‘We created a monster’: companies rein in AI usage as costs strain budgets

{"text":[[{"start":14.3,"text":"Companies that raced to put AI tools in the hands of their workers are starting to rein in their use, as the cost of deploying the technology at scale begins to test corporate budgets."}],[{"start":25.200000000000003,"text":"Amazon, Walmart, Cisco, Uber and Meta are among early adopters that have introduced caps, discouraged wasteful use or pushed employees to cheaper models in a bid to keep AI spending under control. "}],[{"start":38.400000000000006,"text":"The shift marks a new phase in corporate AI adoption. As workers move beyond chatbots to AI agents, which can perform complex tasks autonomously but require far more computing power, companies are being forced to scrutinise whether each query and task is worth the cost. "}],[{"start":56.7,"text":"This has intensified as groups including Anthropic and OpenAI have moved some services from flat subscriptions to token-based billing, which tracks the units of data processed by models. The change has exposed companies more directly to the cost of every prompt and automated workflow."}],[{"start":73.15,"text":"“Compute costs are now beginning to enter the minds of both CFOs and boards. Consumers and businesses have been taught that AI is cheap or free and that is definitely not the case,” said Costi Perricos, global generative AI leader at Deloitte. "}],[{"start":null,"text":"

"}],[{"start":88.15,"text":"Sam Altman, OpenAI’s chief executive, said this month that cost had emerged as a “huge issue” for customers this year. “The issue never came up [last year] . . . People were totally happy with the amount they were spending.”"}],[{"start":101,"text":"Uber president and chief operating officer Andrew Macdonald said it was becoming “harder to justify” its outlay on AI tokens. “It’s very hard to draw a line between one of those stats and ‘OK now we’re actually producing like 25 per cent more useful consumer features,’” he said on a recent podcast."}],[{"start":118.9,"text":"The ride-hailing company has introduced usage caps, limiting employees to $1,500 in monthly token spending on individual AI tools, after blowing through its entire AI 2026 budget by April. Bloomberg first reported the cap. "}],[{"start":132.1,"text":"US retail giant Walmart has similarly limited use of its own internal AI agent by capping the number of tokens employees can use."}],[{"start":140.6,"text":"Suresh Kumar, Walmart global chief technology officer, said usage of the company’s Code Puppy vibe-coding platform had “really skyrocketed”. “This is now an opportunity for us to take a step back,” he said, adding that employees are being tasked with identifying the right tools for tasks."}],[{"start":159.15,"text":"Jeetu Patel, president and chief product officer at Cisco, said companies were having to balance a desire to use the technology to deploy agents with the cost and availability of tokens. "}],[{"start":170.5,"text":"“The amount of infrastructure needed for an agent is meaningfully higher than for a chatbot,” said Patel. “For every human you might have 10, 100 or on the aggressive side 1,000 agents . . . They just keep working and that consumes a chunk of [compute].”"}],[{"start":null,"text":""}],[{"start":187.45,"text":"Goldman Sachs analysts last month predicted that use of AI agents would result in a 24-fold increase in token consumption by 2030 and that the huge rise in demand would exacerbate a shortage of chips over the next 12 to 18 months."}],[{"start":203,"text":"While token usage and AI spending by businesses continue to grow, efforts to curb costs could weigh on the growth of the world’s largest AI labs such as Anthropic and OpenAI, which plan to go public later this year at near-trillion-dollar valuations."}],[{"start":218.15,"text":"Since the start of the year, Chinese AI models have overtaken their US counterparts in token consumption, according to data from OpenRouter, an aggregation platform that allows users to access multiple AI models."}],[{"start":231.05,"text":"China’s cheaper energy and more efficient models have allowed the country’s AI labs to charge less than leading US groups for tokens, giving China a new edge on the AI battleground. "}],[{"start":null,"text":""}],[{"start":242.65,"text":"Smaller companies are also feeling the cost pressure. Software group Workato said its AI use exploded after its 1,300 employees started using AI agents last summer. “It took off like wildfire, people started really transforming their jobs with agents,” said chief information officer Carter Busse."}],[{"start":261.85,"text":"But the company got a shock when Anthropic switched it over to token-based pricing in May. “Our spend went up 7x the first day and I’m like, oh shit, we created a monster,” said Busse. “[Large language model] companies have been subsidising all of our usage and now no longer. User-based pricing shelters you.”"}],[{"start":282.35,"text":"Rather than restrict access, Busse is trying to rein in spending and has pivoted his twice-weekly AI enablement sessions to push cost-saving ideas, such as defaulting to Anthropic’s older, cheaper models. “Instead of innovation, let’s talk about AI financial responsibility,” he said. "}],[{"start":300.35,"text":"Larger companies have also started to change incentives and internal messaging around AI adoption. "}],[{"start":306.15000000000003,"text":"Amazon warned employees last month that they should halt using “AI just for the sake of using AI” after engineers started to deploy agents for the sake of climbing internal leader boards. "}],[{"start":316.6,"text":"The group has been forced to pivot its approach to measuring adoption in a bid to rein in costs attached to tool misuse. Meta took similar steps in April. "}],[{"start":326.5,"text":"Both Amazon and Meta have built and deployed their own models but are also reliant on third parties such as Anthropic for models and tools such as Claude Code. "}],[{"start":335.55,"text":"AI groups and platforms are trying to preserve adoption by steering users away from expensive frontier models when they are unnecessary and replacing them with cheaper versions. "}],[{"start":345.90000000000003,"text":"Kyle Daigle, chief operating officer of GitHub, said Microsoft had engaged customers ahead of implementing pricing changes to discuss “fit and purpose” with developers. "}],[{"start":356.50000000000006,"text":"“The question for the user is, ‘what models are best suited [for the job]?’ You’re not always looking for a frontier model,” he added. "}],[{"start":364.50000000000006,"text":"Microsoft and other AI platform providers, including Amazon and Google, have already launched tools that route customers’ queries and tasks to the most relevant model within a series chosen by a customer to more effectively control costs. "}],[{"start":379.70000000000005,"text":"Meanwhile, some companies have told workers to use open-source models that can be run locally on their own servers or personal devices, reducing the bill they pay to AI labs and cloud providers. "}],[{"start":392.1,"text":"But customers are still weighing higher costs against the promises they have made to investors about AI’s impact on their own bottom line and workers’ productivity. "}],[{"start":401.75,"text":"“Our engineers want more tokens . . . We have to figure out a way to fund it,” Cisco’s Patel said.  "}],[{"start":407.85,"text":"Additional reporting by Gregory Meyer in Bentonville "}],[{"start":419.35,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1781851941_1472.mp3"}