Why companies are burning through AI tokens and racking up eye-watering bills

Companies are struggling with escalating costs from AI adoption, with some allocating massive budgets to AI tools while others reconsider hiring human workers due to expenses. A company reportedly spent $500 million in a single month on Claude licenses from Anthropic, while others are delaying employee raises to offset AI-related spending. AI tokens, the units of data processed by models like OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini, are driving costs up as businesses use them for complex tasks such as coding and multi-step workflows. Token consumption has surged as companies integrate AI into operations. Andrew Forde, head of AI at KPMG Canada, noted that tasks like writing software can require 15,000 times more tokens than a simple text response. Users pay for both input and output tokens, meaning every AI interaction incurs costs. Royal Bank of Canada saw its AI token usage jump over 500% in its second quarter, while Shopify reported 100% daily AI tool adoption among employees, offering unlimited token access. Google revealed it processes over three quadrillion tokens monthly—a sevenfold increase from a year ago. In Silicon Valley, engineers are competing to maximize token usage, a trend dubbed ‘tokenmaxxing,’ to demonstrate productivity. Nvidia’s CEO Jensen Huang has suggested that engineers should focus on high-value AI tasks rather than basic coding to justify their salaries. The rapid adoption of AI for advanced tasks is outpacing cost reductions, forcing businesses to rethink budgets. Some are now finding that human labor may be more cost-effective than deploying AI for certain functions. As token consumption grows, companies are grappling with how to balance innovation with financial sustainability.