Software companies now sell AI work in credits. Salesforce lists Agentforce Flex Credits at $500 per 100,000, or $0.005 a credit, and Adobe, Atlassian and Notion list theirs at about $0.004 to $0.01 a credit, by our arithmetic from their pricing pages.
Figma took the other side of the trade. Its AI credit FAQ says customers get “2x more credits on the Professional plan, and 1.6x more credits on the Organization and Enterprise plans” at the same cost, which halves the Professional price per credit, by our arithmetic. Its third-quarter hosting costs rose $52.8 million, about 45% of its added revenue, as we reported in Figma’s Hosting Costs Rose $52.8M, 45% of Its Added Revenue.
Business Pill · Margin Squeeze
An evergreen Business Engineer explainer on margin squeeze. It teaches the general idea only and does not discuss Figma or any company named here.
The key insight: As we read it, software companies sell AI work in two ways: a meter at about $0.004 to $0.01 a credit, by our arithmetic from the Salesforce, Atlassian, Notion and Adobe pricing pages, or a larger allowance at the same price, as Figma chose while its third-quarter hosting costs rose $52.8 million year over year.
Every AI feature runs on inference, and someone pays for it. Figma’s 10-Q for the quarter to 30 September says its cost of revenue consists primarily of “technical infrastructure and hosting costs, including AI inference”. A software company can meter that cost back to customers in credits, or carry it in its own margins.
The pricing pages we read show both choices. Salesforce, Atlassian, Notion and Adobe publish a price for a unit of AI work. Figma raised the number of credits it includes for the same price, and its finance chief, Praveer Melwani, told CNBC: “you’re potentially trading off near-term revenue”.
List price per AI credit on each company’s pricing page on 5 November 2026: Notion ($10 per 1,000 credits), Atlassian Rovo ($0.01 per credit over the allowance), Salesforce Agentforce ($500 per 100,000 Flex Credits) and Adobe Firefly (US$9.99 for 2,000 and US$199.99 for 50,000 credits a month). Per-credit prices are our arithmetic; credits are not the same unit across companies.
The Meter: What a Credit Costs
Salesforce’s Agentforce pricing page lists Flex Credits at $500 per 100,000 and says “Flex Credits are a new unit of payment for Agentforce that align cost to the businessvalue your AI agents create”. Its own example uses 40 Flex Credits for two actions at US$0.20, which is $0.10 an action, by our arithmetic. It also lists Conversations at $2 per conversation.
Atlassian’s Rovo pricing page includes 25 credits per user a month on Jira and Confluence Standard and 150 on Enterprise. It calls Rovo credits its “unit of measurement for AI feature usage”, charges $0.01 per credit over the allowance, and lists a basic AI interaction at 10 credits, which is $0.10 at that price, by our arithmetic.
Notion’s pricing page says of its AI agents: “Free to try, then $10 per 1,000 monthly Notion credits”, which is $0.01 a credit, by our arithmetic. Adobe’s Firefly plans sell 2,000 generative credits a month for US$9.99 and 50,000 for US$199.99, about $0.0050 and $0.0040 a credit, by our arithmetic.
The Absorber: Figma
Figma’s 10-Q lists three year-over-year increases in technical infrastructure and hosting costs in the third quarter: $29.5 million in cost of revenue, $14.5 million in research and development and $8.8 million for users of its free version. Together they are $52.8 million, about 45% of the $117.2 million rise in revenue, by our arithmetic, as set out in Figma’s Hosting Costs Rose $52.8M, 45% of Its Added Revenue.
The credit change took effect on 25 August 2026, the FAQ says, and Figma moved its agent in Figma Design to general availability on 6 October. The FAQ does not list a per-credit price; it says pricing is shown in each customer’s admin dashboard. The share-price reaction and Melwani’s remarks are in Figma Gave Up to 2x AI Credits at Same Price; Shares Fell 7%.
What the Two Models Do to Margins
As we read the documents, a meter moves inference cost onto the customer’s bill, while an allowance keeps it in the vendor’s cost of revenue. Figma’s 10-Q shows the second case: its AI-related hosting increases sit in cost of revenue, research and development, and the cost of serving free users.
The 10-Q also names an agent risk: “errors or unintended agent actions” that could “expose, alter, or destroy customer code or data”. As we read it, that risk grows with agent use: an agent that acts inside customer systems carries an operating risk next to its inference cost. Figma’s third-quarter results and its decision to cover inference for its agent are in Figma Revenue Up 43% as It Covered AI Agent Inference Costs.
The Business Lever
The lever is the unit of sale. A credit priced at about $0.004 to $0.01 ties revenue to the AI work a customer runs. A larger allowance at the same price, as Figma chose, keeps the customer’s bill flat and moves the cost of extra AI work into the vendor’s own cost lines.
What to Watch
Pricing pages and filings carry the signals. A change in credits per action, in allowances per seat, or in the price of a credit pack shows who is absorbing inference cost. In filings, the line to watch is technical infrastructure and hosting costs, including AI inference, against revenuegrowth.
As we read the pricing pages, a credit moves the cost of each AI action onto the customer’s bill: Salesforce’s own example prices two actions at US$0.20.
As we read Figma’s FAQ next to its 10-Q, the company chose the opposite: 2x more credits on its Professional plan at the same cost, while AI-related hosting increases ran through cost of revenue, research and development and its free tier.
As we read Melwani’s remark to CNBC, Figma frames the choice as a trade between near-term revenue and long-term growth.
Praveer Melwani, Figma chief financial officer, to CNBC, 4 November 2026
“What that means is you’re potentially trading off near-term revenue, investing in your customers, and then allowing yourself to then grow durably over the long term”
Three Implications
FOR SOFTWARE BUYERS Salesforce lists Flex Credits at $500 per 100,000 and Atlassian charges $0.01 per credit over the allowance, so AI use shows up as a metered line.
FOR FIGMA Its 10-Q lists $29.5 million, $14.5 million and $8.8 million of year-over-year increases in technical infrastructure and hosting costs in the third quarter.
FOR PRICING COMPARISONS A credit is a different unit at each company: Salesforce meters actions, Atlassian interactions and API calls, Adobe generative features.
The framework says: “The new job of finance is to manage the ratio between the two — work produced per token burned.”
As we read it, a credit meter and a free allowance are two ways of deciding who pays for the tokens burned: the customer per action, or the vendor in its own cost lines. The framework does not discuss Figma, Salesforce, Atlassian, Notion or Adobe.
What Is Not Established
The pricing pages list prices; they do not show discounts, contract terms or how many credits customers use, so the per-credit figures are list prices, not what any customer pays. A credit is not the same unit across companies: Salesforce meters actions, Atlassian meters interactions and API calls, and Adobe meters generative features.
Figma does not publish a per-credit price on the FAQ page we read, so its halving is a ratio from the FAQ, not a dollar figure. We make no forecast on any company’s pricing or margins.
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.
Salesforce, Atlassian, Notion and Adobe meter AI work at about $0.004 to $0.01 a credit, by our arithmetic from their pricing pages. Figma gave customers 2x more credits on its Professional plan at the same cost while its third-quarter hosting costs rose $52.8 million year over year, about 45% of its added revenue.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. Prices are list prices from the Salesforce, Atlassian, Notion and Adobe pricing pages as read on 5 November 2026. Figma figures are from its Form 10-Q for the quarter to 30 September 2026 on SEC EDGAR and its Help Center FAQ; the CFO’s remark is from CNBC. Per-credit prices, ratios and totals are our arithmetic.
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.
