Amazon spent about $1.8 million on a failed AI project and did not flag it for five months, Yahoo Finance reports

Amazon has been moving aggressively to scale artificial intelligence across its business, but a new Yahoo Finance report points to cases where failed AI work became unusually expensive and slow to surface internally.

The report says Amazon spent roughly $1.8 million on an AI task that did not deliver results, and that the problem was not recognized for about five months. The same coverage frames the episode as part of a broader pattern of “catastrophically expensive” overspending on AI projects that do not pan out.

Yahoo Finance also notes that Amazon is planning to spend about $200 billion on AI during the year, underscoring the challenge of running large AI budgets while keeping costs tightly controlled. When AI systems are tested, iterated, and sometimes rebuilt, teams can accumulate compute, engineering, and data-development costs even after a direction is abandoned.

The report’s central claim is less about the existence of failed projects and more about what happens after failure: whether organizations detect problems quickly and reallocate funding before spending escalates. In the described case, that reallocation appears to have been delayed, leaving a costly bill attached to a task that did not succeed.

Amazon is publicly known for its two main technology engines: retail and logistics operations on the consumer side, and Amazon Web Services on the enterprise side. AWS is the platform where customers buy access to cloud infrastructure and, increasingly, AI services such as model hosting and machine learning tooling. Budget pressure and project governance matter in both settings because AI work can involve expensive compute and specialized engineering.

More broadly, the report adds to ongoing scrutiny of how large technology companies manage AI program portfolios. Even when leaders emphasize efficiency and responsible deployment, AI development timelines can be nonlinear, and teams can end up paying for experimentation that later becomes redundant.

Still, key details about how the $1.8 million figure was calculated, what the “failed” task specifically involved, and what internal review or controls were meant to catch the issue are not provided in the information available here. The Yahoo Finance framing suggests a governance gap, but it does not clarify whether the delay was due to reporting mechanics, approval pathways, or technical complexity.

What to watch next is whether Amazon responds with additional transparency about project oversight, cost-tracking, and the criteria used to stop or salvage AI initiatives. Investors and customers will likely focus on whether governance improvements reduce “runaway” spending, and whether AWS can demonstrate that AI scale can coexist with tighter unit-cost discipline.