AI tokens are on the minds of many chief financial officers. The chairman of OpenAI predicts that in a year, they won’t be thinking about them at all.
Read more Live Q&A: Why Kalshi is wooing Wall Street
Bret Taylor, who is the founder of the AI startup Sierra alongside leading OpenAI’s board, said that a lot of the “problems” related to tokenomics stem from the AI market not yet reaching maturity.
Speaking on CNBC on Monday, Taylor said that over time, he expects other companies to “bear the burden of managing tokens.”
Tokens are units of text that AI models process, and they’re used to calculate AI usage and cost. Earlier this year, spiraling token costs prompted companies to rein in their AI spending and assess whether they were getting a return on investment.
“I believe where the world is going is paying for outcomes,” Taylor said, echoing the view of other executives who are now prioritizing ROI.
Tokenmaxxing to token efficiency
Anthropic official says stopping AI usage is ‘the wrong’ response to AI cost concerns
Adam Mosseri says Instagram has ‘reined in’ AI costs after shutting down ‘the silly things’ that were burning tokens
My startup accidentally went over our AI budget from one Claude setting — it turned out to be worth it
Tokenmaxxing is so over. It’s all about modelmaxxing now.
Taylor compared today’s AI market to the early days of the internet, when it cost substantially more to build a website.
Read more Why ‘Big Short’ investor Michael Burry says Netflix makes milk — while Disney makes wine
“We’re just in the early stages of the technology, and I think it’s a call for entrepreneurs to develop solutions so businesses don’t need to deal with this stuff,” Taylor said.
He pointed to Ramp’s recently unveiled token spend tool, as well as sector-specific AI companies like the legal startup Harvey that manage tokens for their customers instead.
“I think if you fast-forward 12 months from now, IT departments will be really sophisticated about the industrial applications of AI,” Taylor said. “For your marketing department, you might have one thing. For your software engineers, another thing.”
“So you just don’t need to think about the word token at all,” he added.
Token costs are widely expected to come down over time as AI models become more efficient. Kimi K3, a model released last week by the Chinese startup Moonshot, grabbed Silicon Valley’s attention by rivaling the capabilities of powerful models from OpenAI and Anthropic at a lower cost.
Taylor said he’s not fully convinced yet.
“‘Is it cheaper to use’ is the most important part for anyone considering it,” Taylor said. “And I think the jury’s out on that.”
Read more AI is everyone’s job now
About the Author
