The New Economics of Code

01 / SummaryWhat this report actually answers

AI coding tools have gone from novelty plug-in to default install. In 2025 the category was worth roughly US$8 billion, defined by GitHub Copilot, a Cursor valued at US$29.3 billion, and a transforming Windsurf/Devin.

But once adoption plateaus, the real questions begin: where does the money come from, and how much margin is left. This is not a product review, and it does not re-narrate launch events. It takes the chain of growth, retention and unit economics apart: which four layers revenue is built from, how much ARR each layer can carry, how much of gross margin inference cost consumes, and whether the moat sits in the model or in distribution.

Every quantifiable figure carries its source and a reliability grade — verified disclosure, industry estimate, or illustrative model — backed by 15 traceable primary sources. You may disagree with the conclusions, but you can check every number yourself.

02 / ContentsNine chapters, one complete business logic

1. 01Executive summary: five core judgementsPlateau, reshuffle, monetization
2. 02Key metrics: eight numbers that matterAdoption, users, ARR, valuation, market size
3. 03Growth curve: how a $10B category expandsVerified anchor vs 2030 forecast range
4. 04Competitive quadrant: where five player types sitDistribution × product/agent capability
5. 05Monetization: the four-layer revenue stackSeat → enterprise → usage → model spread
6. 06Unit economics: where a monthly fee goesCost structure and margin erosion
7. 07SWOT: strengths, weaknesses, opportunities, threatsIncluding the trust gap
8. 08Outlook: a three-stage evolutionCompletion → chat → agents → outcome pricing
9. 09Full source list15 items, each graded for reliability

03 / SampleFour numbers first — then decide

These four come from chapter two, "Key metrics". Each is a first-party disclosure or an authoritative survey, so you can verify them independently.

The chapter also covers Cursor's three-step ARR climb (US$100M → 500M → 1B), Windsurf's peak ARR of US$82 million, and a note on how analysts arrive at Copilot's ARR estimate.

04 / ReliabilityEvery number can be checked

The report grades data at three levels and labels each one inline, never mixing them:

Verified public data (the Stack Overflow 2025 Developer Survey, GitHub Octoverse, Cursor's official blog, MarketsandMarkets and others) — original source and date given.
Industry estimates (such as Copilot ARR or the broad CB Insights market figure) — explicitly marked "not an official GAAP disclosure".
Illustrative or modelled figures (2026–2030 extrapolations, the unit cost breakdown) — explicitly marked "explains the logic, not a financial disclosure".

Private companies do not publish gross margin, so every cost structure in the report is labelled illustrative. That is not vagueness — it is putting which part is fact and which part is inference on the table.

05 / FitWho it is for

AI founders and product leads building coding agents, developer tools or AI middleware, who need the real constraints on pricing and margin.
Investors and analysts who need a traceable data chain instead of a pile of conflicting second-hand citations.
Corporate strategy teams deciding how much room seat-based pricing still has, and which slice model vendors will take.
General readers who want to understand the category: no technical background required, everything is written in business language and every term is explained.

06 / DeliveryHow to buy, how you get it

1. Email us to order

   Write to hello@neuroai.site with the subject "Report order" and your email address. We reply within two hours with the payment options.

2. Pay $30

   By PayPal, card or WeChat Pay — then send us the receipt. Invoices are available on request.

3. Get your licence code

   Within an hour (usually minutes) you receive a unique licence code. Enter it on this page to read the report online, or print it to PDF in one click.

07 / FAQ