Bloomberg confirmed on June 18, 2026 that Lugano-based Prem AI is raising $100 million in Series A funding at a valuation of at least $500 million, with the round expected to close in Q3 2026. CEO and founder Simone Giacomelli leads the company, which previously raised a $14 million seed in 2024 and a $6.1 million bridge at a $200 million valuation. Reported backers include Jim Breyer of Breyer Capital, Index Ventures, Sequoia Capital China co-founder Fan Zhang, and Marvel Studios founder David Maisel. Prem coupled the funding news with the launch of Fluso, an AI workspace built for organizations that need to operate on sensitive data without routing it through OpenAI or Anthropic.
The product thesis is a direct response to a regulatory and geopolitical reality. Hedge funds, law firms, and clinical-trial operators cannot ship privileged or material non-public information to a third-party API hosted abroad, and the June 12 US directive freezing foreign-national access to Fable 5 and Mythos 5 just made the legal team's job harder. Prem's pitch — own the model, own the inference, own the audit trail — is the inverse of the frontier-lab business model, and the round is priced as if that thesis now has buyers.
Switzerland keeps showing up in this conversation for a reason. The country's data-protection regime, neutrality, and energy mix are good substrates for an on-premise AI vendor courting EU and Middle Eastern enterprise. Prem is the latest of several European AI infrastructure plays — Mistral, H, Aleph Alpha, Pleias — defining themselves against US hyperscaler dependency. The Dream round announced the same day at $3 billion makes a related but distinct bet on governments; Prem is making it on regulated commercial verticals.
Takeaway for learners: "AI is being commoditized" and "private AI is a real category" are both true at once. The frontier model layer is consolidating into a few labs, and the deployment layer is fragmenting into vendors that wrap those models in compliance, residency, and control. If you are picking a problem to work on, the deployment layer has lower model risk and higher procurement-cycle risk — different skill set, often a better career bet at the seed-to-Series-B stage than another model wrapper.