AI companies set low prices after ChatGPT’s launch to attract customers, with investors subsidizing the initial costs, described as "subsidised intelligence" by Kevin Simback of Delphi Labs [1, 2]. However, the cost of AI usage has risen sharply due to expensive AI agents performing multiple tasks and consuming large numbers of tokens [1, 2]. Jack Gold of J.Gold Associates said some companies now pay more in token costs than employee salaries within a month or two because of "tokenmaxxing," or excessive AI usage [1].
Hardware shortages, such as limited computing chips and constrained data center capacity, add further pressure to AI costs [1]. Early in 2026, Meta encouraged employees to maximize AI token use as a productivity metric, but Andrew Bosworth, Meta’s CTO, later reversed this stance, declaring, "Nobody should be using AI tools just for the sake of using them" [1]. Uber’s chief operating officer also reported that rising AI spending has not delivered noticeable productivity improvements [1].
In response to rising expenses, some companies are switching to free, open-source AI models to cut costs [1]. Meanwhile, OpenAI and Anthropic announced plans to go public later in 2026 to attract broader investment from main street investors [1].
Mark Barton from Omniux summed up the financial challenge, stating, "All the costs are really starting to skyrocket" [1]. As companies weigh AI’s benefits against these growing costs, the move toward open-source alternatives and public offerings by leading AI firms marks a shift in the AI business landscape.
OpenAI and Anthropic’s upcoming public listings later this year will provide new funding avenues aimed at sustaining AI development amid rising operational costs [1].
