The first thing you notice about Cohere's product page is what it doesn't promise. There is no chatbot avatar, no demo of a model writing a sonnet, no leaderboard score. Instead, the headline word is private, and underneath it, a quiet line about deploying inside a customer's own virtual private cloud or on-premise environment [Cohere]. For a generative AI company in 2026, this is a kind of statement of values: the model is the easy part, the deployment is the product.
That positioning is the bet Aidan Gomez, Ivan Zhang, and Nick Frosst have been refining since they founded Cohere in Toronto in 2019 [Wikipedia]. While much of the AI industry has spent the last three years racing to win consumer mindshare, Cohere has aimed almost entirely at the enterprise buyer who cannot, for legal or competitive reasons, send sensitive data to a public API. The company sells generation, rerank, and embed models on a usage-based meter, with a separate dedicated tier billed per instance for customers who want guaranteed performance [Cohere]. The pitch to a bank or a manufacturer is straightforward: bring our models into your environment, point them at your documents, and pay for the compute you actually use.
The wedge is working well enough that investors keep writing larger checks. Cohere closed a reported $270 million Series C at a $2.2 billion valuation [Crunchbase], followed by a $500 million round in July 2024 at $5.5 billion [Crunchbase], and another $500 million in August 2025 that lifted the valuation to $6.8 billion [Reuters, Aug 2025]. PSP Investments, Cisco, and Fujitsu are on the cap table, an unusually strategic mix that hints at the kinds of customers Cohere is courting: pension-grade institutions, networking incumbents, and Japanese industrials.
Series C valuation | 2.2 | $B
Jul 2024 round valuation | 5.5 | $B
Aug 2025 round valuation | 6.8 | $B
The momentum on the revenue side appears to be catching up to the funding. According to an investor memo reviewed by CNBC in February 2026, Cohere has topped its revenue target as it builds toward an IPO [CNBC, Feb 2026]. The company has not disclosed specific figures publicly, but the framing of the memo, paired with the August fundraise and the Meta hire that accompanied it, suggests a business preparing for the scrutiny of public markets rather than one stretching for a bridge.
The shape of the opportunity
Because the largest AI buyers in the world (banks, insurers, defense primes, hospital systems, telcos) are structurally unable to use the default deployment model of OpenAI or Anthropic. Their compliance teams require data residency, audit trails, and the ability to run inference inside infrastructure they control. Cohere has built its product line around exactly that constraint, with vertical pages for financial services, technology, and manufacturing that all lead with the same private-deployment promise [Cohere]. The 2021 multi-year partnership with Google Cloud, which gave Cohere access to TPUs for training, established the technical credibility early [TechCrunch, Nov 2021], and the customer-facing posture has stayed consistent since.
The company has also been candid about the economics. In an August 2024 interview, Gomez told TechCrunch that AI's business model is changing fast and that margins on raw model inference are compressing [TechCrunch, Aug 2024]. Cohere's response has been to move up the stack: rerank and embed models that improve a customer's existing search and retrieval workflows, and a Model Vault inference platform for customers who want managed deployment without giving up isolation [Cohere]. It is a more boring product story than building artificial general intelligence, and probably a more durable one.
Team and traction
Gomez, who co-authored the original transformer paper at Google Brain, remains CEO and was elected to Rivian's board in April 2025, a signal that the EV maker views Cohere as a strategic AI partner [TechCrunch, Apr 2025]. Co-founder Ivan Zhang serves as CTO [sherlock-ai GitHub]. Martin Kon, the President and COO, runs the commercial side [The Information]. In August 2025, Cohere hired a former Meta executive as its AI Chief alongside the funding announcement [Reuters, Aug 2025], a move that gives the research org a senior leader with experience scaling models inside one of the few companies operating at frontier scale.
Headcount tells its own story. Cohere went from roughly 150 employees in late 2022 to 275 in late 2023, then 400 by December 2024 [The Information, Dec 2024], and reportedly exceeded 800 by early 2026 [IntuitionLabs]. A roughly 20-person layoff in July 2024, immediately following that summer's $500 million round, was a small recalibration against the broader hiring curve [CNBC, Jul 2024]. The current open roles, including a Machine Learning Engineer focused on tool use and a Senior Research Scientist at Cohere Labs [Lever.co; AshbyHQ], suggest continued investment in agentic capabilities and applied research rather than pure go-to-market expansion.
What the bears say
The honest counterfactual is competitive intensity. OpenAI, Anthropic, and Google are all chasing the same enterprise contracts, often with deeper pockets and stronger consumer brand pull-through, and the Reuters reporting on the August round noted that Cohere remains smaller than its US-based rivals by revenue [Reuters, Aug 2025]. The bull answer is that those rivals are structurally conflicted: their core distribution runs through public APIs and consumer products, and their largest enterprise customers increasingly want a vendor whose business model is private deployment first. Cohere's strategic backers (Cisco, Fujitsu, PSP) are precisely the kind of channel partners that can put the company in front of regulated buyers that competitors reach more slowly.
What to watch
The next twelve months will turn on three things: whether the IPO process implied by the February 2026 investor memo [CNBC, Feb 2026] actually materializes, whether the new AI Chief from Meta produces a model release that closes the perceived capability gap with frontier labs, and whether Cohere can convert its strategic-investor relationships into named, large-dollar deployments in financial services and manufacturing. An S-1 would be the most consequential document the company has ever produced, because it would force the first public accounting of a thesis Cohere has been building quietly for six years.
The cultural question Cohere is implicitly answering: in an AI market dominated by products that ask you to send your data somewhere else, what is the enterprise willing to pay to keep its data home?