What you might not know is that this change has come about from the top announcements from the tech world recently; After all, it’s always sunny in Silicon Valley. Like Google I/O in May, many techies at Microsoft’s Build conference were talking about tokens, the metric by which AI hints and answers are measured (a token, strangely, is about three-quarters of a word on average).
Both conferences also focused on claims of marginal AI that are at least questionable. DeepMind CEO Demis Hassabis at Google I/O: “Artificial General Intelligence is just a few years away… We are standing at the foothills of the Singularity.” Microsoft AI CEO Mustafa Suleiman: “The scaling laws hold… We are building towards what we call humanistic superintelligence.”
Investors also showed no signs of losing their AI optimism this week. Nvidia’s stock fell for a few days, but rallied after CEO Jensen Huang insisted that AI agents will run everything, everywhere in the future (presumably once they stop deleting databases). Anthropic, OpenAI and SpaceX continue to pursue trillion-dollar IPOs, largely based on the untested concept of AI data centers in space.
Elon Musk found the cheat code to capitalism. SpaceX’s IPO proves it.
But outside the AI bubble, a backlash is brewing, and not just among students, with pro-AI commencement speakers being criticized.
According to a Pew survey conducted in March, only 10 percent of Americans say they are excited about the future of AI; That same month, an NBC poll found nearly 80 percent of registered American voters said neither Democrats nor Republicans were doing a good job on the AI front. This number is also reflected in an April survey of white-collar workers: 80 percent are outright refusing to use AI even if it is mandatory. In the past 30 days, 54 percent of workers reported bypassing company AI tools and completing tasks themselves.
These numbers suggest general strike-levels of dissatisfaction with AI in every industry, if not an outright revolutionary mood in real America beyond Silicon Valley and Wall Street.
Data center opposition, driven by the 70 percent of Americans who say they don’t want a data center near them, is likely to increase going forward – especially now that they are producing solid results. According to Data Center Watch, at least 48 data center projects are blocked or delayed in 2025, and the fight is becoming more fierce.
Take the planned Stratos data center in Utah, where local opposition forced VC and Shark Tank investor Kevin O’Leary to reduce its land use by 75 percent. “We messed up,” O’Leary told local TV news on Friday. “We offended a lot of people.”
And the threat of the electoral guillotine may explain why politicians are starting to propose serious action.
Just this week, Senator Bernie Sanders came out in favor of the American public owning a 50 percent stake in AI companies, former presidential candidate Andrew Yang proposed an AI tax, and President Trump finally signed an executive order on AI regulation, which his AI czar, Silicon Valley titan David Sachs has long opposed. Finally, on Friday, New York state legislators sent a one-year data center moratorium to the governor’s desk.
The White House AI executive order was announced as Microsoft CEO Satya Nadella was making rosy announcements on AI at Build, adding to the real sense that we’re watching a tale of two worlds – anti-AI people versus an out-of-touch AI regime that essentially says, let them eat tokens.
But hold on to the revolution: Just beneath the surface (and the Microsoft Surface Ultra), the AI system is showing signs of breaking on its own – and it’s all because of those tokens.
Silicon Valley’s AI response begins
When it comes to companies that are true believers in AI, they don’t come truer than Uber. The rideshare giant says 90 percent of its engineers use AI tools, mostly Anthropic’s cloud code. About 10 percent of Uber’s codebase is written by AI agents. Uber had leaderboards that encouraged the use of AI tokens whenever possible; In Silicon Valley, this is known as tokenmaxing, and it was really hot in 2025.
Then the tokenmaxing bill came due. “I thought I would need a budget [for 2026] “It’s already been blown away – less than four months into the year,” CTO Nappalli Naga told The Information on April 14.
However, at the time, the information didn’t make much of an impact in the AI news cycle — not until Uber’s COO confirmed in late May what it meant. Naga’s busted budget was a “head-scratching moment”, Andrew Macdonald told the Rapid Response podcast. Such spending “becomes difficult to justify because AI is not free…we have to start talking about token consumption.”
Just like that, we started talking about token consumption. Axios reported that an unnamed company had burned half a billion dollars of tokens in a single month “after failing to set usage limits on cloud licenses”. Next, we learned that Amazon and Meta have shut down their own internal AI leaderboards; Other companies like Walmart and Starbucks have scaled back their AI agent plans.
mashable light speed
In a leaked email, a senior vice president at Amazon told employees to “stop using AI just for the sake of using AI.” You’d be forgiven for thinking that this would destroy a huge part of OpenAI and Anthropic’s business model. Both companies have spent years building models that consume increasingly more tokens. Now they are promoting agents that can consume tokens on steroids – often 24 times more than a regular model.
No matter how high-minded their missions may be, both companies are in it to sell tokens.
Why did tokenmaxing die?
A scene from the data center protest in Tucson, Arizona.
Credit: Mamta Popat/Arizona Daily Star via Getty Images
Sensing the direction of the wind, some AI leaders have started saying such things openly. Ravi Kumar S, CEO of AI IT firm Cognizant, called tokenmaxing “a vanity metric” at a Fortune conference on Monday. Kumar took aim at OpenAI’s Sam Altman and Anthropic’s Dario Amodei, accusing them of “fear-mongering”. Altman and Amodei have rejected previous predictions of the apocalypse of AI jobs as they now have an IPO coming.
