Evan Greenberg, chairman and chief executive officer (CEO) of global insurance company Chubb, said the company’s artificial intelligence tokens cost only a fraction of the efficiencies and improvements it will gain.
During Chubb Limited’s second-quarter 2026 earnings call, Greenberg was asked about how the company balances rising technology deployment costs with its ability to save costs and improve profits.
The issue comes as companies grapple with rising artificial intelligence (AI) costs, with a recent JPMorgan analysis noting that spending is being driven by the price of AI “tokens,” prompting companies to reassess whether productivity gains outweigh spending.
However, Greenberg said Chubb’s experience was different, with the insurer finding that the costs associated with the use of AI tokens were still small compared to the efficiencies, insights and improvements generated by the technology.
“This usage of tokens is really about the usage of a lot of tokens in tech companies. Artificial intelligence and tech companies are using a lot of tokens in model development. That comment doesn’t really apply to businesses in general,” Greenberg said.
He explained that Chubb tracks the cost of its tokens as part of its broader fee management framework.
“We know the cost of our tokens. Frankly, it’s our economic model and how we measure fees. The cost of our tokens and their usage is a fraction of the efficiencies, insights and improvements we get, and we measure it in hard currency.”
Later in the call, Greenberg also discussed the use of technology, saying “I think it’s a competitive advantage through technology, data, scale, scale and the breadth of capabilities and insights it brings to you,” adding that he believed it represented “a structural, long-term advantage.”
As the cost of AI tokens rises and companies seek to gain efficiencies from their AI deployments, there has also been a significant shift away from large base models towards cheaper open weight and open source models whose capabilities are quickly catching up.
