Podcast: sovereign VP says order book 'is a power queue now,' confirms staged termination charges
Sovereign demand is a repeatable class carrying a power queue with penalties
In a recorded interview on an infrastructure-industry podcast, NVIDIA's VP of Sovereign Programs, Aisha Bello, said sovereign demand 'is not a backlog in the traditional sense — it is a power queue,' with national programs structuring purchases around grid interconnection timelines of three to five years. Asked about cancellation risk, she said contracts now include staged commitments with termination charges, 'which changes the character of the order book,' and that the company reports sovereign progress by energized megawatts rather than contracted dollars. She declined to quantify the sovereign share of data center revenue. On competition, she described 'a healthy multi-vendor floor for inference' while saying training demand remains concentrated. Bello confirmed NVIDIA participates in two national standards bodies working on compute procurement language and that the company has hired transmission-planning specialists in three regions. The interview was published without a transcript; the host noted Bello's team declined questions about specific countries.
A power queue with staged termination charges and megawatt-based progress reporting is a materially different order book from the two-to-three-quarter hyperscaler cycle. The refusal on sovereign revenue share is the disclosure to track.
'Not a backlog in the traditional sense' is a sentence that answers a question nobody asked on air but everyone is asking off air. And 'energized megawatts' is a metric where the utility can take the blame.
- Sovereign demand framed as a power queue gated by 3-5 year interconnection timelines
- Contracts include staged commitments with termination charges
- Progress reported by energized megawatts rather than contracted dollars
- Transmission-planning specialists hired in three regions