Microsoft Shifts AI Default Model Amid Spree of Unchecked Spending
Microsoft's employees were encouraged to use AI aggressively for years, but when the bills arrived, the company took notice. In 2026, a new column appeared on the employee compensation spreadsheet tracking AI tool spending. One employee in the Customer and Partner Solutions organization spent $28,000 in just 28 days.
The median AI tool spending per 28-day window for US employees was around $300, but team-level medians revealed a wide range. Some departments saw individual maximums of over $10,000, while others had floors as low as tens of dollars.
Microsoft's internal leaderboards made consumption visible, and some employees deliberately burned AI tokens on low-value or useless queries to climb the ranks. In response, Jay Parikh, CoreAI EVP, sent an internal memo stating that 'tokenmaxxing' was not what the company optimized for.
The company has since shifted its workloads to OpenAI's GPT-5.6 Sol as the default model inside GitHub Copilot and related workflows, citing 'greater value from our token investment.'