AI adoption surge sparks corporate backlash over rising costs, says Jefferies
New Delhi, May 29
The rapid adoption of artificial intelligence across industries is triggering concerns over surging costs and inefficient usage, even as companies continue to aggressively invest in the technology, according to a latest report by Jefferies.
The report said that while AI adoption is accelerating globally, many corporations are now facing a backlash over escalating compute expenses driven by employee experimentation with AI tools.
The report noted that "AI for now is costing more money than it is saving", highlighting growing concerns among hyperscalers and corporate users over rising expenditure despite falling token costs.
Jefferies said a specialist AI conference attended by chief information officers this week highlighted expectations of "continuing massive increases in AI capex spending", estimating that global AI-related capital expenditure could touch "an aggregated US$4.7tn globally by 2029 of which US$2.6tn will have been spent on tech hardware".
The report observed that firms initially encouraged employees to experiment freely with AI tools, but the resulting compute bills are now prompting a rethink.
According to the report, "Microsoft has started cancelling its internal Claude Code licenses and is moving its teams back to GitHub Copilot CLI coding tool", while Uber's Chief Technology Officer reportedly warned internally that "AI usage burned through an entire year budget in just four months".
Jefferies also pointed to a growing phenomenon termed "tokenmaxxing", where employees excessively use AI systems to improve internal productivity scores.
Quoting a Financial Times report, Jefferies said Amazon staff were allegedly using AI tools for "unnecessary tasks to inflate usage scores".
The report added that measuring employee AI activity may not accurately reflect productivity gains.
Despite the cost pressures, Jefferies said there is zero sign of the capex slowing yet, particularly in semiconductor stocks, with the Philadelphia Semiconductor Index rising by 80 per cent year-to-date.
The report linked the AI boom to Jevons Paradox, an economic principle stating that greater efficiency can ultimately increase overall consumption.
According to Jefferies, collapsing token costs have accelerated demand for AI tools instead of reducing spending.
The report also warned that "a lot of capital will be destroyed in this AI capex cycle", drawing parallels with the Dot-com bubble and railroad boom-bust cycles in the United States.
However, Jefferies maintained that AI remains a transformative technology in the long term despite the possibility of an over-investment correction in the near future.
The report further noted that Anthropic's annualised revenue run rate reportedly rose from "US$9bn at the end of 2025 to over US$44bn as of early May", underscoring the rapid commercialisation of AI services globally.
— ANI
Reader Comments
Interesting article. I work at a tech startup in San Francisco and we're seeing the same trend. The "tokenmaxxing" phenomenon is real - people are generating thousands of AI queries just to look productive. It's like the early days of the internet where everyone was building a website just because they could. The bubble will burst eventually.
The Jevons Paradox reference is spot on! As Indians, we've seen this before with mobile data - when Reliance Jio made data cheap, usage exploded. Same thing is happening with AI. Cheaper tokens mean more usage, which means higher total costs. Companies need to set clear policies instead of just letting everyone experiment freely.
The comparison to the Dot-com bubble is unsettling but probably accurate. I remember the 2000 crash when I was studying in Hyderabad - a lot of companies burned through investor money chasing the next big thing. Having said that, AI is genuinely transformative. The trick is to focus on specific use cases rather than blanket adoption.
Respectfully, I think the backlash is overblown. Every transformative technology goes through this phase. Yes, costs are high now, but AI capabilities are improving exponentially. The $4.7 trillion investment figure might seem insane, but infrastructure investments always look excessive during the build phase. Remember how people mocked Bezos for Amazon's early losses?
The Amazon employees using AI for unnecessary tasks part made me chuckle 😂. Typical corporate behaviour - you measure something, and people will game the system. In India, we have a term for this: "jugaad". Companies need
We welcome thoughtful discussions from our readers. Please keep comments respectful and on-topic.