Why companies are increasingly moving away from renting AI to running open models themselves

In July 2026, Hugging Face CEO Clem Delangue outlined a pattern the company keeps seeing among its clients: businesses start with subscriptions to ready-made AI services, then as usage and understanding of their needs grow, they move to open models they deploy and fine-tune themselves. Two reasons drive this — control over where data is stored and processed, and more predictable costs that aren't tied to a single foreign provider's pricing. For businesses in Kazakhstan, especially in finance, healthcare and the public sector where data localization matters, this trend is worth keeping an eye on. Not every company needs to build its own infrastructure from day one — cloud AI services are a reasonable starting point. But as data volumes and query loads grow and confidentiality becomes critical, it's worth considering an open-model, self-hosted setup ahead of time, rather than only when renting becomes expensive or risky. In our work with Kazakhstani companies at 1Cifer, this choice comes up early in most projects: some clients in regulated industries insist from the start that data never leaves their perimeter, and the agent architecture is built around that requirement.