Stay informed with free updates
Simply sign up to the Artificial intelligence myFT Digest — delivered directly to your inbox.
Cohere, the artificial intelligence start-up founded by ex-Google scientists, is in talks to raise as much as $1bn in financing as investors race to back large language models that can rival systems under development by Microsoft-backed OpenAI.
Three people close to the talks said the Toronto-based start-up had not yet fixed a valuation for its new round or an exact fundraising target, but two of the people said a range of $500mn to $1bn in fresh capital had been discussed.
That is more than the total Cohere has lured from investors across four fundraising series to date, according to Crunchbase, and is expected to give the company a far higher valuation than the $2.2bn it achieved when it raised $270mn — which came from investors including Nvidia and Oracle and venture firms including Index Ventures and Inovia Capital — in June 2023.
Cohere declined to comment.
The five-year-old company was co-founded by Aidan Gomez, Nick Frosst, and Ivan Zhang. Gomez, Cohere’s chief executive, co-authored a landmark research paper while he was an intern at Google called “Attention Is All You Need” that led to significant advances in how computers analyse text.
Cohere has become one of the three most high-profile artificial intelligence start-ups in North America alongside OpenAI and Anthropic.
AI start-ups require a steady stream of new capital to develop models built on the back of huge computing power and vast amounts of data. “It’s obviously a capital intensive industry,” said one person involved in the fundraising.
But more than a year into an AI boom started by OpenAI’s launch of ChatGPT in November 2022, investors are anxious for signs that hot start-ups can commercialise their technology and provide a return on investment.
Similar to its competitors, Cohere is building a large language model capable of conversing with users. But it is focused on selling to enterprise customers rather than dominating the consumer chatbot market.
As a result of that narrower focus, Cohere’s costs to develop its models are lower than those of OpenAI, whose general-purpose tools are more expensive to build, train and run.
“An enterprise model has a very different scale — there aren’t millions of people using it for free. That needs a different amount of capital,” said the person involved in Cohere’s fundraising efforts.
Cohere’s return to investors, half a year after it last raised hundreds of millions of dollars, will test the appetite and ability of venture capitalists to continue putting money into AI at higher valuations. OpenAI and Anthropic have increasingly relied on partnerships with Big Tech companies as their capital requirements have grown.
Cohere’s equity value has been far lower than that of OpenAI, which is running a secondary sale of employee shares at a valuation of $86bn, and Anthropic, which is in talks to raise new funds at a valuation of more than $18bn, said people with direct knowledge of the discussions.
Cohere has positioned itself as a neutral provider for enterprise customers to use models that are not linked to cloud providers such as Microsoft — the biggest backer of OpenAI and an investor in other fast-growing AI start-ups such as Inflection — or Amazon and Google, both of which have committed billions of dollars to Anthropic.
