Understanding Margins: Cohere Founder Discusses Rapid Changes in AI Business Models

OpenAI and Anthropic invest billions annually in training advanced AI models such as GPT-4 and Claude, yet aggressive price competition is destabilizing the market surrounding these technologies. Aidan Gomez, CEO of AI provider Cohere, highlighted this issue during a recent podcast, declaring that selling access to AI models is rapidly evolving into a "zero margin business." Currently, the costs associated with these AI models surpass their revenue.

“If you're focused solely on selling models, the near future is going to be quite challenging,” Gomez explained in an interview with 20VC’s Harry Stebbings. His reference to “selling models” pertains to providing API access to AI technologies; key players like OpenAI, Anthropic, Google, and Cohere face similar challenges in this arena.

“The landscape is shifting toward a zero margin business due to aggressive price competition. Many are offering models at no cost. While the overall market is still significant due to the rapid growth of technology demand, profit margins are currently extremely tight.”

AI companies at the forefront are fiercely competing to enhance their models. The primary strategy for improving performance involves investing heavily in computational resources, necessitating substantial expenditures on Nvidia hardware to make models more efficient. Simultaneously, a price war has erupted; OpenAI and Google have reduced costs for accessing their models to retain users, while Meta stands out by offering its open-source models for free.

“That's why there's a surge of enthusiasm at the application layer,” Gomez noted, highlighting offerings like OpenAI’s $20 monthly ChatGPT subscription. He believes Cohere’s AI models are positioned for long-term success, but generating revenue through product offerings will be essential in the interim.

In essence, today’s AI models are operating at a significant loss. While tech giants like Microsoft and Google can absorb these losses, startups often lack that luxury. Cohere is among the few remaining startups dedicated to developing cutting-edge AI models, competing alongside OpenAI, Anthropic, and Mistral. Many of their peers—such as Inflection, Adept, and Character.ai—have been acquired by larger cloud providers, leaving behind nonviable business models while retaining their advanced technologies.

However, Big Tech appears to be consuming emerging firms before they can mature into significant competitors.

“It’s precarious to become dependent on your cloud provider,” Gomez remarked, emphasizing that venture capitalists seek solid returns, in contrast to cloud companies which may have broader objectives. “This arrangement often proves detrimental for startups.”

As the race to create advanced AI models intensifies, these companies find themselves in a challenging landscape. There is hope that breakthroughs in model architecture, data efficiency, or computing power will yield substantial returns, but the timeline for such advancements remains uncertain. Unfortunately, it’s clear that not every AI startup will endure long enough to witness these potential rewards.