Businesses on Ramp used 37.1% more AI tokens in the week ending 23 October than in the same week a month earlier, while their token spending rose 8.9%, according to Ramp’s AI Index. By our arithmetic, spend per token fell about 21% in four weeks.
On the other side of the build-out, Alphabet’s quarterly filing lists $924.4 billion of material purchase commitments as of 30 September, with $213.4 billion due within 12 months. Its third-quarter capital spending was about 121% of its operating cash flow, by our arithmetic.
Short explainer · AI Spend per Token -21%
A vertical news Short summarising this story. Figures come from Ramp’s AI Index and Alphabet’s Q3 2026 release and 10-Q; our arithmetic is labelled.
Business Pill · Free Cash Flow
An evergreen Business Engineer explainer on free cash flow. It teaches the general idea only and does not discuss Ramp, Alphabet or any company in this piece.
The key insight: By our arithmetic, businesses on Ramp paid about 21% less per AI token in four weeks, while Alphabet lists $924.4 billion of purchase commitments, $213.4 billion of them due within a year.
Ramp says “Total token spending rose 8.9% compared with the same week last month”, and that total token volume rose 37.1%. Volume grew about four times as fast as spending, by our arithmetic.
Ramp’s table gives a blended price per million tokens for OpenAI and Anthropic of $0.56 in the week ending 23 October, down 26.8% from $0.76. Its blended price for OpenAI fell 42.4%, to $0.46, and for Anthropic 15.7%, to $0.61.
Our 21% figure divides Ramp’s spend growth by its volume growth. It differs from Ramp’s 26.8% because Ramp computes its totals and its blended price separately. The data covers businesses that pay through Ramp, not the whole market.
Ramp AI Index, week ending 23 Oct 2026 vs week ending 25 Sep 2026: token volume and token spending; and Alphabet’s material purchase commitments as of 30 Sep 2026 against its Q3 2026 operating cash flow. Spend per token and ratios are FourWeekMBA arithmetic.
What Alphabet’s Filings Show
Alphabet’s earnings release shows $51.4 billion of purchases of property and equipment in the third quarter against $42.3 billion of net cash from operating activities. That is about 121%, by our arithmetic, and the release reports free cash flow of minus $9.1 billion for the quarter.
Its Form 10-Q says $802.1 billion of the $924.4 billion in commitments relates to “multi-year commitments for technical infrastructure and inventory, energy take-or-pay contracts, and content licenses”. It adds $130.5 billion of future payments for leases not yet commenced, mainly for data centers and energy infrastructure.
The $213.4 billion due within 12 months is about five times Alphabet’s third-quarter operating cash flow, by our arithmetic.
How the Two Sets of Numbers Meet
Ramp’s figures describe what businesses on Ramp pay for AI tokens, with Anthropic at 60.5% of that spend and OpenAI at 35.1% in the latest week; Alphabet’s describe what one of the builders has committed to spend on infrastructure. Neither source links the two.
As we read Ramp’s table, usage and spending are moving at different speeds: token volume rose 37.1% while token spending rose 8.9% over the same weeks.
As we read Alphabet’s filings, the build side is set in multi-year contracts: $802.1 billion of the commitments relates to multi-year infrastructure, inventory, energy and content agreements, per the 10-Q.
As we read the two together, price per token and commitments are measured on different clocks: Ramp’s comparison spans four weeks, and Alphabet’s commitments run largely through 2031, per its filing.
Alphabet Form 10-Q, Q3 2026
“The significant majority of these commitments will be fulfilled through 2031.”
Three Implications
FOR BUYERS OF AI Ramp’s table shows OpenAI’s blended price per million tokens down 42.4% and Anthropic’s down 15.7% over the four weeks.
FOR ALPHABET The release reports third-quarter free cash flow of minus $9.1 billion, after $51.4 billion of purchases of property and equipment.
FOR THE BUILD-OUT The 10-Q lists a further $130.5 billion of future lease payments, mainly for data centers and energy infrastructure, not yet commenced.
The framework sets out to map the financing of AI: “How the money flows, where it cracks, what it buys, who can absorb it”.
As we read it, this piece puts two of those questions side by side: what buyers pay per token, from Ramp’s data, and what one builder has committed to spend, from Alphabet’s filing. The framework was published in May 2026 and does not discuss Ramp’s data or Alphabet’s quarter.
Ramp’s comparison spans four weeks and one sample of companies; it does not show a trend over a year. Ramp does not report Google’s token prices, and Alphabet’s filings do not give its own price per token.
The 10-Q describes the commitments as subject to counterparties performing on their obligations and does not say how they would be funded.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
In Ramp’s data, businesses used 37.1% more tokens and spent 8.9% more, about 21% less per token, by our arithmetic. Alphabet, which spent about 121% of its quarterly operating cash on property and equipment, lists $924.4 billion of purchase commitments, $213.4 billion of them due within a year.
A note on sourcing. Token figures come from Ramp’s AI Index as read on 31 October 2026; Alphabet figures from its third-quarter 2026 earnings release and Form 10-Q. Spend per token and the ratios are our arithmetic. We did not contact Ramp or Alphabet.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
