⚠ Abstraction & Multi-CloudLow threat

The switching-cost moat assumes workloads are hard to move, and a wave of technology is working to make them less so. Containers and Kubernetes let software be packaged in a portable form¹ that runs on any cloud; open-source databases and tools reduce reliance on any one provider's proprietary services; and a generation of enterprises, wary of being captive, now deliberately design for 'multi-cloud' so they can move or split workloads between providers. Each trend chips at the entanglement that keeps AWS revenue sticky.

The danger is that portability turns cloud into more of a commodity, where customers play providers against one another on price. If an enterprise can credibly threaten to move a workload, AWS's pricing latitude thins, and the annuity-like quality of the revenue weakens. The most sophisticated customers, who spend the most, are precisely the ones most able to architect for portability.

In practice, true portability remains more aspiration than reality for most. Multi-cloud in practice usually means using several clouds for different things, not freely moving workloads between them, and the deepest, most valuable services — the managed databases, the AI platforms, the analytics — remain stubbornly proprietary and sticky. Abstraction moves the commodity layer; it has not moved the high-margin core.

Low-to-moderate, on balance. Portability genuinely thins the lock-in at the commoditized bottom of the stack and gives big customers more leverage than before — but the services that earn the fat margins remain hard to leave, and the friction of actually moving keeps most enterprises right where they are.