AI Boss Tools | Diagnostic 01 • Back to the Hub
A single gigawatt-scale AI data centre can demand around US$38 billion in capital, so most governments cannot build alone. This tool scores whether yours should build its own capability, pool it with regional partners, or buy it through strong contracts. Twelve questions produce a recommendation and a downloadable one-pager.
Sovereign AI means a nation controlling its own AI capability: the data, the models, and often the compute. More than 30 countries now fund national programmes, and the costs are real, with frontier training runs alone between US$78 million and US$192 million before staff, energy, or data centres. For most small states the honest answer is the hybrid path: sovereign control of data and open-weight models while renting compute, the route Jamaica took with Project Maestro. Answer for your government as it is today, not as you hope it will be.
Sources: CNAS Sovereign AI Index; Stanford HAI AI Index 2025 (frontier training runs cost US$78M to US$192M); Epoch AI (about US$38B per gigawatt of AI data centre capacity); Project Maestro, Jamaica, a Caribbean case study in the hybrid path.
Free. Your name and email open the full report and a one-pager sized for a board pack.