Ivan Zhang, who played a role in this year’s AI Power Index along with co-founders Nick Frosst and Aidan Gomez, is becoming one of the world’s most promising AI startups, valued at $6.8 billion after a $500 million financing in August. As the company’s chief technology officer, Zhang advocates an efficiency-first approach that contradicts the industry’s “bigger” mentality. Cohere specializes in corporate AI, not chasing consumers’ viral moments. The strategy has proven to be successful as the company’s annual revenue increased from $35 million in March to $100 million in May. Zhang believes that businesses need to be safe, custom, efficient and reliable at the level of consumer products, positioning CONERE as what he calls “a major player focused only on enterprise AI” when he saw AI transition from experience tools to real infrastructure throughout the main organization.
What do you think is an assumption of AI?
Large-scale, resource-hungry models are the only option. The industry is obsessed with Throw away more money and chips to achieve better results, but we have proven it Repeated errors. Our latest models make customers perform incredible on 1-2 GPU, because we found that the enterprise runs the model privately Resourceful in the hardware they have. If our model doesn’t take long Consumer chat use cases are tail, they don’t need the ability to store the internet edge. We can train models that only cost a small amount of calculations to perform excellent proxy tools.
If you have had to choose for a while in the past year, think about: Changed everything about artificial intelligence, what is this?
Honestly, this is not a breakthrough moment – it is watching our customers actually deploy models and deploy North on a massive scale and see the adoption curve start to accelerate. We know that enterprise adoption rates are slower than consumer adoptionbut we need to reach the key point People realize that this is more than just another productivity tool. It is no longer an experience, but It has become a real infrastructure. “Oh shit” is realizing the scale of what happened before This adoption pattern begins to ripple in every major organization.
Something developed by AI allows you to keep most people up at night Aren’t you talking about it?
The gap between safety postures required by Enterprise AI and how some participants The industry is operating. The industry hasn’t talked much about it because it’s ensured Infrastructure is harder than pursuing the next benchmark. We are the only major players Focusing on enterprise AI only, we know that consumer chatbots are not for High-risk security businesses need it.
You focus on enterprise AI, not on the virus moments chasing consumers. What Convince you that this is the right bet?
Simple math. Enterprises will pay for AI that solves their actual business problems. You can’t Just take the general consumer model and expect it to work in a regular environment. Businesses need security, customization, efficiency and reliability at the consumer level The product wasn’t even tried. So when others compete to build the most gorgeous presentation, We built infrastructures that actually work in situations where you need to deploy at scale bet.
How do you separate the tech vision when both of you dive into the tech founders?
We naturally complement each other. I (Ivan) worked hard to transform our model into Products that people can actually use at work help drive the North’s vision. Aidan Very concerned about our direction with models and products and We are best suited to the industry that serves our way. We are all pragmatists, which helps. We are not interested in building AI for this. us Want to solve the real problem. When we disagree with the direction of technology, it usually drops What is actually useful to the customer, not something interesting in theory.
