Skip to content
Let's talk
Nick Patience

Nick is Vice President & Practice Lead, AI Platforms. He is a thought leader on the development, deployment and adoption of AI — an area he has been researching for 25 years. Prior to Futurum, Nick was a Managing Analyst with S&P Global Market Intelligence, with responsibility for 451 Research’s coverage of Data, AI, Analytics, Information Security and Risk. Nick became part of S&P Global through its 2019 acquisition of 451 Research, a pioneering analyst firm Nick co-founded in 1999. He is a sought-after speaker and advisor, known for his expertise in the drivers of AI adoption, industry use cases, and the infrastructure behind its development and deployment. Nick also spent three years as a product marketing lead at Recommind (now part of OpenText), a machine learning-driven eDiscovery software company. Nick is based in London.

Predictions

Sovereign AI's Substance Test Arrives

By the end of 2026, the principle that AI model access is revocable by originating jurisdictions, demonstrated by the June 2026 US export directive, will become a critical factor in mainstream enterprise procurement. Enterprises can no longer assume “sovereign” labels provide immunity from revocation. This directive proved that, regardless of regional infrastructure or contracts, access remains subject to unilateral control. Consequently, regulated buyers will now scrutinize every AI service for this risk. Vendors unable to provide a credible continuity strategy for their sovereign-label offerings will face exclusion from regulated-sector tenders, as compliance and procurement teams prioritize jurisdictional resilience over simple hosting location.

Watch on Vimeo

Why This Is Trending

  • The Test Arrived Earlier and Harder than the Market Expected: A US export control directive suspended global model access for foreign nationals, including provider staff. This event proved that originating jurisdictions maintain unilateral control over model access, invalidating the perceived immunity of sovereign infrastructure or contracts.
  • Enterprise Buyers drew the Wrong Conclusion: Many incorrectly dismissed this as a frontier-model issue unlikely to affect mainstream tiers. This misses the structural reality: the jurisdictional-reach principle established by the directive is tier-agnostic, applying to any model from any US provider, regardless of capability. And the fact the ruling was eventually rolled back also doesn’t matter; it happened, and it has changed the risk assessment for AI.
  • Sovereign-region Infrastructure does not address this Vulnerability: While sovereign-AI discussions focus on data residency and infrastructure control, the recent directive introduced a critical risk: originating jurisdictions can unilaterally revoke model access, bypassing regional infrastructure protections. Even a model hosted on sovereign infrastructure in Frankfurt remains dependent on its US-originating provider.
  • Regulated Sectors Cannot Absorb this Risk Quietly: Regulated entities such as banks and energy providers cannot risk unilateral model revocation. As these organizations shift from pilot to production, legal and compliance teams will increasingly scrutinize this operational vulnerability, intensifying the pressure originally deferred during procurement.

Vendor Impact

  • Hyperscalers Face a Question Their Sovereign Tiers were not Designed to Answer: Sovereign-region offerings were built to address data residency, operational access, and jurisdictional governance of the data path. They were not designed to address model-access revocation by the originating government. Providers will need to explain credibly what continuity protections exist if a directive of the June 2026 type were applied to their mainstream model tiers; “it won’t happen to us” is not a procurement-grade answer.
  • European Sovereign-cloud and Private-cloud Specialists have a Sharper Argument than before: Vendors offering deployable models in genuinely customer-controlled environments, where the customer holds the weights, runs the inference, and does not depend on a live connection to a US-headquartered provider, can now point to a live event rather than a hypothetical risk. Their market just got more credible, and the regulated-sector buyer who dismissed on-premises or sovereign-cloud AI as over-engineered has a new reason to revisit that call.
  • Model Providers Face a Portability Question: The ability to offer genuinely deployable model versions – weights that can run in customer-controlled or regionally isolated environments, not solely via a central endpoint – becomes a procurement differentiator in regulated sectors. Providers whose models are only available as a hosted API are exposed to the revocation risk in a way that providers offering deployable weights are not.
  • AI Platform Vendors need a Continuity Architecture: Policy-aware inference routing – automatically deciding whether a workload can run on a global endpoint, must stay in-region, or must execute in a customer-controlled environment – now needs a fourth dimension: what happens if the primary model endpoint becomes unavailable due to a government directive. Failover, model substitution, and continuity planning become platform requirements rather than edge-case considerations.
  • Systems Integrators have a new Practice Line: Regulated-sector clients will need their AI workloads mapped against revocation risk, not just data-residency requirements. Designing continuity architectures for AI services, i.e., what runs where, what the failover path is, what the contractual protections are, becomes a distinct and billable engagement, separate from the sovereignty compliance work that was already in motion.

Latest from Nick Patience

All work by Nick Patience