A growing number of companies, especially in regulated industries like finance and healthcare, are opting for open-source AI models they can run on-premise. This trend is driven by concerns over data leakage, IP security, and national data sovereignty, creating a distinct market need for more domestic, controllable AI solutions separate from frontier models.

Relying on third-party LLMs is a temporary phase. The ultimate advantage will come from companies training and owning their own models, potentially on physical hardware in their office. This transforms AI from a rented tool into a core, defensible intellectual property.

Glean's co-founder argues that most enterprise tasks don't require expensive frontier models. Open-source alternatives are now capable enough for the vast majority of use cases. The primary adoption driver has shifted from data privacy to pure cost savings, as enterprises seek to control skyrocketing AI bills.

Rather than relying on APIs from major labs like OpenAI and Anthropic, large enterprises like law firm Latham & Watkins are now buying NVIDIA servers to build their own systems. This move gives them control over proprietary data, enhances security, and insulates them from the regulatory and competitive volatility surrounding the frontier model providers.

Echoing crypto's "not your keys, not your crypto," a new ethos is emerging in AI: if a company's core product relies solely on another's model via an API, it has no real ownership. Startups are realizing they must control their own model weights to ensure steerability, capture their data flywheel, and build a defensible business.

Contrary to past momentum, the most advanced AI startups are increasingly adopting and fine-tuning open-source models. This shift is driven by the need for cost-effective speed and deep customization as their workloads mature and scale.

Open and closed source AI models will coexist by serving different parts of the market. Companies with core AI needs and large budgets will "build" on open source for control and customization. Most others will "buy" closed-source APIs for convenience, mirroring the established build-vs-buy dynamic for other technologies.

Renting an API token conceals a complex stack comprising GPU hardware, cluster management, inference engines, and execution sandboxes. Companies attempting to host models on their own infrastructure must operationalize this entire supporting software architecture themselves.

As annual API bills exceed nine figures, enterprises shift from renting tokens from closed providers to owning model weights on private infrastructure. Open-model businesses commercialize by monetizing the necessary orchestration software, bare-metal deployment, and integration services required for complete ownership.