Cohere, the Toronto-based enterprise AI startup founded by former Google Brain researchers, closed 2025 with $240 million in annual recurring revenue - blowing past its $200 million target by 20%. The company is now openly positioning for an IPO that could come as early as this year.
The numbers, revealed through an investor memo reported by CNBC on February 13, show quarter-over-quarter growth exceeding 50% throughout 2025. Gross margins averaged around 70%, expanding by 25 basis points year over year.
The Quiet Operator
In an industry where OpenAI and Anthropic dominate the news cycle with Super Bowl ads and $30 billion funding rounds, Cohere has been building something less flashy but potentially more durable: an enterprise AI business that actually makes money efficiently.
The company’s pitch to enterprises comes down to three things. First, its Command family of models requires fewer GPU resources than competitors, with enterprises reporting 30-40% lower inference costs. Second, Cohere models can run on-premises, in hybrid environments, or through cloud providers - a flexibility that matters to regulated industries handling sensitive data. Third, its investor roster includes Nvidia, AMD, and Salesforce Ventures, giving it hardware optimization advantages and enterprise distribution channels that pure-play AI labs lack.
The result is a company that’s carved out a real foothold in financial services, healthcare, and manufacturing - sectors where data privacy isn’t a nice-to-have but a regulatory requirement.
North Star
Last summer, Cohere launched North, an enterprise AI workspace that lets businesses build secure, custom AI agents and workflows using Cohere’s models. The platform emphasizes security and compliance features that enterprise buyers actually care about, rather than the consumer-facing chatbot features that drive headlines.
In 2026, Cohere plans to continue expanding in Europe and building out North’s capabilities. CEO Aidan Gomez has indicated an IPO might come “soon”, though the company hasn’t committed to a specific timeline.
The IPO Question
The timing is interesting. OpenAI and Anthropic are both weighing public offerings, and the enterprise AI market is hot enough that investors are willing to pay premium multiples. At typical enterprise AI valuations of 15-25x ARR, Cohere could be looking at a public market debut in the $3.6 billion to $6 billion range - modest compared to the eye-watering private valuations of its larger competitors, but grounded in actual recurring revenue from paying enterprise customers.
That positioning might be exactly what public market investors want to see. After Anthropic’s tools sent software stocks into a $2 trillion selloff and the broader market grapples with what AI agents mean for existing software businesses, a company that can show steady enterprise revenue growth, healthy margins, and capital efficiency stands out.
What to Watch
The enterprise AI race is getting crowded fast. Anthropic recently added Allianz to its enterprise client roster. OpenAI launched Frontier for enterprise agent management. Google keeps expanding Gemini’s enterprise footprint. And Databricks, now at $5.4 billion in annual revenue and a $134 billion valuation after raising another $5 billion, is building the data infrastructure layer that all of these companies depend on.
Cohere’s bet is that enterprises want choice - and that a company focused exclusively on their needs, without the distractions of consumer products or existential AI safety debates, can win a meaningful slice of the market. The $240 million revenue figure suggests that bet is paying off. Whether it’s enough to sustain an IPO in a market that’s still figuring out how to value AI companies is the next test.