Torsten Slok, Partner and Chief Economist at Apollo Global Management, published a research note on October 24, 2026, finding that AI tools have delivered zero financial payoff to US firms outside tech. The publication, titled “No Signs of AI Payoff Outside the Tech Sector” in Apollo’s The Daily Spark newsletter, compares the Magnificent 7 tech leaders against the remaining S&P 493 companies. Slok found that revenue per worker and operating margins climb for big tech developers like Microsoft and Alphabet, but financial results for the rest of the market stay flat [1]. Slok’s Apollo AI report argues that mainstream businesses view token bills as an overhead expense that drags on margins instead of driving sales growth [3].
Apollo AI Report Core Findings
Slok evaluated five separate charts covering corporate earnings, labor yield and business card transactions. Slok tracked five charts [1]. He pulled data from Bloomberg, Macrobond and the US Bureau of Labor Statistics to chart inflation-adjusted sales per employee from the fourth quarter of 2019. He also reviewed the San Francisco Fed TFP utilization-adjusted series designed by John Fernald alongside transaction records compiled by Ara Kharazian from credit card processor Ramp [3, 4]. Mainstream firms haven’t seen the revenue gains promised by generative software [1].
Yash Thakker at explainx.ai said that core findings from the Apollo AI report spread quickly to boardrooms after coverage on Zeli and Hacker News [3, 5]. Marc Rowan, Apollo Chief Executive, previously framed the firm’s differentiated strategy around this slower market reality [1, 7]. Slok said that early adopters in non-tech fields face high friction from government regulations, internal data governance requirements and rigid legacy workflows that resist rapid software automation. Startups buy tokens to train tools, but ordinary companies can’t justify high bills without proven gains [1]. In a June 30 note in The Daily Spark, Slok cited work from Apollo Thematic Investing, writing that capital-intensive sectors like logistics, utilities, healthcare and manufacturing need years of process redesign before algorithmic tools translate into higher cash flows. Equity markets face repricing if earnings take five years to arrive [7].
Slok argued that basic differences between software creators and general firms explain why broad returns stay flat [3]. AI functions well as a fast calculator, but real company choices demand context and human judgment. Algorithms don’t grasp accountability, trust, team culture or ethical trade-offs, meaning automated language models can’t handle delicate reputational risks during high-stakes corporate operations [1].
Mag 7 Margin Split with S&P 493
Apollo’s first two charts plot the performance divide between the seven dominant tech firms and the S&P 493 [1]. For the Magnificent 7, inflation-adjusted revenue per employee has risen steadily since late 2019 [4]. Their profit margins grew along the same path. The Mag 7 pulled ahead. For the other 493 companies in the broad S&P 500 index, both revenue per worker and profit margins remained flat throughout the entire period from 2019 to 2026. Slok pointed out that high market prices assume general companies will see a surge in profit, yet quarterly filings show no evidence of that shift [1].
BrainDetox researchers pointed out that group index numbers can obscure separate business causes. Sales per worker is a financial ratio that climbs if company headcount drops at steady sales, or if gross sales rise with an unchanged number of employees. The chart data alone can’t isolate which variable moved [4]. External forces like interest rates, tariffs and wages also affect S&P 493 margins [3].
Ara Kharazian, Ramp chief economist, tracked billing data from more than 85,000 firms to assess software spending [3, 4]. Across Ramp’s records, 55.9 percent of US businesses paid for AI products in September 2026, up 12.1 percentage points from a year earlier. Yet dollar spending stays concentrated in a tiny group of software teams. Mainstream firms spent far less. The top 1 percent of paying firms had a median spend of $9,567.53 monthly per worker, but the median paying firm spent $15.70 ($15.70 per worker). Ramp’s calculation reveals a 609-fold difference between heavy users and ordinary companies [4].
Productivity Measures from the San Francisco Fed
San Francisco Fed TFP data underpins Apollo’s third chart on national business output trends [1, 3]. Slok selected the utilization-adjusted total factor productivity series designed by John Fernald in 2014, which adjusts traditional growth statistics for cyclical shifts in labor effort and capital hours [4]. It shows no sign of an AI lift [1].
BrainDetox quarterly figures show a stark contrast between standard and adjusted metrics. In the first quarter of 2026, unadjusted TFP grew at an annualized rate of 0.87 percent, but the utilization-adjusted series fell by 1.45 percent. The second quarter showed unadjusted growth of 0.62 percent, but adjusted TFP dropped 1.11 percent. Over the four full consecutive quarters ending in mid-2026, standard total factor productivity rose 1.42 percent, but the utilization-adjusted measure fell 0.07 percent across the same period. During that same second quarter, overall output grew 2.67 percent and labor productivity gained 2.31 percent, but a 1.73 percent rise in factor utilization (annualized quarterly rate) dragged down the adjusted metric [4].
