Summary
Cohere says it cleared $240 million in annual recurring revenue in 2025, a number that does more than flatter a pitch deck. It signals that enterprise AI has moved from experiment to line item, and that buyers are willing to commit real budgets even as the market is crowded by OpenAI and Anthropic.
The figure also drags Cohere toward a public market story, whether the company wants the spotlight now or not. At this scale, staying private starts to look less like patience and more like avoidance of the hard questions: margins, defensibility, and what happens when the hype cycle turns into procurement fatigue.
Revenue is not the same as power
$240 million in recurring revenue is a milestone, but it is not a moat. In enterprise AI, customers are not just buying model quality, they are buying risk management, contract terms, predictable uptime, and the freedom to avoid being trapped inside a single vendor’s ecosystem. Cohere’s positioning has long leaned into that sober enterprise sensibility, and the number suggests it is working. Still, the uncomfortable truth is that enterprise adoption often rewards the vendor that feels safest, not the one that is most innovative.
That dynamic matters because the competitive set is brutal in ways that normal software markets are not. The frontier model race is expensive, and the distribution advantage of a platform company can make superior technology feel irrelevant. If a CIO can get acceptable performance bundled with existing cloud commitments, “best model” becomes a philosophical argument, not a purchase order.
The IPO question is really a credibility question
An IPO would not simply be a liquidity event, it would be a test of whether Cohere can turn enterprise enthusiasm into durable economics. Public investors have less patience for romantic narratives about adoption curves. They want to know what it costs to serve a customer, what the renewal rates actually look like once pilots end, and whether inference margins improve or get competed away.
There is also a cultural tension. The AI industry loves to talk about building responsibly, yet the public markets demand velocity and predictable growth. Once quarterly expectations set in, “safety” can become a branding layer rather than an operating principle, unless leadership treats it as non negotiable even when it hurts the numbers.
Enterprise AI is growing up, and that is the risk
Cohere’s $240 million year reads like proof that enterprise AI has crossed a threshold. But maturity is when customers start negotiating harder, standardizing vendors, and asking why they need custom solutions at all. The next phase is less about dazzlement and more about consolidation.
If Cohere does move toward an IPO, the real story will be whether it can stay distinct when the market stops rewarding possibility and starts rewarding inevitability. In AI, inevitability has a habit of looking like whoever controls distribution. The question is whether Cohere can make trust and enterprise focus feel like distribution, not just a posture.
