Cohere’s new CFO believes there’s a “fundamental difference” between the artificial intelligence (AI) startup and its large language model (LLM) competitors that goes beyond its focus on enterprises instead of consumers.
According to Francois Chadwick, Cohere spends money on computing power to train its LLMs but doesn’t carry its customers’ full compute cost. Customers don’t have to pay as much to integrate services, he claimed, allowing the Canadian firm to carve out an advantage in a crowded, well-capitalized, and foundational AI field.
Chadwick spoke with BetaKit at the AI-oriented ALL IN conference in Montréal, just after Cohere announced it had raised an additional $100 million USD from the Business Development Bank of Canada (BDC) and Nexxus Capital Management, bringing its valuation to $7 billion USD.
“They don’t have to throw out the baby with the bathwater…to prove to themselves that they’re going to get that ROI.”
Francois Chadwick,
Cohere
Chadwick, who previously served as Uber’s acting CFO, was hired earlier this month alongside chief AI officer Joelle Pineau as the company announced $500 million USD in fresh funding. A Cohere spokesperson said that the first close was oversubscribed, so the team decided to add on a smaller, second close “to give us additional capital and also to allow really good investors to join.”
Founded in 2019 by former Google researchers, Cohere builds LLMs that power chatbots and other AI applications for enterprises and governments.
It faces stiff competition from larger American LLM developers like OpenAI and Anthropic that have raised more money. Meanwhile, tech giants building their own AI models—Google, Meta, Amazon, and Microsoft—are expected to spend a combined $400 billion USD on capital investments next year, according to the Wall Street Journal. Chip giant Nvidia also announced it would invest up to $100 billion USD in OpenAI to fund more data centres for AI compute.
Cohere’s main product is a workspace platform called North. It allows users to create personalized AI agents to perform and automate tasks like document summaries and emails. It also offers security options, such as on-premise deployment (running software on a company’s computers rather than remote servers).
Chadwick noted that Cohere lets customers continue using their existing cloud services and AI models when deploying Cohere’s North platform.
“They don’t have to throw out the baby with the bathwater…to prove to themselves that they’re going to get that ROI,” Chadwick claimed.
Cohere is on track to bring in $200 million in revenue this year, according to a source familiar with the company’s operations. That means its valuation is set to be 35 times its revenue, which puts it lower than OpenAI, Perplexity, and Anthropic, according to some estimates.
On stage at ALL IN, Cohere CEO Aidan Gomez claimed that 99 percent of the company’s revenue came from outside of Canada two years ago. He now puts that number between 85 and 90 percent due to a greater appetite from Canadian buyers.
Throughout the conference, Cohere was featured on several panels and lauded by government officials like AI minister Evan Solomon and industry minister Mélanie Joly as a leading AI company.
“We will build Cohere, and we will make it a Canadian champion,” Joly said in a closing keynote speech.
Beyond providing Cohere with $240 million via its Sovereign AI Compute Strategy, the government has also signed a non-binding memorandum of understanding with the company to explore using its AI tools within the public service. Joelle Pineau, the company’s chief AI officer, was also tapped as part of the feds’ new AI strategy task force.
As the federal government works on new rules for data and AI, Chadwick believes Canada should strike a balance between the United States’ anti-regulation approach and the stricter laws of the European Union.
“What I think should happen is…you build rules and regulations that allow companies to build, allow capital to flow to those companies, and then you put sensible regulations on how they actually get used,” he said.
Feature image courtesy ALL IN