# Cloud Exit Cases

> Comparable accounts of successful, partial, abandoned, and reversed cloud exits.

---

LLMS index: [llms.txt](/llms.txt)

---

The case library supports a qualified conclusion: **cloud exit can work, but
not for every workload, organization, or moment.** Success stories, partial
moves, abandoned migrations, and returns to managed services receive the same
template and visual weight.

## Six required questions {#six}

1. **What moved?** Data, compute, storage, network, control plane, and operations.
2. **What stayed?** Services deliberately retained and why.
3. **What did it cost?** Cash, labor, interruption, and stranded commitments.
4. **How long did it take?** Decision, build, migration, stabilization, and payback.
5. **Where did it fail?** Surprises, reversals, abandoned scope, and lasting trade-offs.
6. **What would change next time?** The team's own hindsight, not the editor's guess.

## Evidence labels {#labels}

Every case distinguishes **reported** facts, **modeled** values, **verified**
artifacts, and **editorial inference**. A modeled saving is not renamed as an
audited saving. A first-party story is valuable evidence, but its applicability
boundary must travel with it.

The first calibration candidates are the DHH/37signals exit, Tantan's
infrastructure economics, Ahrefs, and publicly documented partial exits. They
will not appear as completed cases until sources and timelines are reconciled.