Paying for AI is now normal. Saying what it earned is not.
One line climbs 6.8 points across eight monthly prints. The other barely moves off the floor.
Sources: Ramp AI Index and Ramp Economics Lab, dated monthly editions from October 2025 to May 2026 (businesses on Ramp’s card and bill-pay platform); Goldman Sachs’ count of S&P 500 earnings calls, reported by Fortune (March 2026) and Investing.com (August 2026). Every figure carries a link to the page it was read from.
Last updated August 17, 2026. How we source this. Each panel keeps its own scale because the two series count two different populations on two different clocks — businesses buying AI tools month by month, and S&P 500 companies speaking on a quarterly call. Neither contains the other, so the distance between the panels is a contrast, not a subtraction.
The climbing line is Ramp’s: 43.8% of businesses on its card and bill-pay platform paid for AI in September 2025, 50.6% by April 2026. The flat one is Goldman’s: 1% of S&P 500 companies put a number on what AI did to their earnings in Q4 2025, and 2% have done so in each quarter since. The two are drawn on separate scales, because one axis cannot show both. This page keeps every figure with its source, and updates when the sources publish.
Last updated August 17, 2026. Latest prints: April 2026 (businesses paying for AI), Q2 2026 (earnings calls). Nothing on this page is modelled, averaged, or filled in.
What the two lines actually say
The spending line is the easy one. Ramp watches AI purchases across the businesses on its card and bill-pay platform. The share buying at least one AI tool was 43.8% in September 2025 — a month it actually fell — then rose through the winter, crossed 50% for the first time in March 2026, and reached 50.6% in April 2026: 6.8 points across eight monthly prints, against about 35% a year earlier. Then the series stops. Ramp’s May, June and July 2026 releases moved on to vendor share and spend per employee and printed no overall rate, so neither do we.
The proof line is the interesting one. Goldman Sachs reads S&P 500 earnings calls and counts who quantifies AI’s effect on earnings. In Q4 2025 it was 1%. In Q1 2026 it was 2%, and in Q2 2026 it was 2% again. In that Q2 season, 65% of the index mentioned AI at all and 11% put a number on a single use case such as coding or customer support. Talking about AI is now ordinary. Costing one use case is rare. Naming what it did to earnings, in a figure you will defend in front of investors, is almost nobody.
Goldman is also explicit about what quantifying did not buy those companies: “Q2 results showed a small and statistically insignificant difference in earnings growth between the companies quantifying AI productivity gains this quarter and other S&P 500 companies.” That cuts against the easy story, so we print it. Measuring is not itself a performance edge, and we will not claim it is. What measuring buys you is the ability to decide. A company that cannot separate what AI cost from what it returned is renewing on faith.
Q2 2026 earnings calls, in three steps
- Mentioned AI on the call65%
- Put a number on one use case11%
- Put a number on AI’s effect on earnings2%
Goldman Sachs US Weekly Kickstart, August 15, 2026 (Ben Snider), as quoted by The Dark Side of the Boom · source. All three steps come from one note, so no cross-quarter comparison is involved.
Every figure, with its source
Download the data (CSV) — the three tables below, one row per figure: period, figure, source wording, population, publication, and source link. The three populations are labelled per row and are not comparable.
Share of US businesses on Ramp’s card and bill-pay platform paying for at least one AI tool. Ramp titles each release with the month it publishes and reports the prior month’s data; the periods below are data months. This is not a census of all US businesses, and Ramp notes its sample skews somewhat toward technology companies.
