That’s according to a report Thursday (July 30) from the Financial Times (FT), citing multiple sources familiar with the matter.
Those sources said that Amazon senior engineers told colleagues at a staff meeting earlier this week that efforts to switch tasks from conventional programming to using artificial intelligence models had caused “unplanned” spending.
The FT notes that the issue underlines the trouble even the largest tech companies are having with weaving AI into day-to-day operations without spending too much.
“It’s difficult to figure out how much anything [AI related] costs,” a senior Amazon employee told the FT.
According to the FT’s sources, employees learned during a presentation this week about an incident in which Amazon spent $1.8 million on matching author details with listings on the company’s eCommerce site using Anthropic’s Claude Sonnet despite the deployment failing.
This meant the project ran 860% over budget, with the spending taking five months to detect, the sources added. Engineers reportedly told staff the overspending wasn’t a one-time thing. In another incident, Amazon incurred around $541,000 in unanticipated costs tied to creating financial auditing tools.
“As with any new technology, we’re experimenting, learning and improving how we use it, including how we drive cost efficiencies,” an Amazon spokesperson said in a statement to PYMNTS.
“Cherry-picking small, isolated examples where teams are learning from one another and portraying them as business as usual doesn’t reflect how teams across Amazon are using AI.”
The news follows reports from earlier this month that AI spending by the world’s biggest tech companies have left investors feeling uneasy.
Meanwhile research by PYMNTS Intelligence finds companies from a range of industries investing more in AI, though for different reasons.
“Financial firms are funding AI to improve productivity, sharpen competitive positioning and reduce risk. Healthcare firms are still using budgets to test what works. Media and advertising firms are moving quickly, often with strong executive backing, but with less reliance on hard financial returns,” the report said.
“The spending pattern suggests that AI is entering a more practical phase. Like a company moving from blueprints to construction, enterprises are beginning to decide which projects deserve real capital and which still need proof.”
