Karl Havard, COO, Nscale on 3 Things Sovereign AI Must Deliver: Control, Compliance, and Growth

Nscale's Sovereign AI Definition

• Carl (CCO at Nscale) defines "sovereign AI" as a full-stack approach spanning energy, data center construction, infrastructure deployment, and model hosting.
• The organization positions itself as an alternative to US hyperscalers for organizations and governments in Europe and the UK.

Industry Shift from Legacy Cloud to AI Infrastructure

• The cloud industry has transitioned from an era where it resolved on-premise burdens (patching, testing, Capex) to a new "level set" driven by AI hardware requirements.
• Hosting advanced accelerators (NVIDIA H100, H200, Blackwell, Grace Blackwell) necessitates a complete data center redesign due to high-density power needs.
• Legacy data centers cannot support current AI hardware; modern requirements mandate liquid cooling, specifically direct chip liquid cooling.

Geopolitical Drivers for Sovereignty

• Geopolitical unpredictability, cited with US President Donald Trump's actions, is forcing the EU, UK, and Far East to seek internal control over AI infrastructure.
• The Patriot Act (Cloud First Act): US-centric services allow the US government to access data at any time; an example cited is the denial of access to the International Criminal Court prosecutor, Karim Khan, via Microsoft Teams.
• Physical Threats: Russian trawlers are reportedly hovering over undersea internet cables in the Atlantic, highlighting vulnerability in physical connectivity.
• US Investment Dominance: The "Stargate" initiative involves a $500 billion US investment in AI, creating a competitive barrier for European organizations seeking access.
• Potential Tariffs: Cloud services may soon face US tariffs, adding financial risk to cross-border reliance.

Three Pillars of Sovereign AI Strategy

• Control: Requires the ability to develop, deploy, and manage infrastructure entirely within sovereign territory.
  • Infrastructure must include the physical energy source, data center ownership, and the management of AI models domestically.
  • Verification criteria: Organizations must be able to physically visit ("touch") their servers; if a passport is required to access the site, sovereignty is compromised.
  • Pricing and tariffs must be set by the sovereign entity, not a foreign jurisdiction.
• Compliance: Extends beyond GDPR to include nation-specific and industry-specific regulations (e.g., UK FCA backup strategies).
  • Data must reside within the sovereign kingdom, not just the data center's location.
  • Disaster recovery backups must remain within the sovereign territory (e.g., within 150km) rather than foreign lands.
  • Access controls must prevent foreign legal entities from gaining legal sight of sensitive data.
• Growth: AI deployment must serve as an economic vehicle to boost GDP and create jobs.
  • Data centers are catalysts for broader ecosystems, potentially generating 10-20x more jobs than the 150 maintenance roles and 500 construction roles often cited in press releases.
  • Success requires utilizing sovereign energy sources, sovereign data centers, and sovereign networking (e.g., leveraging local telcos like BT or Deutsche Telekom).

UK Government Initiatives and Talent Strategy

• The UK AI Opportunities Action Plan is launching five "AI Growth Zones," with one confirmed in Culham linked to the Atomic Energy Authority.
• The goal is to create a national grid of AI through collaboration across research institutions, startups, and incubators within these zones.
• Talent Retention: PhD research contracts (30-60 days) currently favor US tech giants (Meta, Google, Microsoft, OpenAI) offering private enterprise work; sovereign AI aims to provide facilities to retain top UK/EU research brains domestically.
• Public Perception: Governments must demonstrate AI benefits (healthcare, drug research) to the public without relying solely on the "AI" label to drive adoption.

Energy Grid Challenges and Solutions

• UK Energy Surplus: The UK has up to 6.5 gigawatts of annual offshore renewable energy that the national grid cannot currently absorb, resulting in wasted potential.
• Coordination Strategy: A government-led energy council is required to coordinate renewable energy generation, data center construction, and national grid upgrades.
• Community Impact: Without grid coordination, data centers will build independent power sources, bypassing local communities; coordination ensures energy prices drop for local citizens, improving public acceptance.
• Switzerland Context: Switzerland faces unique challenges as a landlocked country unable to utilize offshore renewable energy sources.

Future Outlook

• Nscale envisions a collaborative fabric connecting AI growth zones to support startups and enterprises across the UK and Europe.
• The industry is moving away from the "Cloud First" model toward a "Sovereign First" model where control, compliance, and economic growth are paramount.
• No further Q&A was conducted following the single public question regarding energy resources.