🔍 Read the full analysis: AI Spending Is Exploding. The Benefits Are Harder To Spot on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Companies are committing vast sums to AI infrastructure, including Meta’s roughly $825 billion in future contractual and lease obligations, while broad productivity data show little evidence of an economy-wide boost so far. Apollo economist Torsten Slok’s “zero payoff” description is an interpretation of aggregate trends, not proof that AI produces no returns: studies find measurable gains in some specific tasks.
Companies are pouring money into artificial intelligence infrastructure, but broad measures have yet to show a comparable productivity payoff across most of corporate America. The contrast is laid out in an October 24 note by Apollo Global Management chief economist Torsten Slok and sharpened by Meta’s filings, which list about $825 billion in future commitments, mostly tied to data centers, servers and cloud capacity. The figures do not establish that AI has no benefits; they show that investment is easier to measure than economy-wide returns so far.
Slok’s note, titled “No Signs of AI Payoff Outside the Tech Sector,” compares revenue per worker and profit margins for the seven largest technology companies with the other 493 companies in the S&P 500. Apollo’s charts show revenue per worker and margins rising for the seven, while the measures for the other companies have remained broadly flat. The note also cites San Francisco Federal Reserve productivity indicators that show no clear AI-driven increase in productivity growth.
Those comparisons describe broad trends, not a controlled measurement of AI’s effect. Flat results across hundreds of companies can conceal firms or departments that gained while others did not, and the data cannot isolate AI from changes in labor costs, interest rates or tariffs. Apollo’s conclusion that AI has delivered “zero financial payoff” outside technology is Slok’s interpretation of those indicators, rather than a direct measurement proving no company outside tech has benefited.
Meta’s October 22 filing details the obligations behind the headline figure: 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 purchases. The categories include data centers, energy and network infrastructure, servers and cloud capacity; not every item is exclusively for AI. 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.
Meta also reported $64.60 billion in third-quarter revenue, up 26% from a year earlier, and $32.79 billion in operating cash flow. Its resulting free cash flow was negative $2.12 billion, compared with positive $10.6 billion a year earlier. Separately, the source material reports that SpaceX borrowed $40 billion to buy AI chips, adding to questions about how the infrastructure buildout will be financed.
Infographic 01 · The “zero” in numbers
What “zero financial payoff” actually measures.
Apollo’s headline rests on five charts. Here are their numbers, and what they can and cannot prove.
Who pays for AI? Share of spending by the top 10% of firms
Ramp customer data via Apollo · non-AI categories in grey for comparison
Spending is concentrated in every category, not only AI; AI is the most extreme.
What the five charts do and do not show
shown not shown
- Shown: revenue per worker and margins rising for the 7 tech giants, flat for the other 493
- Shown: no AI lift yet in US productivity statistics (San Francisco Fed TFP, output per hour)
- Shown: AI spending sits in few, mostly venture-funded firms (Ramp, 85,000+ US businesses)
- Not shown: that AI caused the gap; rates, tariffs and wages move the same lines
- Not shown: gains inside single functions, which can cancel out in an average
- Not shown: the S&P 493 itself in the Ramp data; those are Ramp’s customers
“The AI story will be decided in the S&P 493, not the Mag 7.”Torsten Slok, Apollo chief economist, 24 Oct 2026 (opinion)
Sources: Apollo Daily Spark (24 Oct 2026) with charts by Bloomberg, Macrobond, SF Fed, BLS and Ramp; Apollo (20 Oct 2026); Ramp AI Index. Values with ≈ are read off Apollo’s charts. Assessment: Thorsten Meyer.
Powered by Thorsten Meyer AI
The Cost of Building AI Capacity
The difference between infrastructure spending and broad business returns matters because companies are making large, long-dated commitments before the payoff is visible in aggregate productivity figures. Meta’s lease and contract obligations run across several years, while its capital spending plan shows that substantial near-term outlays are continuing. If returns remain concentrated among a small group of technology companies and heavy AI users, other firms may face costs without yet seeing enough revenue or efficiency gains to justify them.
