Ivan Zhang, featured on this year’s A.I. Power Index along with co-founders Nick Frosst and Aidan Gomez, is building Toronto-based Cohere into one of the world’s most promising A.I. startups with a $6.8 billion valuation following a $500 million funding round in August. Zhang has championed an efficiency-first approach that contradicts the industry’s “bigger is better” mentality. Cohere has focused exclusively on enterprise A.I. rather than chasing consumer viral moments. This strategy has proven successful as the company more than doubled its annualized revenue from $35 million in March to over $100 million by May. Zhang argues that enterprises require security, customization, efficiency and reliability at levels consumer products don’t attempt, positioning Cohere as what he describes as “the only major player focused solely on enterprise A.I.” at a time when he sees A.I. transitioning from experiential tool to real infrastructure across major organizations.
What’s one assumption about A.I. that you think is dead wrong?
That massive, resource-hungry models are the only way to go. The industry got obsessed with throwing more money and chips, leading to better outcomes, but we’ve proven that wrong repeatedly. Our latest models are bringing customers incredible performance on 1-2 GPUs, because we’ve found that enterprises running models privately need to be resourceful about the hardware they have. If our models aren’t required to handle the long tail of consumer chat use cases, they don’t need the capacity to store edges of the internet. We can train models that only spend a tiny amount of compute for great agentic tool use.
If you had to pick one moment in the last year when you thought “Oh shit, this changes everything” about A.I., what was it?
Honestly, it wasn’t a single breakthrough moment—it was watching our customers actually deploy models and North at scale and seeing the adoption curve start to accelerate. We knew enterprise adoption would be slower than consumer adoption, but we’re getting to the point where people realize this isn’t just another productivity tool. It’s not experiential anymore, but it’s becoming real infrastructure. The “oh shit” was realizing the scale of what’s coming once this adoption pattern starts to ripple across every major organization.
What’s something about A.I. development that keeps you up at night that most people aren’t talking about?
The gap between the security posture that enterprise A.I. requires and how some players in the industry are operating. The industry doesn’t talk about it much because securing this infrastructure is harder than chasing the next benchmark. We’re the only major player focused solely on enterprise A.I., and we know that consumer chatbots weren’t designed for the high-stakes security enterprises require.
You’ve focused on enterprise A.I. rather than chasing consumer viral moments. What convinced you that it was the right bet?
Simple math. Enterprises will pay for A.I. that solves their actual business problems. You can’t just take a general consumer model and expect it to work in a regulated environment. Enterprises need security, customization, efficiency, and reliability at levels that consumer products don’t even attempt. So while other people were racing to build the flashiest demo, we built the infrastructure that actually works when you need to deploy at scale with real stakes.
How do you split the technical vision when you’re both deeply technical founders?
We complement each other pretty naturally. I (Ivan) work to translate our models into products that people can actually use at work, which helps drive the vision for North. Aidan focuses a lot on the direction we’re going with our models and products and the industries we’re best-suited to serve with our approach. It helps that we’re both pragmatists. We’re not interested in building A.I. for the sake of it. We want to solve real problems. When we disagree on technical direction, it usually comes down to what will actually work for customers, not what’s theoretically interesting.
