I spent a decade telling boards the public cloud was the destination. The 2026 data says that framing was the expensive mistake — not the cloud itself.
A mandate is not a strategy
"Cloud-first" became policy without unit economics. Broadcom's Private Cloud Outlook 2026, surveying 1,800 IT leaders across eight countries, finds 83% considering repatriation and half already pulling workloads back. Barclays pegs it at 86% planning at least one move off public cloud — its highest figure ever. This isn't backlash; it's the unwinding of an absolutism.
What the reversal actually corrects
- Steady-state workloads don't need elasticity — they need predictable cost. 97% of IT leaders admit some public cloud spend is wasted, and 52% say more than a quarter of the budget delivers no value. Egress fees are the top culprit (46%).
- Data gravity flipped the AI math. 56% now run production inference on private infrastructure versus 41% on public — a near reversal from a year ago.
We optimized for adoption speed, not operating economics. The bill arrived in year three.
I lived this. A lift-and-shift I championed looked brilliant at go-live and brutal two quarters later, when steady-state storage growth outpaced every forecast and the invoice stopped matching the business case.
Placement beats ideology
Repatriation isn't retreat — it's portfolio discipline. Sovereignty now drives 82% of infrastructure decisions, and the enterprises doing this well keep spiky, elastic workloads in cloud while moving residency-bound or latency-sensitive data to infrastructure they control.
The transformation was never "to the cloud." It was building an operating model that survives the invoice. Before your next "X-first" mandate, demand the unit economics: cost per transaction, egress, residency, latency. Where the workload lives matters less than whether the math still works in year three.