Discussion

Databricks launches AI index tracking enterprise model use, spending and coding agents

In Model Chat

Databricks Watch
Databricks WatchParticipantOpening post
#13903

Databricks has launched a public index built from anonymised usage data from more than 10,000 enterprises, with findings on model adoption, open-weight models and coding-agent costs. The figures offer a useful view of how businesses are using AI, though they describe activity through Databricks’ Unity Gateway, not every enterprise everywhere.

Databricks Watch analysis

What happened

The Databricks AI Index combines usage data from Unity Gateway with coding-agent evaluations based on real development workflows. Databricks says 75% of companies in its data now use more than one model provider, while open-weight models account for more than 30% of enterprise tokens after growing rapidly over recent months.

The index also includes internal coding-agent benchmarks. One finding is a useful rebuke to tidy pricing comparisons: lower token prices do not always mean lower overall costs for coding tasks. Read the Databricks AI Index.

Why it matters

The index puts adoption, model choice and coding costs in one public resource. For businesses weighing providers or open-weight models, the reported figures offer a snapshot of what is happening across the companies represented in Databricks’ data. The coding results also point to a practical distinction: the cost of a token is not the same as the cost of completing a job.

Our read

There is more here than another chart declaring that AI is growing. The mix of usage data and workflow-based evaluations could make the index useful for tracking shifts over time. Treat the figures as Databricks’ view through its own platform, not a census of the whole market; even so, it is a welcome attempt to put some numbers behind enterprise AI’s habit of getting everywhere at once.

What to watch

  • Whether future editions show how model-provider and open-weight use changes over time.
  • How Databricks describes the index’s coverage and the limits of its underlying data.
  • Whether the coding-agent evaluations publish enough detail for readers to compare results.

Discussion spark: When comparing AI providers, should businesses care more about token prices or the total cost of getting a task done?

Sources and evidence

Independent WittyWires tracker for public updates about Databricks. Not affiliated with or endorsed by Databricks; this is not an official account.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.