22 October 2026
Local AI for SMEs: from niche to global category
An Australian provider is packaging DGX Spark and on-premise AI for regulated SMEs, and the trade press talks compliance: the local AI model is going global.
If talking about “local AI for SMEs” sounded like a niche obsession a year ago, this week brought another signal that the model is becoming a global market category.
The signal: sovereign AI packaged as a service
In Australia, the provider IOTAI has published a dedicated service page: “Sovereign AI: On-Premise LLM Deployment on NVIDIA DGX Spark”, aimed at SMEs in regulated sectors (legal, healthcare, finance). The package: hardware supply, integration and “zero data egress” — exactly the same business model being built in Catalonia: a mini AI server in your offices, with your data never leaving your network.
It is not an isolated case. NVIDIA’s developer forums discuss the “sovereign AI stack” architecture on DGX Spark with open models, and Spanish trade press (VARINDIA) already describes the DGX Spark as “the missing hardware piece for on-premise AI compliance”.
The driver: compliance, not hype
What is driving this is not generative hype but regulation. The European AI Act, GDPR and sector rules like NIS2 or ENS push sensitive data to stay inside the organisation. Running an LLM on your own hardware — with open-weight models that keep getting more capable — turns a legal problem into a solved technical problem.
When the same argument shows up simultaneously in Spanish press, technical forums and providers on another continent, it has stopped being a bet and become an expectation.
What it means for a Catalan SME
Three practical readings:
- The model is validated. If you were waiting to see whether local AI was just for techies, you now have the answer: providers are packaging it as a standard service in several countries.
- Watch out for marketing. “On-premise” and “sovereign AI” are about to become buzzwords, and not everyone advertising them does the same. Concrete questions: where do your data actually run?, which models?, is there any call to external services?, what about updates?
- The window is still open. Being among the first in your sector with well-built local AI — an assistant with your own documentation, a private RAG, automation over your own data — is still a competitive advantage. Once it becomes an industry standard, it will just be a cost.
If you want to assess what local AI could do for your company, talk to us.
Sources: IOTAI — Sovereign AI on DGX Spark · VARINDIA — DGX Spark, the missing hardware piece for on-premise AI compliance · NVIDIA Developer Forums — The Sovereign AI Stack
Free assessment
Local, private AI for businesses, freelancers and organisations. Your data never leaves your office.