Enterprise Contracts Drive Cohere ARR

Diving deeper into

Cohere

Company Report
creating more predictable ARR compared to pure API-based competitors
Analyzed 5 sources

This revenue mix shows Cohere is building an enterprise software business, not just a model utility. Pure API revenue rises and falls with token volume, but private deployments and model licenses are sold on multi year contracts, often with dedicated instances and implementation work. That makes revenue easier to forecast because the customer is buying committed capacity, security, and deployment support, not just unpredictable bursts of model usage.

  • About 85% of Cohere's revenue comes from private deployments. These deals include customers like Oracle, Fujitsu, RBC, LG, and Notion, and they usually involve secure deployment inside the customer's own cloud or servers, which naturally lends itself to annual or multi year subscription commitments.
  • Cohere's enterprise plan is sold differently from the self serve API. Large customers pay custom contract pricing for dedicated model instances, support, and deployment options. That is closer to classic infrastructure software than to a pure consumption API, where spend can swing sharply with application traffic or experimentation cycles.
  • This also explains why Cohere looks different from OpenAI and Anthropic. Those companies have huge fast growing API businesses, but Cohere has leaned harder into private cloud, on prem, and workflow products like North. Mistral is moving in a similar direction, which suggests enterprise AI labs are discovering that contract revenue is more durable than raw token sales alone.

The next step is more bundling of models, retrieval, connectors, and workflow software into a single enterprise contract. As AI buyers move from experiments to company wide rollouts, the winners are likely to be the vendors that can lock in multi year deployment deals and become part of a customer's internal system of record.