Toronto-based AI startup Cohere has reached a significant milestone, doubling its annualized revenue to $100 million as of May 2025, according to sources familiar with the matter. This growth follows a strategic shift initiated in the third quarter of 2024, when the company pivoted toward private deployments tailored for enterprise customers in highly regulated industries such as finance, healthcare, and government.
Cohere CEO Aidan Gomez outlined this new direction in a year-end memo, emphasizing the company's focus on building customized AI models for enterprise users rather than competing in the race to develop ever-larger foundation models. The strategy has paid off handsomely, with approximately 85% of Cohere's business now coming from private deployments, delivering impressive profit margins of around 80%.
Founded in 2019 by former Google AI researchers, Cohere has raised more than $900 million from investors including Nvidia, Cisco, and Inovia Capital. The company was last valued at $5.5 billion following a $500 million funding round in July 2024. Unlike consumer-focused AI companies like OpenAI and Anthropic, Cohere has maintained its enterprise-only approach, serving clients such as Oracle, Fujitsu, and Notion.
In January 2025, Cohere launched North, a ChatGPT-style application designed to help knowledge workers with tasks like document summarization and data analysis. This secure AI workspace platform allows users to create and deploy AI agents without coding expertise while maintaining strict data privacy controls. The platform can be deployed in the cloud or on-premises, including in air-gapped environments for maximum security.
Cohere's shift toward smaller, specialized models reflects a broader trend in the AI sector. As companies report diminishing returns from scaling up model size, the industry is increasingly prioritizing domain-specific tools over large, generalized systems. This approach has resonated with enterprise clients seeking secure, customized AI solutions that address specific business challenges while maintaining control over sensitive data.