Open Models in U.S. Earnings Calls: Why Businesses Are Talking About Them

Open AI models are becoming a bigger part of enterprise conversations.

Recent reporting based on U.S. earnings calls found that mentions of open models increased sharply compared with the previous year. One analysis reported a sixfold increase in references, while companies such as Vercel and AT&T have disclosed significant use of open models in their AI workloads.

The trend reflects a broader change in enterprise AI.

Businesses are no longer evaluating models only by benchmark scores.

They are considering:

Cost

Data control

Deployment flexibility

Vendor dependence

Customization

Latency

Infrastructure requirements

Open models can be attractive because companies can deploy them in environments they control.

That can be particularly useful for regulated industries.

A bank, healthcare provider or government organization may not want every sensitive workload routed through an external API.

An open-weight model deployed inside controlled infrastructure can offer another option.

But open models aren't automatically cheaper or easier.

Organizations still need GPUs, inference infrastructure, model updates, security controls and engineering talent.

That means the real comparison isn't simply:

Closed model vs open model.

It is:

Managed API economics vs infrastructure ownership.

For some workloads, a hosted frontier model may remain the most practical choice.

For others, running an open model internally may make more sense.

The increase in open-model discussions on earnings calls suggests that enterprises are moving beyond experimentation.

They are beginning to treat model choice as an infrastructure and financial decision.

And that could make the next phase of AI much more competitive.