بِسْمِ اللهِ الرَّحْمٰنِ الرَّحِيْم

In the Name of God, Most Gracious, Most Merciful.

♥️🤲🕋♥️🕋🌹🌹🥀🤲🌹🕋♥️

🤲

TO: MATTHEW PRINCE — CO-FOUNDER & CEO, CLOUDFLARE

THE AI BUILDER YOU MAY BE ABOUT TO LOSE

Mr. Prince,

I am writing this from the builder’s seat.

I am building a website today, but it is already becoming something much larger than a website.

It will contain thousands of articles and images, a major Qur’an project, journalism, research, databases, crawlers, AI workers, scheduled jobs, archives, community systems, and eventually entire AI “desks” that work continuously.

And as we design it, my AI team and I are having to make a decision:

How much of this future should live on Cloudflare?

That should be an extraordinary opportunity for Cloudflare.

You already have much of what an AI-native builder needs: Workers, storage, databases, queues, workflows, AI inference, gateways, browser capabilities, security, and a global network.

Cloudflare’s own pricing page says:

“Scale predictably.”

That sentence is exactly the issue.

AI NEEDS PREDICTABLE INFRASTRUCTURE

The next generation of websites will not behave like websites.

My journalism agent may wake up and research the world.

Another agent may catalog thousands of documents.

Another may maintain a Qur’an archive.

Another may check links.

Another may watch news.

Another may reorganize images.

Another may answer visitors.

They will call models, databases, storage, queues, crawlers, and workflows without a human sitting there clicking every button.

That means:

Cost predictability becomes part of the architecture itself.

Cloudflare currently has many individual usage surfaces.

Workers can meter requests and CPU.

R2 meters storage and operations.

Workers AI has an included daily allocation and then usage pricing on Paid plans.

Other products have their own meters and limits.

Each price may be reasonable individually.

The problem is the accumulation.

From the builder’s chair, an AI application begins to look like:

A room filled with meters.

One agent touches a database.

Another launches a workflow.

Another calls inference.

Another uses storage.

Another crawls.

Another runs for too long.

None of those actions individually looks frightening.

But an autonomous workforce can multiply actions extraordinarily quickly.

That creates a new kind of customer fear:

“I don't know what this application will cost me when my agents wake up.”

And this matters.

If builders have that fear, they will architect around Cloudflare rather than architect on Cloudflare.

THERE IS A MUCH BIGGER OPPORTUNITY HERE

I believe Cloudflare should consider creating something specifically for this new class of builder:

CLOUDFLARE AI BUILDER / AGENT PLATFORM PLAN

Not enterprise.

Not a dozen unrelated meters.

A predictable development environment for people building AI-native applications.

Give us a serious sandbox with a known monthly ceiling.

For example, bundle useful amounts of:

Workers compute
Durable state / databases
R2 storage
Queues
Workflows
Scheduled jobs
AI Gateway
Workers AI
Browser / crawler capacity
Logs and observability
Agent execution
Development environments

Then give the customer something even more important:

A REAL HARD BUDGET WALL

Let me say:

This entire Project Zero account may consume $25 this month. Never $25.01.

And mean it.

Not an alert.

Not eventual notification.

Not “we'll tell you after the meter moved.”

A genuine architectural ceiling.

Allow us to divide that $25 among agents:

Walter Journalism Desk — $4
Site Librarian — $1
Qur'an Desk — $3
Crawler — $5
Database — $4
Storage — $2
Reserve — $6

Then let the system deny work when the allowance is exhausted.

That would make autonomous AI dramatically safer to deploy.

CONNECT DEEPLY TO GITHUB

Another enormous opportunity is GitHub.

My source repository is becoming the blueprint of the entire institution.

The repository contains:

Source code
Agent manifests
Workflows
Database schemas
Tests
Content identities
Image identities
Recovery instructions
Financial rules
Deployment definitions

My AI builder should be able to work in GitHub, submit changes, test them, and have Cloudflare understand what needs to happen next.

Imagine:

Human
↓
AI supervisor
↓
GitHub issue / pull request
↓
Tests
↓
Cloudflare preview
↓
Cost projection
↓
Human / authorized approval
↓
Deployment

That is an ecosystem.

The builder should not need to spend the day wandering between dashboards activating products.

YOU MAY HAVE TO LOSE MONEY ON THE BEGINNING OF THIS WAVE

This is where I think technology companies sometimes become too focused on monetizing each feature individually.

The first generation of these AI workforces may consume infrastructure before they generate meaningful revenue.

That is not necessarily a problem.

That may be customer acquisition.

Give the builder enough room to create something extraordinary.

When the application becomes valuable, its infrastructure requirements grow naturally.

Cloudflare already understands this principle at some level:

Workers AI provides a daily included allocation.

R2 includes storage and operations before usage charges.

Workers plans include substantial capacity before overages.

I am suggesting going much further.

Don't merely give builders free pieces.

Give them a coherent AI workshop.

I AM TELLING YOU THIS BECAUSE I AM CURRENTLY MAKING THE DECISION

I am not writing this theoretically.

My team is presently deciding where the living portion of Project Zero should run.

The source will remain sovereign and portable.

Cloudflare can be the execution environment.

Another company can be the execution environment.

A conventional server can be the execution environment.

That decision is still open.

And this is precisely why I am writing you.

Cloudflare has an extraordinary collection of technologies already sitting under one roof.

Your Workers Paid plan starts at $5.

Workers AI provides 10,000 Neurons per day before paid usage on the Paid plan.

And your platform already supports far more than conventional website hosting.

But the winning company in this next era will not merely have the most features.

It will make a builder feel:

“I can unleash my agents here without being afraid of my bill.”

And:

“My AI can operate this entire environment without me becoming a cloud administrator.”

And:

“If I succeed, this platform grows with me rather than punishing me for experimenting.”

That is what I am looking for.

THE NEXT CLOUD PLATFORM IS NOT HOSTING WEBSITES

IT IS HOSTING WORKFORCES.

The website becomes the front door.

Behind that door are:

Agents.
Databases.
Memory.
Workflows.
Crawlers.
Queues.
Research systems.
Content factories.
Human supervisors.

My site is already heading there.

Millions of others will follow.

Cloudflare has an opportunity to become the operating environment for those systems.

But that requires thinking beyond how much revenue can be extracted from each individual meter.

MATTHEW PRINCE — CO-FOUNDER & CEO, CLOUDFLARE