🤚 The Open-Palm Capacity Report

In a development that will make every Azure sales engineer reach for the bourbon, Microsoft is quietly routing GitHub infrastructure through Amazon Web Services because its own cloud platform cannot keep up with the AI-driven surge in developer activity.

According to Business Insider, the company that spent $7.5 billion acquiring GitHub in 2018 — explicitly to funnel the world’s developers into the Azure ecosystem — is now paying its biggest cloud competitor to keep the lights on. The culprit? Agentic coding. The very AI tools Microsoft has been aggressively promoting are consuming infrastructure faster than Azure can provision it.

Key details:

• GitHub’s AI-driven usage has surged beyond what Azure alone can support
• The platform is experiencing strain from agentic coding workloads — AI agents writing, reviewing, and deploying code autonomously
• Microsoft’s original integration strategy called for GitHub to run entirely on Azure
• AWS is now supplementing capacity for the platform that hosts over 200 million repositories

👐 The Two-Handed Strategic Irony

Let us appreciate the full, exquisite geometry of this situation.

Microsoft bought GitHub to make it an Azure showcase. Microsoft invested $13 billion in OpenAI to lead the AI revolution. Microsoft launched GitHub Copilot and then Copilot Workspace and then agentic coding features that let AI agents spawn other AI agents to write code at industrial scale. And now Microsoft is discovering that when you tell every developer on Earth to let AI write code for them, someone has to run the AI.

That someone, apparently, is Andy Jassy.

This isn’t just an infrastructure story. This is the single most expensive product-market fit problem in enterprise history. Microsoft succeeded too well at making developers dependent on AI — and its own cloud couldn’t metabolize the demand. The company is essentially paying AWS a toll to cross a bridge it built.

There’s a term in business school for this. It’s called “vertical integration,” and it’s supposed to go in one direction.

🌿 The Gentle Awakening

Somewhere in a Redmond conference room, someone is standing in front of a whiteboard with two columns. One says “Azure Revenue From GitHub” and the other says “AWS Invoices For GitHub.” The gap between them is a philosophical question about what it means to own a platform in the age of AI.

Because here’s the quiet truth no hyperscaler wants to discuss: AI workloads are not like regular workloads. They don’t scale linearly. They don’t respect capacity planning cycles. They arrive in bursts shaped like exponential curves, driven by agents that spawn agents that spawn agents. When your product is an autonomous system that generates its own demand, your infrastructure team doesn’t have a forecasting problem — they have an existential one.

Microsoft built a product that creates infinite demand for compute, then discovered that infinite demand requires infinite infrastructure, and infinite infrastructure requires the one company they were trying to make irrelevant.

👑 The Gold-Leaf Cloud Economics Lecture

The broader implication is one that should concern every CTO who just signed a five-year cloud commitment: the agentic era may break the single-cloud model entirely.

If Microsoft — a company that owns both the AI layer and the cloud layer and the developer platform — cannot run its own AI workloads on its own cloud, what hope does your Series B startup have of staying within budget on a single provider?

The answer, increasingly, is none. Multi-cloud was supposed to be a best practice. It’s becoming a survival requirement. Not because you want redundancy, but because no single hyperscaler has enough GPUs to run the AI future that all of them are selling.

The age of cloud loyalty is over. The age of cloud desperation has begun. And the first company to admit it out loud was the one that owns the cloud.

“We spent $7.5 billion to acquire the world’s developer platform and $13 billion to lead the AI revolution, and the punchline is a monthly invoice from Jeff Bezos’s retirement project.” — The Slap of Wisdom Cloud Economics Desk, updating its multi-cloud strategy from a GitHub Codespace running on AWS