- Anthropic CEO Dario Amodei said AI training could someday cost $100 billion.
- Amodei told CNBC he's not worried about the commoditization of large-scale language AI models.
- The number of companies that can afford to develop new models will remain small, he said.
The market for large-scale language models may be heating up, but the CEO of AI giant Anthropic isn't too worried about competition, saying the price tag for training such models could end up being as low as 100 billion. We predict that it could increase to US dollars.
Anthropic CEO and co-founder Dario Amodei talked about the future of Anthropic, its chatbot Claude, and the billion-dollar AI industry in a wide-ranging interview with CNBC last week.
Amodei said the “number of players” with the financial resources to train professional-grade AI models at the highest scale “will be relatively small to begin with.”
Amodei and his sister Daniela Amodei co-founded Anthropic in 2021 and quickly attracted major backers including Amazon. The e-commerce company pumped $1.25 billion into Anthropic last year and pledged another $2.75 billion in March, giving Amazon a minority ownership stake in Anthropic and giving Anthropic access to Amazon's cloud servers and chips. We have strengthened our strong partnerships to give you access to
The company's Claude is comparable to similar models such as OpenAI's ChatGPT and Google Gemini. Meanwhile, Amazon is working on its own AI, Olympus, and Elon Musk open sourced the Grok model last month.
But despite the boom in model development, Amodei dismissed concerns about rapid commoditization, pointing out that creating and training large language models is astronomically expensive. did. Amodei told CNBC that the current model has already cost $100 million to develop, and that price will only increase as the technology improves.
“I think next year the cost of training a model will be around $1 billion,” Amodei told the outlet. “And in 2025, 2026, it's going to reach $5 billion or $10 billion. And I think it could go beyond that to $100 billion.”
The number of companies that can afford to train models at that cost will remain small, he said.
Amodei told CNBC that diversity in development technology could also help prevent commoditization, likening the different models to differences between humans.
“We humans are all fundamentally designed the same way, but we are very different from each other, and I think the models will be the same,” he said.
Amodei said some AI models may specialize in subjects such as law or national security, while others may gain expertise in biochemistry.
“I think this momentum will lead to different model providers specializing in different things, even if the basic model they create is the same,” he added.
