Cohere raises $500M to beat back generative AI rivals

Startups

Cohere, a generative AI startup co-founded by ex-Google researchers, has raised $500 million in new cash from investors including Cisco, AMD and Fujitsu.

Bloomberg says that the round, which also had participation from Canadian pension investment manager PSP Investments and Canada’s export credit agency EDC, values Toronto-based Cohere at $5.5 billion. That’s more than double the startup’s valuation from last June, when it secured $270 million from Inovia Capital and others, and brings Cohere’s total raised to $970 million.

Josh Gartner, head of communications at Cohere, told TechCrunch that the financing sets Cohere up for “accelerated growth.”

“[W]e continue to significantly expand our technical teams to build the next generations of accurate, data privacy-focused enterprise AI,” Gartner said in a statement. “Cohere is laser-focused on leading the AI industry beyond esoteric benchmarks to deliver real-world benefits in the daily workflows of global businesses across regions and languages.”

Reuters reported in March that Cohere was seeking to nab between $500 million and $1 billion for its next round of fundraising, and that it was in talks with Nvidia and Salesforce Ventures to raise the money. Both Nvidia and Salesforce ended up contributing, Gartner confirmed in an email to TechCrunch.

Aiden Gomez launched Cohere in 2019 along with Nick Frosst and Ivan Zhang, with whom Gomez had done research at FOR.ai, a sort of progenitor to Cohere. Gomez is one of the co-authors of a 2017 technical paper, “Attention Is All You Need,” that laid the foundation for many of the most capable generative AI models today, including OpenAI’s GPT-4o and Stability AI’s Stable Diffusion.

Unlike OpenAI, Anthropic, Mistral and many of its generative AI startup rivals, Cohere doesn’t have a big consumer focus. Instead, the company customizes its AI models, which perform tasks such as summarizing documents, writing website copy, and powering chatbots, for companies like Oracle, LivePerson and Notion.

Cohere’s AI platform is cloud agnostic, able to be deployed inside public clouds (e.g., Google Cloud, Amazon Web Services), a customer’s existing cloud, virtual private clouds or on-site. The startup takes a hands-on approach, working with customers to create tailored models based on their proprietary data.

Cohere also runs a nonprofit research lab, Cohere for AI, and releases open models like multilingual models for understanding and analyzing text. Its latest flagship model, Command R+, is designed to deliver many of the capabilities of more expensive models (e.g. GPT-4o) while costing less.

Cohere’s has proven to be a winning strategy — even as both OpenAI and Anthropic ramp up their respective enterprise sales efforts. Bloomberg reports that, at the end of March, Cohere was generating $35 million in annualized revenue with a customer base of hundreds of companies, up from around $13 million at the end of 2023.

Generative AI at Cohere’s scale is a costly endeavor — particularly as the company looks to train more sophisticated systems. The new tranche will surely help, as will Cohere’s ongoing partnership with Google Cloud, wherein Google provides cloud infrastructure to train and run Cohere’s models. Cohere also has close ties to Oracle, which is an investor as well as a customer; the startup’s AI is built into many of Oracle’s software products, including Oracle NetSuite.

According to Bloomberg, Cohere is planning to double its 250-employee headcount this year.

Products You May Like

Articles You May Like

Nuro expands driverless autonomous vehicle testing in push to attract customers
Beyoncé to perform live on Netflix during NFL Christmas Gameday stream
Nearly half of AI data centers may not have enough power by 2027
After raising nearly half a billion dollars, ABL Space pivots from launch vehicles to missiles
Here’s the full list of 44 US AI startups that have raised $100M or more in 2024

Leave a Reply

Your email address will not be published. Required fields are marked *