# "No Signs of AI Payoff Outside the Tech Sector" — What Changed, Why It Matters ## The core claim (published today, Oct 24, 2026) Apollo Chief Economist **Torsten Slok** argues AI has delivered "zero financial payoff to corporate America outside of tech." The evidence he marshals: - **Revenue per worker** is rising for the Magnificent 7 but **flat for the S&P 493** (Bloomberg/Macrobond data). - **Profit margins** are rising for the Mag 7 but **flat for the S&P 493**. - **No sign of AI lifting productivity growth** — using the San Francisco Fed's utilization-adjusted total factor productivity series. - **Mainstream corporate America is pulling back**: token spending is increasingly viewed "not as a source of growth but as a cost center that weighs on margins." VC-backed firms keep spending on tokens, but mainstream firms "do not need the latest model and are unwilling to pay for it." - Bottom line: *"the AI story will be decided in the S&P 493, not the Mag 7."* Source: [Apollo Daily Spark, Oct 24, 2026](https://www.apollo.com/wealth/insights-news/insights/daily-spark/no-signs-of-ai-payoff-outside-the-tech-sector) ## What actually changed vs. earlier in 2026 This is an escalation, not a new thesis. In **July 2026**, Slok's argument was conditional: AI valuations rest on the *promise* of S&P 493 margin gains, and in regulated sectors (healthcare, banking, energy, pharma, manufacturing) productivity gains could arrive "well beyond what the market currently projects" — if it takes five years instead of five months, AI stocks face painful repricing. That framing was about *timing risk*. ([The Decoder, Jul 7, 2026](https://the-decoder.com/apollo-economist-warns-ai-profit-gains-outside-tech-could-take-well-beyond-what-wall-street-expects/); Bloomberg covered the same warning on Jul 7, 2026.) The October 24 note is a **harder claim**: not "the payoff is late," but "there is no evidence of a payoff at all," plus a new behavioral datapoint — non-tech firms actively cutting back token spend. The argument has also shifted from a macro-timing story to a **structural one**: Slok now argues intelligence is commoditizing (cheap tokens) while business value depends on judgment, accountability, trust, culture and reputational risk — things AI doesn't supply. ## The supporting data he leans on (Ramp AI Index) Apollo's 4th and 5th charts use Ramp's transaction-based data (Ara Kharazian). The live index shows the split clearly: - **Adoption is broad but shallow**: 55.9% of U.S. businesses on Ramp paid for AI in Sept 2026, up 0.6 pts from August and 12.1 pts YoY. - **Spend is extremely concentrated**: the top 1% of AI-paying firms spent a median **$9,567.53 per employee per month**, vs **$15.70** for the median firm — a ~600x gap. - **AI is only 5.77% of business spend** (excl. payroll), and that share *fell* 0.19 pts in September. - **Token prices are collapsing**: blended OpenAI/Anthropic price per million tokens fell 25.4% month-over-month ($0.86 → $0.64); OpenAI alone fell 34.5%. - **Adoption is lopsided by sector**: Information 79.9%, Finance & Insurance 73.7%, Manufacturing 60.9%, Retail 49.9%, Construction 43.2%, Healthcare 42.5%, Accommodation & Food 33.0%. Source: [Ramp AI Index](https://ramp.com/data/ai-index) The St. Louis Fed's FRED Blog (Oct 19, 2026) independently confirms the sector ordering and notes adoption is spreading fastest at large and medium firms, with small firms lagging — and cautions the index misses free-tool usage and intensity of use. ([FRED Blog](https://fred.stlouisfed.org/blog/2026/10/which-businesses-are-using-ai-technology)) ## Independent corroboration on the productivity gap St. Louis Fed researchers (Ozkan, Kalyani, Sullivan) analyzed ~490,000 earnings-call transcripts from 5,198 U.S. public firms. Two findings that directly support Slok: - **Utilization-adjusted TFP grew only 0.07%** over the four quarters ending Q1 2026. - **~95% of AI-related productivity sentences on earnings calls refer to *future* gains**, vs ~three-fourths for non-AI productivity talk — i.e., executives are promising, not reporting. Source: [St. Louis Fed, On the Economy, Jul 31, 2026](https://www.stlouisfed.org/on-the-economy/2026/jul/ai-productivity-what-firms-say-earnings-calls) ## The strongest counter-argument J.P. Morgan Asset Management (Meera Pandit, Stephanie Aliaga) concedes monetization is "concentrated in the infrastructure layer" and "elsewhere… still early, uneven, and opaque," but argues the earnings case holds: hyperscaler AI-segment revenue grew ~35% y/y in 4Q25, mega-cap tech 2026 EPS was revised to **24.3%** vs **13.9% for the S&P 493**, and hyperscaler net leverage is 0.9x vs 2.6x for the average IG issuer. Notably, their own forecast *embeds* Slok's gap — the S&P 493 is expected to grow earnings far slower than mega-cap tech. ([J.P. Morgan AM, "Evaluating AI"](https://am.jpmorgan.com/wr/en/asset-management/liq/insights/market-insights/market-updates/on-the-minds-of-investors/evaluating-ai/)) ## Why it matters 1. **It reframes the AI trade as a breadth problem.** If the payoff never reaches the S&P 493, the market's ~60%+ concentration in mega-cap tech is a single-factor bet on continued Mag 7 margin expansion — and Slok's July note explicitly flagged that as a repricing risk for AI stocks. 2. **The "cost center" shift is the new, actionable signal.** Falling token prices (down 25–35% m/m) are good for buyers but compress model-provider revenue; simultaneously, mainstream firms are reported to be trimming spend. Cheap intelligence that nobody scales up is deflationary for the AI revenue story. 3. **It sets a measurable test.** Slok's framing gives a clean scoreboard: watch S&P 493 revenue-per-worker and margins, not Mag 7 earnings. The St. Louis Fed's 0.07% TFP reading and the 95%-forward-looking earnings-call language say the test is not yet being passed. 4. **Caveats worth holding.** The Ramp data covers 85,000+ firms on one spend platform and excludes free tools and usage intensity; Apollo's charts are proprietary and the underlying series aren't fully public. And J.P. Morgan's point stands that infrastructure-layer monetization is real and large — the disagreement is about whether it *diffuses*, and how fast. **Bottom line:** As of October 24, 2026, the strongest available evidence — Apollo's margin/revenue-per-worker charts, Ramp's 600x spend concentration and 5.77% share of business spend, and the St. Louis Fed's 0.07% TFP growth with 95% forward-looking AI talk — supports the claim that AI's measurable financial impact remains confined to tech. The open question is timing versus structure: Slok now argues it's structural; J.P. Morgan and consensus still price it as a 2027–2028 diffusion story.