Generative AI shifts to usage-based pricing amid investment gaps

Major AI providers, including Anthropic, are shifting from flat-rate monthly subscriptions to pay-as-you-go pricing models. This transition is driven by the rise of AI agents, which perform autonomous, repetitive tasks that significantly increase computational demands. Goldman Sachs predicts that global monthly token processing will reach 120 quintillion by 2030, a 24-fold increase from 2026 levels.

In Japan, a survey by FIXER reveals a significant gap in AI investment. Approximately 47.1% of surveyed companies with over 300 employees invest less than 1 million yen in generative AI, while only 2.3% invest more than 500 million yen. This low investment level, combined with the risk of sudden high costs from usage-based billing, may hinder Japanese productivity and international competitiveness.

Despite the widespread adoption of generative AI, macro-level productivity gains remain elusive. While many individual workers report that AI helps them complete tasks faster, data from both Japan and the United States suggest that total working hours have not decreased and organizational productivity has not significantly increased. This discrepancy highlights a challenge in translating individual efficiency into broader organizational or economic value.