Policy playbook

Plan A is a set of narrow technical commitments, not one large institution. Each can look too small to matter. Together they let a government change providers, audit a deployment, and keep a service running under stress.

The capacity commitment is modest. National programmes offer one another reciprocal access and maintain a queue for public-interest work. Its value lies in the fallback it creates when a commercial provider withdraws, reprices, or fails.

Evaluation costs the most and matters most. Accreditation, mutual recognition, and a funded pipeline for displaced digital-service workers give disclosure rules teeth, because people in the region can test a provider's claim.

The energy commitment protects the rest. New large loads should fund their own firm renewables and network reinforcement, then publish water draw. Those terms give governments an answer when datacentre politics turns ugly. Without them, they have a press release.

Plan B is the likely result when regional consensus does not arrive. ASEAN's structure rarely produces binding agreement quickly, and countries can still protect essential services through bilateral deals.

The aim is to prevent the worst outcome. Governments refuse exclusivity with any provider, even on generous terms. They keep two viable inference paths for critical services and write contracts that make switching costly but possible.

This weakens their negotiating position. Five states approaching the same provider separately get five standard contracts. The same five, working together, can amend the terms. Plan B trades that advantage for speed and domestic control.

It can also close off later coordination. Each sensible bilateral deal narrows the room for a regional agreement, until the option disappears just as the appetite for it returns.

Plan C is the most likely path. Governments wait until a shock becomes politically undeniable, then act under pressure.

Its costs are predictable. Transition programmes launched after layoffs reach fewer people and cost more per placement. Emergency compute procurement buys buildings instead of control. Retroactive terms on existing datacentre approvals invite disputes and deter the investment the region wants.

Preparation makes Plan C less damaging. A drafted portability specification, an accreditation scheme on paper, and a named list of critical workloads cost little before a crisis. None demands political capital until it is needed.

The low-cost preparatory work in Plan A still helps under Plan C. It is worthless if it starts only after the shock.