Hidden tricks, extended depreciation, and debt growing faster than chip performance
The AI boom is enormous. A race for the biggest data centers, the fastest chips, and the most powerful models. And just like in every bubble, the first cracks appear early — accounting creativity, artificially extended asset lives, hidden debt, and financial structures that look far more like 2008 than 2025.
To truly understand this industry, we need to look beneath the surface — into the numbers behind the technology, and even more importantly, behind the accounting.
The lifecycle of GPUs and servers is brutally short — usually 2 to 3 years, often less. AI evolves far faster than data centers can deploy hardware.
Yet major tech companies have done something that accounting standards would at best call aggressive creativity:
they extended the useful life of these assets two to threefold.
The effect is simple:
Longer useful life → lower annual depreciation → higher reported earnings.
Michael Burry summarizes it sharply:
“Extending asset life is the most common accounting trick of the modern era.”
According to his estimates, hyperscalers (Meta, Google, Amazon, Microsoft, Oracle) will collectively understate depreciation by $176 billion between 2026 and 2028.
In practice this means (Burry’s model):
Oracle may overstate earnings by 26.9%
Meta by 20.8%
The rest are not much better.
When depreciation doesn’t reflect reality, revenues look impressive —
but the cash flow behind them tells a different story.
Here’s where things get more interesting.
Companies pre-pay for chips that will be delivered months or even years later. But these are not just ordinary preorders.
Undelivered hardware is being:
used as collateral,
recognized as an asset,
leveraged to raise new debt,
and that new debt is used to place even more chip orders.
A closed loop.
Demand that isn’t organic, but financially manufactured, designed to convince investors and lenders that the AI boom is accelerating.
What AI giants are doing today looks increasingly similar to the pre-crisis years.
The only difference?
Instead of mortgages, they’re packaging GPUs, servers, and data centers.
What’s happening:
Meta moved ~$30B of debt off balance sheet via SPVs.
Oracle’s leverage is exploding and the market sees it — its CDS just hit the highest level in two years.
Musk’s xAI is raising $20B via an SPV where its only obligation is a five-year lease of Nvidia chips.
Google is backstopping third-party data center debt using credit derivatives.
Off-balance-sheet financing has become the new normal in the AI sector.
Banks and private credit know there’s money to be made.
AI companies know it’s a convenient way to hide leverage.
Morgan Stanley estimates AI firms will need $800B in private, off-balance-sheet financing by 2028.
UBS warns AI-related debt is increasing at $100B per quarter.
This isn’t organic growth.
This is a debt addiction.
To make things even more concerning, leaked Nvidia internal emails point to a deeper issue:
Nvidia’s hardware business is booming —
but its software division is struggling.
Key takeaways:
Clients don’t understand what Nvidia’s software actually is.
Legal and procurement teams of large enterprises cannot approve it.
Nvidia struggles to explain the value of its subscription model.
There is a “fundamental disconnect” between the product and the market.
Translated:
Hardware sells itself.
Software does not.
And that’s a problem — because Nvidia needs software revenue to protect margins once the GPU boom slows down.
Michael Burry summarizes the situation bluntly:
depreciation is manipulated,
debt is growing faster than revenues,
reported results are overstated,
and the entire industry runs on “flipped money” —
debt → chips → more debt.
And yes, Burry is short.
Exactly on the companies we’ve been talking about.
AI is a massive opportunity.
But also one of the biggest emerging risks — and at this stage, for us, the risk is starting to outweigh the potential upside.
The issue is not the technology — that part is incredible.
The issue is how the boom is financed.
We’re witnessing clear signs of:
excessive leverage,
accounting creativity,
hidden obligations,
and “growth” driven not by real demand, but by debt recycling.
Eventually, the numbers will have to reflect reality.
When that moment comes, the market will separate companies that generate real cash flow from those that survive only by rolling debt into more debt.
Note: This article is for informational purposes only and does not constitute investment advice. Investing in financial markets involves risks, and independent analysis is essential before making investment decisions.