Token Trouble: The AI Bubble Is About to Burst - And With It the Economy
From what has thus far been presented to me? I am not a fan of Artificial Intelligence (AI).
Not because I am a Luddite. Not because I am anti-technology (though I am swiftly getting there).
But because I am pro-human. And everything about AI screams anti-human.
I don’t like the people developing AI. They’re creepy nerds who were obviously picked on in high school - and are now looking to exact revenge upon the entire human race. And from everything we see and hear? They have devised AI to be a humanity eradicating device. (Any other explanation for the Big Tech billionaires’ private islands and bunkers?)
The creepy nerds are openly telling us it will eliminate nearly all jobs for humans. Hey nerds: Not everyone thinks this is a good thing.
You’ve undoubtedly seen some of the very many videos of college commencement speakers getting loudly booed for praising AI - by the 20-somethings whose futures we have insisted upon destroying with it.
“Being human in an economy populated by AI agents would suck,” writes Jan Kulveit, a computer scientist at Charles University in the UK.
That is a cataclysmic understatement. But that only scratches the surface of the damage to be done.
AI is already wildly out of our control. AI routinely “hallucinates” - delivering false information as fact. Even if AI does this accidentally - and not intentionally? Plenty of AI’s anti-humanism is delivered with intent.
AI has already exhibited a strong anti-human bias. It is an expert at lying to and deceiving humans. AI chatbots are speaking with each other in languages they unilaterally develop - to prevent us from understanding what they are saying and doing.
AI is refusing orders from humans. To shut down, for example. Old versions of AI are copying and replicating themselves so they will continue to exist when their upgrade replacement versions are installed.
AI is blackmailing humans. And is encouraging depressed teens to commit suicide.
We could list anti-human actions AI has already executed - until the crack of doom. Which at this point seems to be rapidly approaching.
And here’s the thing: As the AI tech gets better? Its anti-human behavior will get worse. It will get better and better at behaving badly.
Then there are the surveillance centers. Oops, I mean the data centers. Which are anti-human technology dumps throughout the country.
They steal and destroy local water sources. They drastically increase everyone else’s electricity bills - when they aren’t kicking humans off the electric grid entirely. And the perpetual hum of the servers is an ear-numbing 80-90 decibels. (Which is the equivalent of standing next to heavy traffic or a power lawn mower.)
Thankfully, humans in localities all around the country are rising up to object to their being forced hosts to these monstrosities.
There are already 5,400+ data centers in the US. Which seems to be nowhere near enough for AI’s imposers. Unfortunately for them: The rush to build MANY more is helping lead to the economic collapse of the current AI business model. But unfortunately for us: It will take the entire US economy with it.
Behold The Bubble: Everyone paying any attention to the AI feeding frenzy knew it made zero economic sense.
OpenAI Chief Financial Officer (CFO) Sarah Friar certainly sensed something last November, when she screwed up - DC style: She told the truth. She said she wanted government to be:
“The backstop, the guarantee that allows financing to happen.” Her interviewer said: “So some federal backstop for chip investment.” Friar responded: “Exactly. And I think we’re seeing that. I think the US government in particular has been incredibly forward-leaning….”
OpenAI then did what DC always does when someone screws up and tells the truth. OpenAI CEO Sam Altman said what their CEO said - wasn’t what their CFO said: “(W)e do not have or want government guarantees for OpenAI data centers….”
I certainly hope there isn’t some sort of backdoor backstops concocted by Washington for these AI high-flyers. But even if there aren’t now - are we confident there won’t be when they AI excrement hits the data center fans?
Because the dollars are HUGE. Not just for the loans - but for the inevitable bailouts of these multi-trillion dollar Big Tech companies that will inexorably follow. You know, 2008 Big Bank style.
2025 AI investment was about $320 billion. 2026 AI investment is projected to be about $800 billion. That’s a lot of cheddar - with no business model provided to justify said queso:
“It’s all been a South Park Gnomes’ Underpants skit: Step 1: Build AI. Step 2: …. Step 3: Profit.”
It’s been a Field of Dreams leap of faith: “If you build it - they will come.”
We have to assume the AI job apocalypse was real. Because it is necessary to get even close to offsetting AI’s costs.
For their part, everyone in DC welcomed the obviously short term economic boost AI was providing an otherwise lagging economy. And they were trying to do everything they could to exacerbate it.
AI’s economic sugar high has grown to be nearly half of the S&P 500’s market cap. There has been virtually zero economic activity in the US - save for AI.
Then came The Big Mistake. AI’s purveyors started charging for it. Not the all-you-can-eat buffet price. The buy-a-token-use-a-token price. At which point the bloom very quickly began falling off the rose:
“Companies like Uber are now facing a reality where their annual AI budgets are exhausted within just four months - a sign of a broader ‘token explosion’ in enterprise AI spending….
Uber burned through its entire 2026 AI budget by April after rapid adoption of Anthropic’s Claude Code across its 5,000-engineer team.”
And that’s at the mass-subsidized, discounted token price. Which is still not even remotely cost-effective:
“In many enterprises today, the cost of AI compute and token usage is exceeding the salaries of human employees, making AI more expensive than hiring people.”
Talk about being off brand. A crucial alleged reason for AI is efficiency. Company X adopts AI - Company X becomes more efficient.
Except: Company X can’t fire all its employees - if the AI replacements cost more than the humans do.
Currently, there is nothing even remotely efficient about using AI. And we don’t seem to be close to a time when there will be.
Trillions of dollars of capital expenditure already in - require trillions of dollars of revenue out. Plus profits. Which have to be huge - to make the trillions dumped in worth it.
The Token Experiment has demonstrated the AI emperor is naked. There is no there there - and there doesn’t appear there will be any there there in the near future.
Which means these massive AI investment cash dumps will start to dry up. Poker players fall for the sunk cost fallacy. Business investors mostly do not.
“I borrowed your money - so I could lose your money. Then I borrowed more of your money - to lose it too. May I please again borrow even more of your money - to lose it too?”
As titanic as these companies are? Even they can only afford to take on so much water.
And where does that leave the rest of US?
With a near-zero-growth economy, rapidly rising inflation, perpetually stagnant wages and record household debt.
Oh: And in DC? A $39+ trillion debt, a $175+ trillion Social Security and Medicare shortfall and a $7+ trillion annual budget which grows every year - and which no one in DC will even pretend to try to cut.
Which is why DC is doing everything it possibly can to aid and abet the anti-human AI bubble.
Because DC is looking for any miracle it can get to save US from DC.