- Sep 202543.8%The share of businesses paying for AI models and services fell to 43.8% in September, a 0.7% drop — the second decline of 2025.Ramp Economics Lab, October 2025 · source
- Oct 202544.8%AI adoption rose 0.9 percentage points to 44.8% of businesses in October.Ramp Economics Lab, November 2025 · source
- Nov 202545.0%AI adoption held flat at 45% of businesses in November.Ramp Economics Lab, December 2025 · source
- Dec 202546.6%Business AI adoption rose 1.6 percentage points to 46.6% of businesses in December.Ramp AI Index, January 2026 · source
- Jan 202646.8%Overall business AI adoption rose to 46.8% of businesses in January.Ramp AI Index, February 2026 · source
- Feb 202647.6%Overall business AI adoption rose to 47.6% of businesses in February, a record high.Ramp AI Index, March 2026 · source
- Mar 202650.4%Business AI adoption crossed 50% for the first time in March, reaching 50.4% of businesses.Ramp AI Index, April 2026 · source
- Apr 202650.6%Overall AI adoption rose 0.2 percentage points to 50.6%.Ramp AI Index, May 2026 · source
The series starts in September 2025 because the mid-2025 editions carrying June and July are no longer published at a working address, and it ends in April 2026 because Ramp printed no overall rate for May, June or July 2026. The March 2025 comparison (about 35%) appears only as a year-ago clause inside the April 2026 note, so it is quoted here and not plotted.
Share of S&P 500 companies quantifying AI’s earnings impact on their quarterly call, as counted by Goldman Sachs. Read this as disclosure, not as internal measurement: a company can measure AI’s effect on earnings and choose not to discuss it publicly. What the figure does show is how few are willing to put a number on the record.
- Q4 20251%A mere 1% quantified AI’s impact on earnings.Goldman Sachs, via Fortune, March 3, 2026 · source
- Q1 20262%The Q1 2026 share matched Q2’s: 2% quantified AI’s impact on earnings.Goldman Sachs, via Investing.com, August 17, 2026 · source
- Q2 20262%Only 2% quantified AI’s impact on earnings directly.Goldman Sachs, via Investing.com, August 17, 2026 · source
Median monthly AI spend per employee at businesses that spend on AI at all — a narrower group than the series above. Three published months, and they do not rise in a straight line, which is why they are shown as three figures rather than a trend. Ramp’s own August 2026 reading was that businesses are hitting their limit on AI spend even as more of them buy.
- May 2026$11.38Median monthly AI spend per employee, businesses that spend on AI.Ramp AI Index, June 2026 release · source
- Jun 2026$10.59Median monthly AI spend per employee, businesses that spend on AI.Ramp AI Index, July 2026 release · source
- Jul 2026$11.95Median monthly AI spend per employee, businesses that spend on AI.Ramp Economics Lab, August 2026 · source
How to read this tracker
- The two series are kept apart on purpose. They count different populations, by different methods, at different intervals. We do not average them, index them, or subtract one from the other.
- Each series is drawn in its own panel, on its own scale. The spending panel runs 42–52% so a 6.8-point move is visible as a move; the earnings-call panel runs 0–3% from a true zero, so 1% and 2% are legible and the flatness between quarters shows as flatness. Both scales are printed on the chart, and the size of the move is printed next to it.
- A rising line is not a smooth line. September 2025 was a decline, and we plot it as one.
- A series stops where its source stopped. Missing months are missing, not zero, and never filled in.
- An earnings-call count measures what was said on a call. It is the best public read on who can put a number on AI, and it is not a measure of who quietly knows.
- Where a source gives a figure only as a rounded aside, we quote it in prose and leave it off the chart.
Update history
- August 21, 2026Added the CSV download and the embeddable chart. No figures changed.
- August 17, 2026Checked every figure against its source. Latest prints: April 2026 (share of businesses paying for AI), Q2 2026 (earnings calls).
Embed this chart
Paste this snippet to put the live chart on your own page. It carries its own scales, sources, and last-checked date, and it links back here.
<iframe src="https://oaboscorecard.com/tracker-embed" title="The AI Value Gap Tracker — Oabo" width="100%" height="760" style="border:0" loading="lazy"></iframe>The 2% is a scoreboard. Oabo is how you get onto it.
Oabo tracks what your AI costs, records what it returns, and marks how much of that return is verified. A company running Oabo with a populated scorecard can answer the question this tracker keeps asking — not because of a claim we make, but because the number and its evidence are already written down.