Valarian, a London-based startup founded by a former Palantir executive, has just secured $50 million in a Series A round led by US venture fund NEA. The stated goal sounds simple: enable governments and companies to leverage US-made cloud and artificial intelligence without letting Washington reach into their data.

The timing is no coincidence. The fear that sparked Valarian's creation – losing real control over information hosted on someone else's infrastructure – stopped being hypothetical this very year. Between renewed geopolitical tensions, European data protection rulings, and the extraterritorial reach of the US Cloud Act, the digital sovereignty market suddenly became tangible. No longer a debate for insiders, but an operational constraint.

Valarian's solution fits this gap: a software layer – technical details aren't public, but typically such systems rely on confidential enclaves, always-on encryption, and local key management – that sits between the user and the US cloud provider's infrastructure. In practice, a public administration or a European company can use AWS, Azure, or Google Cloud services while retaining full sovereignty over its data, because the provider never sees the information in clear and the keys stay under local control.

What it means for those evaluating on-premise infrastructure

The influx of significant capital on this front raises a structural question. If sovereignty layers like Valarian's become reliable and certified, they could slow the migration to pure self-hosted setups for sensitive but non-classified workloads. Why invest in an on-premise data center with dedicated GPUs for LLM inference if you can use the same cloud VMs while keeping control?

The answer isn't straightforward. For truly critical data – defense, intelligence, healthcare with strict requirements – the physical isolation offered by on-premise deployment remains the gold standard, and no software enclave can fully replace it. Moreover, adopting a cloud sovereignty solution introduces dependency on a third party (Valarian itself), merely shifting the trust problem.

The most telling signal, however, lies elsewhere: the rush of governments to find workarounds for using American AI without ceding sovereignty shows that the demand for models and compute power is so irresistible it bends regulatory rigidities. This will also accelerate on-premise hardware evolution, because savvy organizations will use the transition period (cloud with a sovereignty layer) to design scalable local infrastructure, anticipating when latency constraints or the recurring costs of sovereign cloud prove unsustainable.

Ultimately, Valarian is not just a startup raising money. It is an indicator that the sovereignty market is moving from theory to contracts. And for those focused on on-premise deployment of AI workloads, it represents both a credible short-term competitor and an indirect accelerator toward genuinely independent solutions in the medium term.