"Unlimited" AI Tokens Run Out as U.S. Army Burns Through Resources at Record Pace

U.S. Army's AI Tokens Disappear Faster Than Expected

"Unlimited" doesn't always mean infinite. The U.S. Army learned that lesson after its AI token allocation vanished in the blink of an eye. Just one month after the Department of Defense announced that nearly half of its 3.5 million employees were using AI at work, personnel received an email warning that they were consuming tokens much faster than anticipated.

What began as a promise of unlimited access quickly faded. In May 2026, the Army's CIO announced unlimited AI tokens for users. By mid-June, however, the available pool had already been exhausted, forcing the Army to reinstate usage limits. The platform in question is Ask Sage, which provides access to large language models including Alphabet's Gemini, Meta's Llama, and OpenAI's ChatGPT.

The Consumption That Raised Concerns

The scale of usage surprised even insiders. An Army employee, speaking anonymously, said the organization consumed an entire year's worth of tokens for a single service. Ask Sage is used for tasks such as reclassifying personnel descriptions, aligning job roles, experience, and career history. To encourage adoption, the Army had been providing employees with a monthly token allocation that could be expanded when needed.

During Operation Epic Fury in Iran, the Department of Defense reportedly burned through billions of tokens. The surge in usage has sparked questions about whether the tokens used by everyday employees come from the same allocation as those supporting classified operations. The issue remains unresolved as the Department of Defense continues expanding its investment in AI tools while facing resource management challenges.

A Warning for Businesses Adopting AI

The Army isn't alone in rethinking nearly unlimited AI access. Meta, for example, has removed its token usage leaderboard and is reportedly considering limits on token consumption per engineer. Uber has faced similar challenges after engineers reportedly consumed a year's allocation of tokens in just four months, highlighting the importance of governance and cost control.

One Army employee also noted that generative AI tools have not always delivered the expected results. In some cases, the models claimed to have completed tasks that they had not actually performed. While AI can certainly reduce government bureaucracy, the employee argued that deploying it without thoughtful oversight is unlikely to produce reliable, efficient, or trustworthy outcomes.

The broader takeaway is clear: as AI capabilities continue to grow, managing usage and infrastructure costs becomes just as important as adopting the technology itself. The U.S. Army learned that lesson the hard way. Is your organization ready to manage its AI resources?