Amazon Ends AI Usage Leaderboard After Rising Costs and Employee Tokenmaxxing

Amazon has discontinued its AI usage leaderboard after employees began using artificial intelligence for unnecessary tasks, which increased company costs. The decision follows reports that some workers engaged in “tokenmaxxing,” a practice where employees inflated their AI usage to improve their ranking on the company’s internal leaderboard.

Key Highlights

• Amazon discontinued its AI leaderboard after employees inflated usage to improve rankings.
• The company replaced raw AI usage metrics with a focus on useful code and meaningful work.
• Amazon expects to spend nearly $200 billion on AI and data center infrastructure in 2024.
• Industry-wide, companies are shifting to consumption-based AI pricing models, raising costs.

Amazon Shuts Down KiroRank System

According to the Financial Times, Amazon informed staff that it has taken its “KiroRank” system offline. This system scored users of the company’s Kiro developer platform based on their AI activity. Dave Treadwell, a senior vice-president at Amazon, told employees that the leaderboard was created with positive intentions. However, it led to higher computing costs as employees used AI tools for tasks that were not necessary, only to boost their scores.

The term “tokenmaxxing” refers to the practice of increasing AI token consumption. AI tokens are units of data processed by AI models. Some employees reportedly used AI for routine or unnecessary work to climb the leaderboard. Treadwell advised staff not to use AI simply for the sake of using it.

Shift to New Evaluation Metrics

Amazon has now adopted a new metric called “normalised deployments” to assess AI tool adoption within the company. Instead of tracking total AI token usage, the new system evaluates whether engineers use AI to produce useful code and complete meaningful work. This change aims to encourage practical and efficient use of AI rather than high-volume consumption.

The move comes as the technology sector faces growing concerns about the costs of advanced AI systems. Companies like Anthropic have switched to consumption-based pricing models, which has increased expenses for some customers. Some industry experts argue that hiring skilled human workers can sometimes be more cost-effective than relying heavily on AI coding tools such as Claude.

Industry-Wide AI Cost Concerns

Microsoft has responded to rising AI costs by reducing Claude Code licenses to lower operating expenses ahead of the new financial year in July. Earlier this year, Praveen Neppalli Naga, CTO of Uber, stated that the company’s AI spending had already surpassed expectations due to rapid adoption of advanced coding tools like Anthropic’s Claude Code.

Amazon relies heavily on Anthropic’s AI models and continues to invest in AI infrastructure while also implementing layoffs to manage costs. Reports indicate Amazon is expected to spend nearly $200 billion in capital expenditure this year, with most funds allocated to AI systems and data center infrastructure.

The technology industry continues to debate the balance between AI investment and operational efficiency as companies seek to control costs while advancing AI capabilities.