Remember when everyone was freaking out about “circularity” in AI deals last month? Well, buckle up buttercup, because there’s a new villain in town: depreciation. And it’s got the AI trade running scared like it just saw its browser history.
Here’s the deal: All those shiny, expensive GPUs and semiconductor chips that Big Tech has been hoarding like digital dragons? Some very smart (and very pessimistic) people think they’re going to lose value way faster than anyone expected. Think of it like buying the latest iPhone, except it costs billions and might be obsolete in two years instead of six.
The numbers are getting ugly. The Nasdaq 100 has dropped 6.3% recently, while tech stocks have fallen over 9%. It’s like watching your crypto portfolio in 2022, but with more corporate jargon.
The Short Sellers Are Having a Field Day
Michael Burry (yes, the guy who predicted the housing crash and got a movie made about him) is basically doing victory laps on social media. He’s claiming that Big Tech companies will understate depreciation by a whopping $176 billion between 2026-2028. His math? These fancy AI chips will last 2-3 years, not the 6 years companies are banking on.
It’s like planning your budget assuming your car will last 20 years, when it’s actually going to die after 7. Except instead of a car, it’s billions of dollars worth of computer chips.
The Math That’s Making Everyone Nervous
Peter Berezin from BCA Research dropped some sobering numbers: By the end of this decade, hyperscalers could be sitting on $2.5 trillion in AI assets. If those depreciate at 20% annually, that’s $500 billion in yearly depreciation expenses. To put that in perspective, that’s more than these companies’ combined profits for 2025.
Kai Wu from Sparkline Capital painted an even grimmer picture, suggesting depreciation could jump from $150 billion to $400 billion annually in just five years. He compared current AI spending to historical booms and found that when you adjust for faster depreciation, today’s AI buildout is off the charts – even more intense than the railroad boom of the 1800s.
But Wait, There’s Pushback
Not everyone’s buying into the doom and gloom. Bernstein analyst Stacy Rasgon thinks GPUs can actually run profitably for about 6 years, and that most companies’ depreciation accounting is reasonable. It’s like having that one friend who insists their 10-year-old laptop still works fine while everyone else has moved on to newer models.
The Bottom Line
Whether you’re team “AI is the future” or team “this is all hype,” the depreciation debate highlights a fundamental question: How long will today’s AI infrastructure actually stay relevant? In a field where yesterday’s breakthrough becomes tomorrow’s paperweight, that’s not just an accounting question – it’s an existential one.
The AI trade has been the market’s golden child, but even golden children have to deal with reality eventually. And reality, as it turns out, might involve a lot more depreciation than anyone bargained for.
The moral of the story? In tech, the only constant is change – and apparently, the only certainty is that your expensive equipment will be worth less tomorrow than it is today.