Answer: Goldman Sachs Research projected token consumption could multiply 24 times to 120 quadrillion tokens per month between 2026 and 2030 as consumers and enterprises adopt AI agents. The report linked agentic AI growth to increased computing volume for hyperscalers running LLMs.
Goldman Sachs Research also said it expects inference cost per token to fall rapidly and cited a decline of 60%–70% per year. The piece said this dynamic is connected in the report to expectations around a gross margin “inflection.”
Source: Goldman Sachs