Ramp industry records explain why national output gains stay quiet outside software. In Ramp’s September 2026 data, the information sector reached 79.9 percent AI adoption, finance reached 73.7 percent and manufacturing stood at 60.9 percent. Retail recorded 49.9 percent, construction hit 43.2 percent, healthcare reached 42.5 percent, and food services trailed at 33.0 percent. Adoption in heavy industries lags software [4].
Examining Apollo AI Report Threads on Hacker News
Apollo AI report discussions climbed past 300 points on Hacker News before reaching 470 points with 462 comments [3, 6]. Technologists compared Slok’s conclusions to earlier computing waves. User jstanley quoted economist Robert Solow’s famous 1987 line about seeing the computer age everywhere except in productivity statistics. Commenter briHass argued that advanced reasoning models have circulated for only nine months, saying business changes take far longer. Commenters drivebyhooting and bob1029 pointed out that tech firms have outpaced other industrial sectors for fifteen straight years, and cmiles8 predicted enterprise budget cuts as corporate planning cycles begin [3].
Botonomous writer Ananya Reddy compared current AI adoption delays to factory electrification in the 1920s. Early factory owners saw little initial gain from electric motors because production lines were still arranged around central steam shafts, with major output leaps arriving decades later when engineers rebuilt facilities around distributed electric power. Electrification took decades to show results [2].
Ananya Reddy described non-tech firms as operating in “adaptation deserts” where old rules and legacy systems turn new software into minor tweaks [2]. Hacker News user secabeen agreed, noting that physical industries need costly plant redesigns before software yields returns. Commenter tapoxi pointed out that software teams use test suites as validation gates to check model outputs automatically. Checking generated text by hand takes time. Without automated software tests, manual checking creates what analysts call an “AI verification tax,” where thirty minutes saved drafting new text is consumed by twenty-five minutes of subsequent human auditing. Commenter realusername reported slightly negative productivity returns from workplace AI [3].
Budget Pressures and Model Sizing for Enterprise Teams
Finance teams are re-evaluating model token expenses as annual planning begins. Yash Thakker at explainx.ai said that chief financial officers now treat token consumption as an operational overhead cost, requiring teams to tie software usage directly to tangible business metrics like revenue per representative, ticket volume or defect counts. Prompt counts don’t prove business value [3].
Survey data reinforces this spending caution among enterprise buyers. Gartner research found that only 13 percent of CIOs report significant value from AI deployments. The McKinsey State of AI ROI review similarly found that leaders demand clear cost benefits before widening budgets [3]. Slok said general buyers don’t need expensive frontier tiers for basic office tasks and refuse to pay premium prices [1]. Major research labs protect proprietary commercial world models behind enterprise firewalls, but mainstream corporate teams choose smaller open weights to control monthly bills. Meanwhile, silicon designers fund physical AI chips to power robotics on factory floors.
Data from the Apollo AI report shows that non-payroll business spend on artificial intelligence dropped to 5.77 percent in September 2026, down from 5.96 percent in August [4]. Slok concluded that the broader AI payoff will depend on corporate performance outside tech, particularly among the hundreds of companies in the S&P 493. Until mainstream businesses redesign their core workflows around software tools, aggregate economic lines will likely stay level [1].
- REPORT Slok, T. (2026, October 24). No signs of AI payoff outside the tech sector. The Daily Spark, Apollo Global Management. [Article Link]
- ONLINE NEWS Reddy, A. (2026, October 24). No signs of AI payoff outside the tech sector. Botonomous. [Article Link]
- ONLINE NEWS Thakker, Y. (2026, October 24). No AI payoff outside tech? Apollo Slok note, checked. explainx.ai Blog, AISOLO Technologies. [Article Link]
- WEBSITE BrainDetox. (2026, October 25). Apollo on AI payoff outside tech: S&P 493 charts explained. BrainDetox. [Article Link]
- ONLINE NEWS Zeli. (2026, October 24). AI delivers zero payoff to corporate America outside tech, Apollo economist says. Zeli. [Article Link]
- ONLINE NEWS HotON.ai. (2026, October 25). No signs of AI payoff outside the tech sector. HotON.ai. [Article Link]
- REPORT Slok, T. (2026, June 30). AI: The ROI runway could be long outside the tech sector. The Daily Spark, Apollo Global Management. [Article Link]
