How Does Cohere Make Money?
Cohere is a Canadian enterprise-focused AI company that develops large language models specifically optimized for business use cases. Founded in 2019 by Aidan Gomez, Ivan Zhang, and Nick Chicken, Cohere has a unique claim to fame: co-founder Aidan Gomez is one of the co-authors of the original 'Attention Is All You Need' paper that introduced the Transformer architecture — the foundational technology behind virtually every modern large language model including GPT-4, Claude, and Gemini. Unlike OpenAI and Anthropic, which pursue both consumer and enterprise markets, Cohere has deliberately focused almost exclusively on enterprise customers. The company's product lineup includes Command (a model optimized for following business instructions), Embed (for converting text into numerical representations used in search and retrieval), and Rerank (for improving search result relevance). This enterprise-first strategy means Cohere's models are designed for deployment in business workflows rather than consumer chatbots, with emphasis on data privacy, customization, and integration with existing enterprise systems. With a valuation of approximately $5.5 billion and an estimated $35 million in ARR, Cohere represents the enterprise-focused approach to AI monetization. The company has raised over $970 million in funding from investors including NVIDIA, Oracle, Salesforce Ventures, and the Canadian pension fund. Cohere's positioning as a specialized enterprise AI provider, rather than a general-purpose frontier lab, reflects a bet that many businesses will prefer purpose-built AI tools over the general-purpose models offered by larger competitors.
Revenue Breakdown
How Cohere makes money, broken down by revenue stream.
Revenue from enterprise customers accessing Cohere's models through its API, including Command (instruction-following), Generate (text generation), and Summarize (document summarization) endpoints. Enterprise API contracts typically involve committed usage tiers, custom fine-tuning, and dedicated support.
Revenue from licensing Cohere's Command model family for on-premise and private cloud deployment. Many enterprise customers — particularly in regulated industries like finance, healthcare, and government — require models to run within their own infrastructure, making licensing a significant revenue stream.
Revenue from Cohere's specialized Embed (text embeddings for semantic search and retrieval-augmented generation) and Rerank (search result reranking) products. These are essential building blocks for enterprise AI applications like knowledge bases, document search, and customer support systems.
Business Model
Cohere sells enterprise-optimized AI models through API access, on-premise licensing, and specialized products (Embed and Rerank), targeting businesses that need customizable, privacy-conscious AI solutions.
How Cohere Actually Makes Money
Cohere generates revenue primarily through its enterprise API, which accounts for approximately 55% of its estimated $35 million ARR. Enterprise customers pay for access to Cohere's suite of models through tiered API plans that include committed usage volumes, custom fine-tuning capabilities, and dedicated support. Unlike OpenAI's API, which serves everyone from individual developers to Fortune 500 companies, Cohere's API business is heavily skewed toward larger enterprise deployments with higher average contract values. The company's models are optimized for specific business tasks like document processing, customer support automation, and internal knowledge management.
Command model licensing represents the second-largest revenue stream at approximately 25%. Cohere allows enterprises to license and deploy its Command model family within their own infrastructure — whether on-premise data centers or private cloud environments. This is a critical differentiator for customers in regulated industries (banking, healthcare, government, legal) where data cannot leave the organization's control boundary. By offering flexible deployment options that competitors like OpenAI initially resisted, Cohere has carved out a niche among security-conscious enterprises willing to pay premium prices for data sovereignty.
Cohere's Embed and Rerank products generate roughly 20% of revenue and serve as essential infrastructure for enterprise AI applications. Embed converts text into dense vector representations that enable semantic search — finding documents based on meaning rather than keywords. Rerank takes search results from any source and reorders them by relevance using AI. These products are particularly valuable for retrieval-augmented generation (RAG) systems, where AI models need to access and reason over large document collections. Many enterprises adopt Embed and Rerank as their entry point to Cohere before expanding to Command for more complex AI tasks.
Cohere's financial position reflects the challenges and opportunities of an enterprise-focused AI strategy. The company's $35 million ARR is modest compared to OpenAI and Anthropic, but enterprise AI sales have longer sales cycles and typically generate stickier, higher-margin revenue once established. Cohere's $5.5 billion valuation implies that investors believe the enterprise AI market is large enough to support multiple major players and that Cohere's specialized focus gives it advantages in specific verticals. The co-founder's role in inventing the Transformer architecture provides unique brand credibility, but the company must continue demonstrating that its models can compete technically with better-funded rivals.
Key Takeaways
- •Co-founder Aidan Gomez co-authored the foundational Transformer paper ('Attention Is All You Need'), giving Cohere unique credibility in the AI space.
- •Cohere's enterprise-only focus differentiates it from consumer-oriented competitors like OpenAI and Anthropic, targeting businesses that need customizable, deployable AI.
- •The company's Embed and Rerank products serve as essential infrastructure for retrieval-augmented generation (RAG), one of the most popular enterprise AI patterns.
- •Flexible deployment options including on-premise licensing give Cohere an advantage with regulated industries where data sovereignty is non-negotiable.
- •At $35M ARR against a $5.5B valuation, Cohere's commercial traction lags behind its valuation, but enterprise sales cycles are longer and tend to produce stickier revenue.
Related Companies
Anthropic
Artificial IntelligenceAnthropic is the AI safety company behind Claude, generating revenue through API access and Claude Pro/Team subscriptions for its advanced AI assistant.
OpenAI
Artificial IntelligenceOpenAI is the creator of ChatGPT and the GPT series of large language models, generating revenue through API access and subscription products.