A record that is 1.65% high
Ramp's monthly index has set another record, and the size of the record is the useful part. AI spending by the businesses on its platform rose 1.65% in October from September. That is growth, and it is also close to flat. Ramp's lead economist Ara Kharazian, who writes the index, frames the post as a continuation of his "Cracks in the AI Thesis" series, not as a victory lap: the headwinds he listed there, he says, remain.
One of those headwinds is price. The average price per million tokens fell 21% in a single month, and in Ramp's token data, volume growth continues to outpace token spend. Read together, those two figures describe customers consuming more tokens without a matching rise in what they pay for tokens. That is good for buyers. For the labs selling the tokens, this sample and period show increased consumption translating into slower spending growth.
$16.32 against $10,200
The distribution matters more than the total. The median firm on Ramp spent $16.32 on AI in October, up 93 cents. The top 1% of firms spent a median $10,200 per employee per month, up 9%. For the top 10% group, Ramp reports $801, roughly flat, without specifying the figure's unit.
One caution on the contrast: these figures are not points on a single scale. Ramp reports the median-firm figure per business and the top-1% figure per employee, so they should not be read as two ends of one distribution or divided into a ratio. What they show separately is a typical firm spending very little, and an elite group spending heavily per head.
That group is small and specific. Ramp says its top spenders are largely VC-backed companies in tech, finance and technical services, and puts the consequence plainly: 1% of businesses drive 80% of the revenues it attributes to OpenAI and Anthropic, and spending is failing to spread outside those sectors. Growth in the median and top 10% groups, in its words, is relatively sluggish.
Any claim that AI spending is now broad-based has to answer the $16.32 figure first. On this data, it is not.
Where the money sits, and the one share that moved
Ramp published a department breakdown for the first time in this edition, and it confirms where the money has been. Engineering teams drove 72% of classified spend in October. Ramp also says engineering spend has fallen from highs observed earlier in the summer, even as usage rose.
The share that moved belongs to customer success and support operations: 13.3% of classified spend in October, up from 0.8% in June. Ramp suggests customer service chatbots as a possible driver, and it is careful about the scale. That growth has not matched the rapid growth engineering posted early this year, and Ramp notes these use cases are a good candidate to rely on cheaper and faster models rather than the highest-margin frontier ones. For now the frontier share still edged up: the models Ramp names (Opus, Fable, Sol and Astra) took 69.5% of spend, up from 67.7%. Both things can be true for a while. The question the next editions answer is which trend compounds.
On vendor adoption, October made the race three-sided on Ramp's platform — Ramp's own framing for this edition. SpaceXAI was paid by 6.3% of Ramp businesses, up 2.2 percentage points in a month, which Ramp calls breakout growth; in an earlier post it called SpaceXAI the fastest-growing software company in its dataset. Anthropic rose 1.1 points to 44.8% and OpenAI held flat at 40.6%. Overall, 56.8% of Ramp businesses paid for some AI, up 1 point. Adoption percentages are not revenue, and Ramp does not present them as such. A firm paying SpaceXAI $20 and Anthropic $20,000 counts once in each column.
What a card platform cannot see
Ramp's index is payment data, and its limits are stated in its own methodology notes, summarized by The Decoder. The index draws on more than 85,000 US companies that use Ramp's cards and bill pay. It misses free tools and anything bought on personal accounts. Its token volume and token spend figures come only from businesses using Ramp Token Spend Management, a sample Ramp itself says is not representative of overall AI spending on the platform. The customer base also skews growth-oriented, tech-savvy and larger than US businesses overall, so the 56.8% adoption figure is likely higher than a representative sample would show.
None of that makes the index useless. It makes it a measurement of paying businesses on one platform, published monthly. The trend lines (a record total, a 21% monthly price fall, a median that barely moved) are the product. The level claims (how much "US business" spends on AI) are the part to discount.
Sources
- [1] Ramp, "November 2026 Ramp AI Index: New Fronts in the AI Trade," by Ara Kharazian, published November 11, 2026 (read first-hand in full; all spend, adoption, department and price figures above are from this post)Read source
- [2] The Decoder, "US corporate AI spending hits a new high in October, but growth slows to 1.65 percent," November 11, 2026 (read first-hand; used for its summary of Ramp's methodology limits: the 85,000-company base, the missing free and personal-account use, the Token Spend Management sample caveat and the customer-base skew)Read source
- [3] MLQ, "US business AI spending hit a record in October but grew less than 2%, Ramp data shows," November 11, 2026 (corroboration of the headline figures only; no load-bearing claim rests on it alone)Read source