Major players in the AI sector like OpenAI and Anthropic are pouring billions into developing advanced models such as GPT-4 and Claude. However, the rampant undercutting of prices poses a significant threat to the profitability of these endeavours. Cohere’s CEO, Aidan Gomez, highlighted in a recent podcast how the sale of model access is on the verge of becoming a business with no profit margins. As things stand, the costs associated with these AI models exceed the revenue they generate.
“The challenge is particularly acute for those whose business is solely model sales in the immediate future,” Gomez remarked during a conversation with Harry Stebbings on 20VC. Here, “selling models” refers to the provision of API access to such AI technologies; a service provided by giants like OpenAI, Anthropic, Google, and Cohere, all of whom are grappling with this issue.
“This is evolving into a no-profit margin scenario due to the extensive price undercutting. Some are even offering their models at no cost. While the sector will continue to be substantial due to the increasing demand for this technology, the profits, at least for now, are expected to be minimal,” he explained.
The competition among firms developing state-of-the-art AI models is intense. Today, the most effective way to enhance AI capabilities is through greater computational power, which often involves significant investment in hardware from companies like Nvidia. Concurrently, there’s a push towards lowering prices. Both OpenAI and Google have reduced their model access fees to keep users, whereas Meta has made its models freely available.
“That’s why there’s so much interest in the application layer,” Gomez pointed out, nodding towards OpenAI’s $20 monthly subscription for ChatGPT. He believes that while Cohere’s AI models promise long-term business viability, short-term revenue might be driven through product offerings.
In simple terms, today’s AI models are operating at a loss, a financial hit that giants like Microsoft and Google might absorb, but startups often cannot. Cohere stands among the few startups still in the race to develop pioneering AI models, along with entities like OpenAI, Anthropic, and Mistral. Other startups, such as Inflection, Adept, Character.ai, have been acquired by larger cloud services, leaving behind a business model that struggles to stand on its own while their innovative technologies thrive under new ownership.
Yet, the relentless assimilation of these emerging firms by larger tech companies is troubling, as they’re barely given a chance to emerge as potent rivals.
“Falling into a dependency on your cloud service provider is a risky move,” Gomez cautioned, indicating that while venture capitalists are in pursuit of attractive returns, cloud services might have broader ambitions. “It’s a risky business strategy.”
For companies at the forefront of AI innovation, the road ahead is fraught with challenges. There’s hope that advancements in model structures, data efficiency, or computational capabilities might eventually reap substantial benefits. However, the timeline for such achievements is uncertain, and it’s clear not all AI ventures will survive to witness these potential breakthroughs.
Compiled by Techarena.au.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
