Sovereign AI and the Agentic Cloud with Peter Damm

In this episode of Nerding Out with Viktor, host Viktor Petersson sits down with Peter Damm, co-founder of Granite, a European company building infrastructure for sovereign AI, to unpack why the next phase of AI may be decided by infrastructure as much as by models. The UK pays offshore wind operators around three billion a year to switch their turbines off, while AI companies struggle to find power for the GPUs they have already bought.

Peter explains why agentic workloads need a different kind of cloud, one that balances GPUs, CPUs, memory and storage, and why the hardware ratio is flipping from eight GPUs per CPU toward eight CPUs per GPU. Granite builds modular, container-sized data centers that go where the stranded energy is, deploying in about four months rather than the five to seven years it takes to bring a traditional data center online.

The episode reframes sovereignty as a question of control rather than regulation. Closed models can be switched off or repriced overnight, while open models like Kimi now match frontier quality and can be audited and run on your own hardware. Viktor and Peter also cover unpredictable agent costs, the YOLO era of agent permissions, AI-driven security threats, and whether OpenAI and Anthropic will survive the revolution they helped start. For founders and engineering teams building on AI, it is a practical look at keeping models, data and costs under their own control.