Earlier this month, Databricks CEO Ali Ghodsi gave a keynote to his entire engineering department. During the speech, he singled out one engineer, who had spent more than $7,000 in AI tokens — an atomic unit of data processing for an AI model to generate text — over a two-week period in January. Ghodsi’s intent wasn’t to chide the employee for overspending. Instead, it was quite the opposite.

“We had everybody in engineering clap for him, and celebrate what he did,” Ghodsi tells Forbes. “I’m trying to get everybody to use this stuff.”

The engineer used the tokens through Isaac, the company's internal coding tool, which pulls from several different AI models, including those from Anthropic and OpenAI, Ghodsi says. Databricks, the $134 billion-valued data infrastructure company, isn’t alone in its fervor to get its engineers to embrace AI. Over the past few months, as vibe-coding has exploded and AI-generated code has become more reliable, Silicon Valley has pushed its software developers to go all in. How to quantify it? By measuring the amount of money that engineers are spending in AI tokens (with the bill usually footed by employers). The idea is to max out AI budgets to supercharge productivity.

Token prices vary. For cheaper models and basic tasks, they can typically cost a few cents per million tokens. For more complex computations and premium models, prices can spike to anywhere from $20 to more than $100 per million tokens. For example, with Anthropic’s Claude Opus 4.6, the AI giant’s latest flagship........

© Forbes