NET · Forward model · Agentic Internet (Act 4) · Bear case
What has to happen in Agentic Internet (Act 4)
Model as of
This page changes Agentic Internet (Act 4) inside the complete NET model. Every other vertical, corporate programme and valuation assumption stays in the selected portfolio case.
Shares this vertical and portfolio case. Slider and horizon edits stay in your browser.
Agentic Internet (Act 4)
Basis quarter$0
Final quarter$178M
Final revenue mix13%
Charging machines to use the internet: Pay Per Crawl for AI crawlers, the Monetization Gateway extending that to any API, dataset or MCP tool call, and the x402 payment rail underneath both. Cloudflare sits in front of roughly a fifth of the web, which makes it one of the few companies that could impose a toll booth. Today the whole layer earns approximately nothing and is justified internally as demand generation for Acts 1 to 3.
Last four quarters
2025 Q3
$0
Estimated
2025 Q4
$0
Estimated
2026 Q1
$0
Estimated
2026 Q2
$0
Estimated
Pay Per Crawl — publishers charging AI crawlersMonetization Gateway — per-call charging for APIs, data and MCP toolsx402 payment rail take rateAgent identity and authorisation
Subscribers
150K
0.8% of a 20.0M addressable base
150K properties monetising machine traffic when the line opens — a sliver of a paying base in the millions.
Addressable subscribers
20.0M
the S-curve ceiling
20M internet properties on Cloudflare with content or an API worth charging a machine to reach.
Net adds
60K/qtr
ramping toward 400K/qtr, throttled as the base approaches the TAM
60K properties a quarter at the start. Turning it on is a switch, so trust gates adoption, not engineering.
Net-add ceiling
400K/qtr
what supply can deliver at full rate
400K a quarter once charging machines is normal rather than novel, not a step beyond a secondary line.
ARPU
$20.00/mo
drifting +2.0% per quarter, floor $5.00
$20 a month per monetising property — a take rate on small payments, not a price Cloudflare sets.
Non-subscriber revenue
$0/qtr
growing 0.0% per quarter
Zero. Any enterprise contract for agent identity or authorisation would land here; none is modelled.
Agentic Internet (Act 4)
Latest: $178M (2031Q2E)
| Period | Value |
|---|---|
| 2025Q2 | $0.00 |
| 2025Q3 | $0.00 |
| 2025Q4 | $0.00 |
| 2026Q1 | $0.00 |
| 2026Q2 | $0.00 |
| 2026Q3E | $0.00 |
| 2026Q4E | $0.00 |
| 2027Q1E | $0.00 |
| 2027Q2E | $0.00 |
| 2027Q3E | $13M |
| 2027Q4E | $19M |
| 2028Q1E | $27M |
| 2028Q2E | $35M |
| 2028Q3E | $44M |
| 2028Q4E | $55M |
| 2029Q1E | $65M |
| 2029Q2E | $77M |
| 2029Q3E | $89M |
| 2029Q4E | $101M |
| 2030Q1E | $113M |
| 2030Q2E | $126M |
| 2030Q3E | $139M |
| 2030Q4E | $152M |
| 2031Q1E | $165M |
| 2031Q2E | $178M |
Assumptions & reasoning
- Zero across every historical quarter is a statement, not a placeholder. Pay Per Crawl and the Monetization Gateway do collect money today, but nothing Cloudflare has disclosed suggests it is material, and the company frames Act 4 as demand generation for the other three. Carving a number out of the reported total to represent it would be inventing the disclosure this model exists to avoid.
- Modelled on monetising properties x take rate because Cloudflare is a toll collector here, not a seller. It does not set the price of a crawl or an API call; it takes a slice of whatever the publisher charges. That makes property count the adoption question and revenue per property the volume question, and they are worth arguing about separately.
- This is the only vertical that could be worth more than the rest of the model combined and the only one that could be worth nothing. If agents pay for content, the company sitting in front of a fifth of the web collects a fraction of every transaction. If publishers cannot make crawlers pay, or the AI labs route around Cloudflare, the entire line is zero and the base case is overstated by whatever it contributes.
- The margin path is the aggressive part. It starts at -25% and reaches 55%, above every other line, on the reasoning that a payment rail on infrastructure that already exists has almost no marginal cost. That is true of the rail and not of the disputes, fraud and support that come with handling other people's money.
- The base case keeps this deliberately secondary: about $295M a quarter by 2031, roughly a ninth of revenue, against a first quarter of $15M in 2027 Q3. That is a real business and not a re-rating. The case where Act 4 is the whole story is the Bull case, where it reaches about a fifth of revenue — and even that is an assumption about publishers and agents, not about Cloudflare.