Amazon increased budget for AI projects by 860 percent

The rapid implementation of AI technologies is causing unexpected financial challenges for major corporations. According to the Financial Times, Amazon significantly overspent on internal AI projects and drastically exceeded its allocated budget, reports Ixbt.com. reports .

Specifically, a project on the marketplace designed to automatically compare author data with product cards cost the company $1.8 million. Anthropic's Claude Sonnet model was used for this task, and it turned out that this figure was a staggering 860 percent higher than the initially approved budget.

Lack of control and hidden costs

The situation was further aggravated by the fact that this financial discrepancy was not noticed immediately. The exorbitant costs were only discovered five months later, and most regrettably, despite all the funds spent, the project was ultimately never implemented in practice.

According to the Financial Times, such cases were not isolated exceptions. The company overspent an additional $541,000 on developing a financial audit tool and nearly $134,000 on an AI project aimed at optimizing logistics processes.

Internal criticism and a change in approach

Amazon engineers described such incidents as «catastrophically expensive» during a closed internal meeting. One employee specifically emphasized that it is very difficult to accurately pre-estimate the real costs of AI projects.

Following this, Amazon management decided to shut down the KiroRank internal rating system, which had been encouraging employees to actively use AI agents. Now, company management is urging its employees «not to use AI simply for its own sake».

Under the current circumstances, soaring costs have forced the corporation to rethink its approach to using advanced technologies. In turn, Amazon representatives stated that the data cited in the Financial Times is based only on isolated cases and does not fully reflect how the majority of teams apply AI in internal processes.