The CEO of Anthropic, a company valued at $18 billion, acknowledged that most startups can't afford to join the AI race.
In a recent “In Good Company” podcast interview with Norges Bank CEO Nikolai Tangen, Anthropik CEO Dario Amodei discussed the nuts and bolts of AI development, including how much it costs to launch an AI model.
Amodi said it costs about $100 million to train an AI model, but “there are models currently being trained that cost over $1 billion.”
Related: OpenAI rival Anthropic's models show 'metacognition': report
Amodei predicts that today's prices will soon be considered on the low side compared with what it will cost to develop competitive AI in the coming years.
“Once we get to $10 billion, $100 billion, which I think will be 2025, 2026, maybe 2027, I think we'll likely have models by then that are better than most humans at most things,” Amodei said.
Antropic CEO Dario Amodei. Photo by Chesnot/Getty Images
The prohibitive cost of AI development creates a high hurdle for startups that want to create their own models to compete with OpenAI and Anthropic.
On average, U.S. startups raised about $59 million in fourth-stage Series C funding in the first quarter of 2023, a far cry from the $450 million Anthropic raised in its Series C in May 2023.
Anthropic has raised more than $8 billion to date.
Related: Hugging Face CEO says acquisition requests for AI startups are on the rise
Amodei acknowledged that many companies won't be able to join the AI race.
“I think there's definitely going to be a vibrant downstream ecosystem and an ecosystem for smaller models,” he said.
Anthropik's March release of Clode, the latest version of its flagship AI chatbot, sparked discussion about “metacognition,” or the ability of an AI to monitor its own activity and self-correct. Anthropik engineers said at the time that Ops seemed aware that engineers were running tests on it.
Amodei also said in the interview that Anthropik is working with Google and Amazon, which are investing up to $2 billion and $4 billion, respectively, in Anthropik to develop AI chips to rival Nvidia.
The costs of AI could also mean negative environmental impacts: Google reported last week that its emissions have increased by nearly 50% in four years because of AI.
Related: AI has caused Google's greenhouse gas emissions to increase 48% since 2019
