Panel
Sovereign Stacks: Building Trusted AI on National Terms | RAISE Summit 2026
RAISE SummitMohsen Moazami, Thomas Di Giacomo, Cory Hawkvelt, Simon Muskett, Gleb Budman, Claire Thomas-Daoulas
- Europe is perceived as potentially behind the early race for sovereign models, with a view that replicating existing cloud solutions may be too late given multi-year legislative timelines and the futility of rebuilding current major providers like AWS.
- Future demand for compute is expected to reach levels requiring $100 billion orders, forcing providers to accept contracts in remote locations, while infrastructure roles will shift from storage to managing complex decisions regarding networks, cooling, and compliance for clusters of 10,000 GPUs.
- The terminology and concepts surrounding data control will evolve every few years as vendor lock-in transforms, potentially pivoting the debate from "sovereignty" to "resilience" or managed economic interdependency by 2030 or 2035.
- AI adoption faces risks where many pilots fail to achieve millions in ROI, and an over-reliance on regulatory compliance could cause Europe to miss business opportunities compared to the US and China, which are driven by massive demand.
- Enterprise strategies are predicted to require nuanced multi-model approaches using up to 20 different models with varying security and cost requirements, while "control and choice" will emerge as a critical decision driver alongside performance and cost.
- Agentic workloads are anticipated to emerge in the next few years, significantly altering sovereignty definitions and workforce dynamics, with innovation increasingly guided by US foreign policy constraints forcing competitors to develop independent hardware and models.
- Business and technology needs, rather than regulations, are expected to drive the adoption of sovereign solutions, as regulations alone cannot enforce sovereignty and restrictive geographic or vendor requirements may limit effective data utilization.
- Predictions regarding the specific state of the political and technological landscape in 2035 are considered speculative, and while innovation is driven by constraints, forcing pragmatic goals unrelated to pragmatic outcomes will limit engineering creativity.