Should companies host "owned" AI models on their own servers?

"Intelligence is not yet a commodity, but it will be for 99% of knowledge work."

Says Jeff Huber in his “12 factor companies” post. This is my Part 2 response to this post.

There is a lot of momentum right now around the idea of “own your intelligence.” We hear Alex Karp talk about it. We hear Satya Nadella talk about his reverse information paradox. Companies competing with Anthropic and OpenAI are aggressively pushing open source and open weight models.

But for the businesses I work with, owning intelligence, having open source models on your own servers, and building your own infrastructure just does not make any sense.

Maybe building infrastructure makes sense for a very small percentage of companies who are making billions of dollars.

But my customers are $100 million-dollar e-commerce brands. They cannot do this. It is just not on the table. Thinking you have to buy your own Mac minis or servers to host your AI models is a distraction.

The way to maintain optionality and pricing power is to make it easy to swap intelligence. Renting intelligence makes a lot of sense, at least for now.

By using something like OpenRouter to have access to different kinds of models, you ensure you are not locked in. You do not want vendor lock-in.

Depending on the tasks, you can seamlessly swap models. Some of them are cheaper and faster, some of them are more intelligent and more expensive. That is totally fine.

Let the multi-billion dollar companies build their own infrastructure. For small, medium, and large brands, rent your intelligence, avoid vendor lock-in, and route your tasks to the best model for the job.

Bora Celik
Founder, Gentic

--

Reply to this email or DM me on LinkedIn to discuss your brand's agentic transformation via gentic.co