Daily Spark note

What the Aggregate Data Cannot Show

Apollo’s “zero financial payoff” wording is an interpretation of broad trends, not a direct measurement showing that AI has generated no return. A flat average across 493 companies could conceal firms with gains and losses that offset one another. The comparison also cannot isolate AI from other economic forces. It remains unclear how much of the observed difference between the large technology companies and other S&P 500 firms is attributable to AI investment.

There are limits to the supporting datasets as well. Ramp’s figures cover its own business customers, while the productivity and executive survey measures use different methods and populations. Meta’s future commitments span years and include items not exclusively tied to AI. The source also reports higher insurance costs for debt issued by Meta and CoreWeave, but does not establish how those costs will affect their ability to fund projects. The eventual commercial returns, and the share of current spending that will prove productive, remain unknown.

The Next Test Is Profitable Use

The next evidence will come from company filings and operating results showing whether AI deployments lift revenue, reduce costs or raise output per worker. Meta’s capital-spending plans and cash flow will be watched alongside the commitments listed in its filing. Investors will also be looking for clearer information about how much spending is directed to AI infrastructure and how quickly that capacity is used.

For companies adopting AI, the relevant test is increasingly specific: whether a deployment improves a defined workflow enough to cover its computing and implementation costs. The source does not provide a single upcoming milestone that will settle the broader question. For now, spending totals are easier to count than economy-wide returns, and further financial and productivity data will be needed to determine whether the gap is narrowing.

Key Questions

What did Apollo conclude about AI’s financial returns?

Apollo’s October 24 note said its data showed no visible broad payoff outside the technology sector. That is an interpretation of company and productivity trends, not proof that every non-tech business has received no benefit.

What does Meta’s roughly $825 billion figure represent?

It combines about $355.73 billion in leases not yet started, $454.52 billion in non-cancelable contractual commitments and up to $14.72 billion in contingent cloud purchases. The total spans future years and includes infrastructure and other commitments, not just AI spending.

Is there evidence that AI helps some businesses?

Yes. The source cites customer-support research showing 14% more issues resolved per hour, with a 34% gain for novice workers, and retail experiments ranging from no sales increase to a 16.3% increase. Results vary by task and workflow.

Does Apollo’s analysis prove AI has no return?

No. Its comparisons show that broad financial and productivity gains are not yet clearly visible in the data it examined. They cannot rule out gains at individual companies or determine how much of the observed trends was caused by AI.

Source: ThorstenMeyerAI.com