Deeptech · July 2026
“Sovereign” is easy to print on a pitch deck. It is hard to underwrite. The real question is who owns the capacity, who attests the stack, and who can serve regulated workloads without exporting trust.
From policy language to underwriting
Governments and regulated industries across ASEAN are tightening expectations around data residency, cross-border transfer and the processing of sensitive workloads. At the same time, AI inference and training are pulling more of those workloads into high-density compute.
That collision turns sovereignty from a communications theme into a set of underwriting questions: Where does the data sit? Who can read it in use? What happens when a foreign cloud region changes terms? Can a bank, hospital or ministry run models without surrendering plaintext?
Confidential computing as infrastructure
Encryption at rest and in transit is table stakes. The scarce capability is computation on data that remains protected while it is being used — and an architecture that makes that the default path, not a boutique mode.
That is why encrypted compute belongs in the infrastructure conversation alongside power and land. Without it, “sovereign AI” collapses into rented GPUs with a local marketing wrapper.
What Redwood is building toward
Through Platform I holdings such as ZipLogic’s encrypted compute stack, Redwood’s view is that Southeast Asia will need owned, attestable capacity — not only more imported cloud. Capital that understands both the cryptography and the regional buyer is scarce. That scarcity is the opportunity.