Microsoft Employees Spend Up to $28K Monthly on AI Tokens

Microsoft workers rack up massive AI compute bills

Some Microsoft employees are spending extraordinary amounts on AI compute resources, with one worker in the Customer and Partner Solutions division consuming $28,000 worth of tokens in a single 28-day period, according to internal compensation data first reported by Gadget Review.

The figures emerged from voluntary, self-reported entries in an internal spreadsheet where Microsoft employees share compensation details. In 2026, the spreadsheet added a column for monthly AI usage, allowing workers to track their AI consumption in dollar equivalents.

Of nearly 600 U.S. employees who contributed to the spreadsheet, approximately 350 shared AI usage data. The median monthly spend was significantly lower at around $300 per employee, suggesting the $28,000 figure represents an outlier rather than typical usage. Microsoft employs more than 223,000 people globally, so this voluntary sample captures only a fraction of the workforce.

Usage varies dramatically by department

The internal data reveals substantial differences in AI consumption across Microsoft's business units. CoreAI leads with a median usage of $975 per employee monthly, followed by Security at $526 and Microsoft AI at $490. Cloud and AI, Experiences and Devices, Azure, and Customer and Partner Solutions show progressively lower median figures.

Individual spending within departments shows even wider variation. Some CoreAI employees spent up to $16,000, while Cloud and AI workers reached $15,000 and Security personnel hit $10,000 in monthly token consumption.

Company shifts from encouragement to oversight

After years of actively incentivizing employees to adopt AI tools, Microsoft is now monitoring token consumption more closely to identify heavy users. The shift reflects growing concern about a practice called "tokenmaxxing," where employees deliberately consume large volumes of tokens to appear on internal tracking dashboards.

The phenomenon isn't unique to Microsoft. Similar patterns have emerged at Amazon and other technology companies as AI tools become more accessible to workers.

CoreAI Executive Vice President Jay Parikh addressed the issue directly in an internal memo. "Tokenmaxxing is not what we are optimizing for," Parikh wrote. "I want all of us focused on maximising outcomes that move the needle for our customers and our business."

Why it matters

This data offers a rare window into how enterprise AI adoption translates to actual compute costs at scale. As companies rush to integrate AI capabilities, the gap between median usage ($300) and outlier spending ($28,000) suggests organizations need clearer guidelines about appropriate use cases. Microsoft's pivot from blanket encouragement to outcome-focused monitoring signals a broader industry maturation—moving past "AI for AI's sake" toward measurable business value. For technology leaders, the lesson is clear: adoption metrics matter less than impact metrics.

The details were first reported by Gadget Review based on internal Microsoft spreadsheets.