The AI Bubble Blowback? Why Big Tech’s Accounting Tricks Have Wall Street Worried
A Warning That’s Getting Louder
Is the global AI boom unstoppable, or is it hiding cracks that could shatter trillions?
Two heavyweight investors say the second option is far more likely.
Michael Burry, the investor who accurately predicted the 2008 housing collapse, and Jim Morrow, the founder of Callodine Capital, are both sounding the alarm. Their message is blunt: Big Tech’s AI-era profits may look dazzling, but the numbers behind them are being propped up by accounting tactics that delay the real cost of the AI gold rush.
According to them, the world’s biggest companies aren’t growing as fast as they appear. They’re stretching depreciation schedules on servers and AI chips to inflate earnings—masking what they claim is one of the most crowded, dangerous trades the market has ever seen.
And if they’re right, the fallout could be enormous.
Why These Warnings Matter Now
Burry Steps Back, Then Speaks Out
Burry recently deregistered his firm Scion Asset Management, stepping away from managing outside money. Without public reporting obligations, he returned to social media with force.
In one post, Burry claimed that Big Tech will understate depreciation by a massive $176 billion between 2026 and 2028. He pointed specifically at Meta and Oracle, estimating that inflated profits could be as high as:
- 26.9 percent at Oracle
- 20.8 percent at Meta
That’s not a small rounding error. Those numbers can move markets.
Morrow Calls It the Most Crowded Trade Ever
Jim Morrow echoes Burry’s warnings and goes a step further. He argues the AI boom is built on an illusion—companies have quietly extended the useful life of their AI hardware to make their bottom lines look stronger than they truly are.
He calls it a tsunami of depreciation waiting to hit earnings. According to Morrow, these adjustments aren’t accidental. They’re strategic.
He sums it up like this:
“It’ll happen slowly, and then all at once.”
The Core Problem: Depreciation Sleight of Hand
What Depreciation Is Supposed to Do
When a company buys expensive hardware—like AI chips, servers, and cooling systems—it records the cost over the period the equipment is expected to remain useful. This reduces profits each year, and it’s meant to reflect reality: hardware gets old and loses value.
What’s Happening Instead
Burry and Morrow argue that Big Tech is stretching the timelines. Hardware that once depreciated over three or four years is now being listed with useful lives of five or six years.
Meta, for example, moved from four to five years, then announced in 2025 it would extend that lifespan to 5.5 years for servers and network equipment.
By stretching these timelines, companies reduce reported expenses now and push them years down the road. That makes today’s profits look bigger—even if the equipment is becoming obsolete faster than ever.
But Chip Cycles Are Getting Shorter, Not Longer
Nvidia, the company at the center of the AI hardware race, now releases major chip upgrades roughly once a year. Three-year technology is ancient by current standards.
This creates a stunning mismatch:
- Chips become obsolete within 12 to 24 months
- Big Tech is reporting that they will remain valuable for five to six years
Morrow argues the math simply doesn’t add up.
The Numbers Behind the Fear
The Economist Weighs In
The Economist recently called this mismatch “the $4 trillion accounting puzzle at the heart of the AI cloud.” They estimated:
- A three-year depreciation schedule would cut Big Tech profits by around $26 billion annually
- A two-year depreciation schedule would double that hit
- If companies fully matched the pace of Nvidia’s chip cycle, as much as $4 trillion in market value could evaporate
Other Analysts Are Speaking Up Too
Richard Jarc from Uncovered Alpha agrees that current accounting ignores the real-world speed of AI hardware obsolescence.
He argues that the market is being propped up by the subsidization of compute costs, not sustainable demand.
But Not Everyone Agrees
Some Say the AI Boom Is Still Just Beginning
Bank of America’s semiconductor team pushed back on the gloom. They argue the recent tech selloff is due to temporary macro jitters, not a collapsing AI thesis.
They point out:
- Memory and optical sectors rose 14 percent last week
- Nvidia disclosed more than $500 billion in upcoming data center orders
- Demand for compute infrastructure remains robust worldwide
In other words, they believe this is just a correction—not the end of the story.
The Bigger Picture: Are We Mistaking Spending for Growth?
The AI Boom Is One of the Most Capital-Heavy Bets of Our Time
Morrow warns that investors are confusing massive spending with genuine economic progress.
To build a single hyperscale data center capable of supporting advanced AI models, companies may spend around $50 billion. Multiply that by dozens of facilities, and you get an unprecedented wave of capital expenditure.
Yet many of these data centers can’t even run yet due to power delays.
Some facilities in key U.S. markets are sitting idle waiting for grid access.
Morrow’s blunt assessment:
“Every month a $35 billion stack of GPUs sits without power, that’s a billion dollars of depreciation burning a hole in the balance sheet.”
The Looming Power Crunch
Companies Are Buying Their Own Power Plants
Because of long wait times for grid hookups, tech giants are considering or already building their own energy sources—turbines, substations, and in some cases, private micro-grids.
Morrow believes this will create a power bubble:
- Overbuilt infrastructure
- Over-leveraged utilities
- And eventually, ratepayers footing the bill
He compares it to past capital booms that ended badly:
shale oil, fiber optics, railroads.
A Market Concentrated Like Never Before
Nearly Half of Retirement Money Is in Six Stocks
Morrow argues that market risk is off the charts. With an enormous amount of 401(k) and passive investing money tied to a handful of megacap tech stocks, any abrupt correction could ripple through the entire financial system.
“This is the most crowded trade in history,” he warns.
“When it turns, it’s going to turn fast.”
Conclusion: Boom, Bubble, or Something In Between?
The debate now raging across Wall Street isn’t just about AI—it’s about whether investors and companies are being honest about the costs, timelines, and true economics behind the largest tech buildout in modern history.
Is AI the next industrial revolution, or the next dot-com bubble dressed in cutting-edge hardware and clever accounting?
Burry and Morrow argue that the numbers don’t lie—and that when the depreciation wave hits, reality will come crashing in.
Others say the AI future is brighter than ever and today’s concerns are temporary noise.
For now, the only certainty is that the stakes have never been higher, and every investor—from hedge funds to everyday savers—is riding the same rollercoaster.