Both CEOs are also beneficiaries of user confusion over the complex costs of AI. Earlier this year, Anthropic quietly changed cloud pricing for many customers, charging them per token. OpenAI is considering abandoning its “unlimited” ChatGPT plans – quite a change from a year ago, when Altman promised to deliver “intelligence much cheaper than the meter”.
The change isn’t just happening at the two AI giants. Microsoft starts cutting token costs for itself And Driving up token prices for everyone else – even before those rosy announcements from the build.
Thanks to the AI Industry for Tech Value Growth: See the Full List
In May, Microsoft began revoking developers’ access to Cloud Code, pushing them onto Microsoft Copilot instead. On June 1, Github Copilot users were switched from a fixed subscription to a per-token subscription model. Reddit is full of angry users complaining about how expensive their AI prompts have suddenly become. In an extreme case, one cloud user blew 50 percent of his monthly credit on a single prompt.
“At the beginning of the year,” Altman said in an OpenAI livestream this week, “people were perfectly happy with the amount they were spending… now, suddenly [it’s] A huge issue.” In a CNBC interview on Monday, Altman acknowledged “a lot of waste” in AI spending, and said companies were asking, “How long do I have to wait [AI benefits] To show in revenue?” Altman said, adding that it was a “fair issue.”
And the closest Altman came to answering the pertinent issue? “The industry will figure it out very quickly…in the next year or two.”
Will Vibe Shift Burst the AI Bubble?
However, how long OpenAI and Anthropic will have to resolve this issue largely depends on what happens with their IPOs.
“No one knows when it will all collapse, but 2026 will be remembered as the year retail investors were left holding the bag,” Gary Marcus, a professor and leading generic AI critic, predicted on Monday.
Marcus, who is increasingly being proven right in predicting AI problems after 2022, may still be off base here. But he has a hunch, based on comments from Anthropic co-founder Daniela Amodei, that both companies had spent so much money that they were “months away from bankruptcy” and had “run out of options” other than filing for a trillion-dollar IPO.
In particular, OpenAI has long been losing more than a billion dollars per month – the cost of offering the ChatGPT service to millions of people for free.
This tweet is currently unavailable. It may be loading or may have been removed.
Financial bubbles built around technologies always end with a the Emperor’s New Clothes Moment. After all, plenty of people are pointing and laughing that the courtiers can’t campaign any longer.
This is what happened to end the dotcom bubble in 2000. A business deal was struck that was so ridiculous on the surface (the world’s largest media empire, taken over by the guys who delivered dial-up Internet via CD?!) that the market couldn’t help but point and laugh. The atmosphere changed. Over-hyped, unprofitable dotcom companies began to look naked and soon the stocks were in decline.
human fare and hallucinations
Times have changed, and the AI bubble is a much tougher thing than its dotcom predecessor. It’s built on top of the company that is currently making money from all this. NVIDIA has sold picks and shovels to AI gold rush seekers for so many years that they have begun to seem invincible. Yet Nvidia is also learning a lesson about the skyrocketing cost of AI.
“The cost of compute far exceeds the cost of employees,” an Nvidia executive told Axios in April. Therefore Nvidia is also vulnerable to tokenmaxing. And that’s why the hottest thing in AI these days is hiring humans, because they’re becoming cheaper than AI – and they’re needed for quality control on AI’s output anyway. Cognizant’s Kumar claimed that his AI company hired 20,000 graduates last year, and even more this year – a vibe change if we’ve ever seen it.
So the Jobspocalypse vibe has changed. The token vibe has changed. And the AI data center-building vibe has changed, too – not just in terms of public and environmental opposition, but in the fact that there aren’t as many data centers under construction as we expected. (Gadfly journalist Ed Zitron has done yeoman’s work here, scouring satellite photos of data center sites for signs of construction).
What is left? Arguably, the only vibe that hasn’t transferred is the hallucination vibe, in which users still don’t know how often most AI models hallucinate. For example, Google won’t say how often Gemini 3.5 flashes hallucinations, but a December Google study found that Gemini may be just as accurate 68.8 to 83.8 percent of the time.
Gemini 3.5 How often does Flash hallucinate or lie? Google isn’t saying.
And hallucinations aren’t hard to spot these days. The hallucination that OpenAI, Anthropic, and SpaceX are real trillion-dollar AI giants that deserve to be listed in a top index fund despite being unprofitable (breaking news: as I wrote this, the S&P 500 officially opted out of that hallucination).
The hallucination that Nvidia will always be on top, even if the companies making the majority of its business are developing their own AI chips (that’s why Michael Burry, big small Boy, continues to short the stock).
The hallucination that customers want AI in everything is counterproductive when survey after survey says the opposite. The hallucination that AI stuff will dominate the future is when the generation that got us there will point and laugh at the predicament of AI.
If these hallucinations fade away from the fevered minds of Silicon Valley and Wall Street, the great AI vibe shift of 2026 will be complete.
This article reflects the opinion of the author.
Disclosure: Mashable’s parent company Ziff Davis filed a lawsuit against OpenAI in April 2025, alleging it infringed Ziff Davis copyrights in the training and operation of its AI systems.
<a href