CEO of Anthropic — the AI company Amazon is betting billions on — says it could cost $10 billion to train AI in 2 years

Amazon Web Services CEO Adam Selipsky speaks with Anthropic CEO Dario Amodei during a 2023 conference. Noah Berger/Getty

• Anthropic CEO Dario Amodei said training AI could one day cost $100 billion.
• Amodei told CNBC that he isn’t worried about the commoditization of large language AI models.
• The number of companies able to afford to develop new models will remain slim, he said.

The large language model market may be heating up, but the CEO of AI powerhouse Anthropic isn’t all that worried about competition, predicting the price tag of training such models could eventually balloon to $100 billion.

Anthropic CEO and cofounder Dario Amodei discussed the future of Anthropic, its chatbot “Claude,” and the billion-dollar AI industry in a wide-ranging interview with CNBC last week.

“The number of players” that have the financial capability to train professional-level AI models at top scale, Amodei said, “is going to be relatively small to start with.”

Amodei and his sister, Daniela Amodei, cofounded Anthropic in 2021, quickly drawing major backers, including Amazon. The e-commerce company poured $1.25 billion into Anthropic last year and pledged an additional $2.75 billion in March, cementing a powerful partnership that grants Amazon a minority ownership stake in Anthropic and allows Anthropic access to Amazon’s cloud servers and chips.

The company’s Claude rivals similar models like OpenAI’s ChatGPT and Google Gemini. Meanwhile, Amazon is working on its own AI “Olympus,” while Elon Musk open-sourced his “Grok” model last month.

But even as model development booms, Amodei brushed off concerns of rapid commoditization, pointing to the astronomical price of creating and training large language models. Amodei told CNBC that current models already cost a company $100 million to develop — and that price will only increase as the technology advances.

“I think we’re going to see models trained in the next year are going to be about $1 billion,” Amodei told the outlet. “And then 2025, 2026, we’re going to go to $5 billion or $10 billion. And I think there’s a chance it may go beyond that to $100 billion.”

The number of companies financially able to train models at that cost will remain slim, he said.

Diversity in development techniques may also help stave off commoditization, Amodei told CNBC, comparing the various different models to differences in human beings.

“We as humans, we all have — our brains are all basically designed the same, but we’re very different from one another, and I think models will be the same,” he said.

Amodei said some AI models may specialize in topics like law or national security, while others could gain expertise in biochemistry.

“I think that force is going to lead to different model providers specializing in different things, even as the base model they made is the same,” he added.

If you enjoyed this story, be sure to follow Business Insider on Microsoft Start.