This article reviews the latest developments, with examples from Japan and overseas, concerning rising token costs and governance barriers as companies integrate AI into their operations.
At a media briefing held by Akamai Technologies, the challenges of token costs and governance associated with the rapid growth in AI use were discussed. For global companies, the choice was described as either imposing restrictions or resolving the issue through systems and processes.
Examples from global companies showed that Uber Technologies consumed its annual AI budget in a short period and set a monthly per-user cap of $1,500, while Microsoft had an example of one user consuming $28,000 in a month and switched to a cheaper default model.
Japanese companies, meanwhile, face the characteristic of the Japanese language that makes token counts more likely to increase, and it was pointed out that AI costs are shifting from flat-rate pricing to usage-based billing. Costs are expected to continue rising.
Mori Yuko (森祐孝) cited research showing that per-model pricing falls to roughly one-thousandth over three years, while usage fees increase tenfold in 12 months, and said total costs are expected to continue expanding. It was also pointed out that the use of AI agents makes it more difficult to maintain visibility into the situation.