AI token pricing challenges

Major technology firms including Microsoft, Alphabet Inc., and Anthropic have invested heavily in large language models, but pricing AI services remains difficult due to unpredictable token consumption. Token costs per unit have fallen, yet overall usage has surged, making total expenses hard to forecast. Goldman Sachs projects token consumption will increase 24 times between 2026 and 2030, reaching 120 quadrillion tokens monthly. Companies like Microsoft and Uber have reportedly faced unexpected token costs, prompting efforts to manage usage. Industry executives are exploring various pricing models, including flat fees, per-result billing, and bundled incident charges, but no consensus has emerged. The non-deterministic nature of AI outputs complicates cost prediction, and any pricing structure could be disrupted by changes in provider pricing. Smaller firms sometimes use flat-fee personal accounts, but observers expect major vendors to clamp down as shareholder pressure for profitability grows.