AI spending nightmare: Companies spend over $500 million in 30 days on Anthropic’s Claude

Companies across industries are rapidly adopting artificial intelligence (AI)-powered tools to fast-track operations and improve productivity. However, the reality comes at a heavy price, urging companies to rethink how they manage AI usage. According to an Axios report, an AI consultant’s client spent over half a billion dollars in just 30 days due to unrestricted employee access to Anthropic’s Claude chatbot. This highlights the ongoing company-wide trend of maximising AI usage, or “tokenmaxxing.”

Must read: Why AI agents are costing companies more than the employees they replaced

Tokenmaxxing is costing companies millions

Tokenmaxxing is the term that refers to maximised AI usage. Therefore, if employees start generating massive amounts of prompts and outputs and automating workflows, they quickly spiral out of control, costing companies a millions in AI bills.

This comes as a warning on how quickly AI usage could go out of control if companies don’t set budgets, quotas, or monitoring systems for employee usage. Reportedly, Microsoft cancelled its Claude Code licenses for employees and is encouraging them to use the internal AI tool by June 30.

On the other hand, Uber's COO highlighted that AI costs have started to get "harder to justify." Now, it turns out “just keep prompting” wasn’t a sustainable business model after all. According to a Financial Times report, Amazon has also shut down its internal AI-use leaderboard after employees started using AI unnecessarily just to climb the rankings. This increased Amazon’s infrastructure and AI operating costs.

Must read: Indian government, tech firm join forces to test Anthropic’s Claude Mythos: Report

The leaderboard was reportedly built to encourage employees to use AI-powered tools. However, these tools come with “usage-based costs,” hence each AI prompt adds to higher computing demand, so a few extra clicks per employee can snowball into a huge expense. Now, Amazon plans to bring tighter controls over AI usage.

AI adoption challenges

The Axios report further highlighted four major challenges companies face amid AI adoption pressure. Firstly, Sophia Velastegui, CEO of Velastegui Ventures and former chief AI officer at Microsoft, highlighted that the first challenge is “Most people default to automating tasks they dislike rather than tasks most valuable to the company.”

Another challenge would be the growing AI usage costs. Axios quoted a CTO saying that a simple prompt, like asking for the weather, can carry heavy token costs. Another challenge is said to be companies struggling to bring efficient adoption and limited AI understanding among staff.

Lastly, the concerns of sharing proprietary data reduce the effectiveness of AI agents. Therefore, companies may need to bring stronger governance, clearer usage guidelines, and better AI literacy across teams to generate the expected ROI for the cost it's paying.