Funding is also becoming a focus alongside access to chips and electricity. The source material reports a $40 billion SpaceX borrowing for AI chips and says insurance against debt default for Meta and cloud provider CoreWeave became more expensive. It also cites Slok’s view that some companies are starting to treat token spending as a cost center. These developments point to financing and operating costs as questions for investors and businesses, but they do not by themselves establish that the investment will fail to earn a return.
The aggregate figures also do not mean AI has no practical value. A customer-support study cited in the source found a 14% increase in issues resolved per hour, rising to 34% for novice workers. Retail experiments reported sales results ranging from no increase to a 16.3% gain, depending on the workflow. Those results suggest that benefits may be measurable in specific uses even when company-wide productivity statistics have not moved.
AI workstation computer with high-performance GPU
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Apollo’s Charts Measure
Apollo’s October 24 note brought together five sets of indicators: revenue per worker and profit margins at large technology firms compared with other S&P 500 companies; productivity measures; the concentration of AI-related spending among businesses using Ramp; and the funding profile of the biggest AI spenders. The distinction between the technology leaders and the rest of the index is central to its argument.
Ramp’s figures cover its own base of more than 85,000 U.S. business customers, not all U.S. firms. In that data, the top 10% of companies accounted for 99.5% of spending on model serving and inference, 99.0% of neocloud spending and 95.6% of spending on OpenAI and Anthropic. Apollo also reported that 80.4% of the top 1% of AI spenders were venture-backed, compared with 13.1% of the median spender. These figures indicate concentration within Ramp’s customer data; they are not national market shares.
The spending estimates for other major technology companies require a separate distinction. The source cites projections of almost $780 billion in 2026 capital spending by Amazon, Alphabet, Meta, Microsoft and Oracle, and more than $1 trillion annually from 2027 through 2030. Those are forecasts, not disclosed commitments. Meta’s roughly $825 billion figure, by contrast, is a total of specified leases and contractual obligations, with some contingent purchases included.
“The company narrowed its 2026 capital-spending outlook to $135 billion to $145 billion.”
— Meta, in its third-quarter results and 2026 spending outlook
How Much AI Is Paying Back
The available indicators do not isolate AI’s contribution to company revenue, margins or productivity. It remains unclear how much of the difference between technology leaders and other S&P 500 companies comes from AI, and how much reflects their other businesses or market conditions. Apollo’s Ramp data also represent one platform’s customers, while the productivity measures track the broader economy rather than particular AI deployments.
The roughly $825 billion figure for Meta combines several kinds of obligations and is not a forecast of spending in a single year. Some commitments cover infrastructure beyond AI, and some cloud purchases are contingent. The source does not provide a single, comparable measure of realized AI returns across the companies making the largest investments. The cited task-level gains, meanwhile, come from particular studies and experiments and should not be treated as results that every employer will reproduce.
Spending and Productivity in Focus
The next evidence will come from further company filings and earnings reports, which can show whether infrastructure outlays continue to rise and whether companies describe measurable revenue or cost savings from AI use. Investors will also watch financing conditions for data-center operators, cloud providers and firms seeking large amounts of computing capacity. The source material does not identify a single upcoming milestone that will settle the question.
For now, the central test is whether gains recorded in particular tasks spread into company-wide results. Until productivity, margins or revenue show a clearer connection to AI use across a wider group of businesses, the scale of spending will remain more visible than the return.
Key Questions
Does Apollo’s analysis prove AI has produced no financial benefits?
No. Apollo’s “zero payoff” language summarizes aggregate trends in productivity, revenue per worker and margins. The measures do not rule out gains at individual firms or in specific tasks, and they do not isolate AI from other economic factors.
What is included in Meta’s roughly $825 billion commitment figure?
Meta’s filing lists 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 obligations include data centers, energy and network infrastructure, servers and cloud capacity; not all are exclusively for AI.
Are companies seeing any measurable gains from AI?
Some task-specific results cited in the source show gains, including a 14% increase in customer-support issues resolved per hour and a larger increase for novice workers. Retail experiments ranged from no sales increase to a 16.3% gain, depending on the workflow. These findings do not establish that every business will see similar results.
Why does the source say AI spending is concentrated?
In Ramp’s data from more than 85,000 U.S. business customers, the top 10% of firms accounted for most measured spending on model serving, inference, neocloud services and the named AI providers. The figures describe Ramp customers, not all businesses or the full U.S. economy.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
