The past few days have been full of bad news for the AI industry. The headlines paint a picture of an industry confronting growing pushback on multiple fronts, from political and regulatory headwinds to disappointing financial returns to poor results from real AI deployments. These are the stories that fed the narrative.
The AI industry’s shifting narrative on jobs
In a notable shift in tone, OpenAI CEO Sam Altman acknowledged that artificial intelligence is unlikely to trigger the “jobs apocalypse” he had previously warned about. Speaking virtually at a Commonwealth Bank event in Sydney, Altman downplayed earlier predictions of widespread job displacement, admitting his early economic intuitions were “pretty wrong” regarding immediate white-collar layoffs. Altman said the hit to entry-level office work has been significantly smaller than he expected.
AI company executives have sometimes dramatized the potential negative effects of AI as a way of hyping the power of their models. But now that popular resistance to the technology, born partly of job-loss fears, is threatening the construction of new data centers, AI companies may be trying to tone down the rhetoric.
Communities often lure big AI data center projects with tax breaks. Last week, Pennsylvania lawmakers introduced bills that would repeal tax breaks for AI/data centers and give municipalities authority to impose an 18-month moratorium on projects. The fact that Republicans are sponsoring data center moratoriums in a state that’s aggressively courting AI infrastructure is notable. It’s another sign of shifting public opinion on AI.
A mid-May Gallup poll found that more than two-thirds of adults oppose the construction of AI data centers, with a majority saying they’d prefer to have a nuclear power plant in their backyard instead.
Illinois passes its AI law
Adding to the industry’s regulatory headaches, Illinois passed a major new AI accountability bill, SB315. The law is the first in the country to mandate independent, third-party safety audits, risk disclosures, and incident reporting for large frontier AI developers. Industry groups warned that the law could increase compliance costs and slow innovation.
The AI industry has had a strong ally in the Trump White House but has failed to persuade Congress to ban new state-level AI legislation. Thousands of AI-related bills have been introduced at statehouses across the country. So AI companies are now bracing for a potential patchwork of state-level regulations that could complicate nationwide operations, and they are changing their strategy accordingly.
OpenAI’s chief lobbyist and political operative Chris Lehane says the industry is increasingly engaging state lawmakers to promote weaker or toothless AI safety laws. Lehane told Politico that OpenAI hopes to shape AI policy in a “critical mass” of key states such as California, New York, and Illinois, with the hope that other states will pass similarly industry-friendly laws.
Most AI hyperscalers apparently aren’t making money
Fresh profitability modeling data from the investment bank Panmure Liberum suggests that most tech companies spending massive amounts on AI data centers and other infrastructure are nowhere near seeing a return on the investment. The numbers were part of a new Financial Times opinion piece written by Panmure Liberum director Joachim Klement titled “The impossible maths of the AI boom”.
They show that, under best-case scenario models, Microsoft’s AI initiatives are returning -9% on investment, while Google’s ROI stood at -15%, Meta’s at -28%, and Oracle’s at a steep -35%. Only Amazon managed to eke out a slightly positive return. The figures cast doubt on the near-term profitability of the massive capital expenditures many tech giants have poured into AI infrastructure and model development.
The end of Uber ‘tokenmaxxing’
Enterprise adoption is also showing signs of strain. Among the first impactful applications of AI in big companies are AI coding tools such as Anthropic’s Claude Code and OpenAI’s Codex. Big company executives have been urging software engineers to rely more on the tools to increase productivity. But the tools aren’t cheap.
An Uber exec revealed that the company had burned through its entire annual AI token budget in just four months after giving thousands of developers access to Anthropic’s Claude Code tool. Some engineers were racking up monthly bills between $500 and $2,000. Now Uber is saying its massive token spend is becoming “harder to justify,” and that the company will rethink its budgeting strategy.
Starbucks’ AI counting flop
Starbucks is another big company that was talking big about implementing AI tools. News broke last week that the company had quietly discontinued an AI-powered inventory tool that was supposed to optimize store operations less than a year after rollout.
Reuters reports that the coffee giant quietly discontinued the Automated Counting system developed by NomadGo after store employees reported persistent inaccuracies on basic tracking tasks, including miscounting milk carton volumes and failing to accurately track back-of-house beverage syrups.
There’s a lot riding on AI deployment
Taken together, the developments last week paint a picture of an industry confronting growing pushback on multiple fronts. Societal resistance to the technology may only grow as AI adoption’s negative side effects begin to appear, job losses among them.
Big tech and its investors have a lot riding on the productive deployment of AI in enterprise settings. The generative AI boom that started with ChatGPT has now entered a phase where enterprises expect the technology to bring measurable increases in efficiency and productivity to operations. Because of the hype and high valuations of AI companies, failures will be magnified, even as successful deployments increase. (That pressure is now colliding with Wall Street’s expectations: Anthropic announced on Monday it has confidentially submitted a draft S-1 to the SEC, giving it the option to pursue an IPO pending review.)
Political and social resistance to the technology is likely to grow as the negative side effects of AI adoption begin to appear—job losses being one of them. Many in the middle class believe AI will enrich and empower a small set of Silicon Valley types, while the technology itself is used to distract, addict, surveil, and even control normal people.
