The AI... Bubble?

“We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.” – Roy Amara, Stanford computer scientist.

Dotcom, housing, crypto, and now, AI.

The Fastest Scaling in History

An asset bubble is a period where the price of an asset skyrockets beyond its intrinsic value before violently crashing. They have plagued economies for centuries.

The artificial intelligence industry has experienced what is considered the fastest technological scaling in history. Exploding from an emerging tech niche into a multi-trillion-dollar foundational economic engine in roughly a decade. The global AI market is now set to rise from $189 billion in 2023 to $4.8 trillion by 2033 

On the one hand, it is still early days but FOMO is an important factor in its hype. The race to ‘’win the AI race’’ has driven a lot of investment into the sector. Banks are flooding capital into it encouraged by profit reports from the likes of Boston Consulting Group’s Nov. 2025 report that forecasts $370 billion, according to Forbes.

In 2025, Alphabet, Amazon, Meta, and Microsoft alone reportedly planned to spend a combined $320 billion on AI technologies and infrastructure. Wall Street analysts now estimate total AI capital expenditure could climb above $1 trillion by 2027.

Banks at the Limit

Here is where the picture gets complicated.

According to the Financial Times, banks are now stretched close to their limits. JPMorgan Chase, Morgan Stanley, and SMBC have lent unprecedented sums to cash-in on the race to build AI infrastructure, and they are now actively exploring ways to offload the risk.

In the last 6 months, JPMorgan and MUFG have been trying to offload nearly $38 billion of risky debt related to building data centres. Money that was lent to Oracle for infrastructure in Wisconsin and Texas. Banks are using synthetic risk transfers (SRTs) to offload this debt. 

SRTs are used “to offload potential losses in their loan portfolios to non-bank investors while retaining the loans on their balance sheets”, whilst some banks have even turned to selling directly to non-bank lenders – i.e. those without a banking licence or deposits, funding themselves instead through wholesale markets, funds or private capital – in private deals.

The question, then, is as simple as it is consequential: what happens if the lending stops? The AI sector requires enormous and continuous capital flows to sustain its growth. If banks pull back, will the entire infrastructure build-out stall?

According to research from Bain & Company “Two trillion dollars in annual revenue is what’s needed to fund computing power needed to meet anticipated AI demand by 2030. However, even with AI-related savings, the world is still $800 billion short to keep pace with demand’’

The Generation That Isn't Buying It

While boardrooms celebrate AI's potential, something different is happening in university auditoriums across the United States.

At commencement ceremonies in 2026, graduation speakers who praised artificial intelligence were met with boos. 

When former Google CEO Eric Schmidt praised AI at the University of Arizona, he was booed and jeered multiple times, notwithstanding his acknowledgement of fears that AI threatens to deprive people now entering the workforce of a future. When Gloria Caulfield, a real estate executive, called AI “the next Industrial Revolution”, she was continuously booed and heckled during a graduation ceremony at the University of Florida. Of AI, students have said they “don’t want AI to trump our academic, maybe our political, maybe our cognitive processes”. Many universities have established anti-AI societies on campus.

According to a Harvard Youth Poll, by more than a 3:1 margin young Americans believe AI will take away opportunities. The Class of 2026 is entering one of the toughest entry-level job markets in years. For many graduates AI does not feel like liberation. It feels like competition. The backlash at graduation ceremonies may be one of the clearest cultural warning signs yet about how younger generations view the technology that their employers are betting everything on. And interestingly, according to Harvard’s poll, this scepticism is one of the few issues that cuts across education levels and gender. Indicating this is not solely the preoccupation of those in university on path to jobs in finance, technology or law, but a generational worry across the field.

Remember Michael Burry?

We have all probably seen The Big Short. The 2015 film follows a congeries of contrarian investors who predicted the collapse of the US housing bubble in the mid-2000s, betting against a reckless banking system and winning billions while the global economy crumbled.

One of the central figures in that story was Michael Burry. He was an eccentric, data-obsessed hedge fund manager who sounded the alarm about the collapse far before anyone else was listening.

He has recently placed a $1.1 billion bet to short the two of the most hyped AI companies in the world. He has bet that Nvidia and Palantir are headed for a fall. To be clear, he is not betting on the downfall of AI, simply that the stocks are being overestimated compared to their true value.

In 2025 he posted on X “Sometimes, we see bubbles. Sometimes, there is something to do about it. Sometimes, the only winning move is not to play”. And, as of June 2026, it seems to be paying off: Palantir is down nearly 40% from its all-time high, but the company's growth rate hasn't slowed.

What Happens If It Bursts?

History offers an interesting parallel. At the height of the industrial revolution, spending on American railroads eventually peaked at around 20% of US GDP. AI spending is already several percent of US GDP and rising. As Forbes Australia noted, AI expenditure is currently one of the few forces preventing Trump’s tariffs  and wider economic turbulence from tipping the United States into recession.

That dependence is mutual. If the bubble were to pop, the aftermath may not be contained to Silicon Valley. AI is a recession buffer today; but could be its downfall tomorrow. If jobs are lost and mortgages unpaid then banks holding already risky AI debts will face extraordinary pressure.

The AI boom will likely be the same as previous technologies, it will transfer vast amounts of wealth from the broad mass of investors into the hands of the small number of companies that ultimately win the race. Most of the capital being deployed today will not generate the returns its holders expect.

Every few years, markets fall in love with a story.

In the 1990s it was the internet.

In the 2000s it was housing.

In the 2010s it was crypto.

Now, it’s AI.

Each of those stories contained genuine truth. The internet did transform the world. Housing does hold real value. Blockchain technology did introduce real innovations. And artificial intelligence is, without question, a genuinely transformative technology.

But value and overestimation are not mutually exclusive. The dotcom crash wiped out huge amounts of market value, and the internet still changed everything. The question is always whether the prices being paid reflect reality.

Roy Amara's quote cuts in both directions. We overestimate the short run. We underestimate the long run. The danger is not that AI fails to deliver on its promise. The danger is that it delivers, just not when promised. Michael Burry is not predicting the death of artificial intelligence. He is asking a simpler, older question: when will the price be right?

Image: Flickr/The White House (Andrea Hanks) Licence: public domain.

No image changes made.