The credit intelligence platform leading firms trust

When credit markets move, seconds matter. 9fin delivers breaking alerts 30–60 minutes ahead of the competition, giving you time to act, not just react.

From CLO impact to covenant risks, 9fin offers real-time restructuring intel, private credit data, and deep distressed coverage; all backed by 20 years of historical data.

AI-powered tools compress hours of research into minutes:

• Comparable analysis

• Earnings breakdowns

• Credit screening

Thousands of leading firms rely on 9fin to evaluate risk and seize opportunities before the rest of the market.

We’re offering subscribers to The Wall Street Skinny a 30 day free trial. Start your free trial today!

With everyone asking whether the AI boom is a bubble, all eyes inevitably turn to the people who called past bubbles correctly.
Enter Michael Burry. 

Burry was one of the first major investors to short subprime in size leading up to the 2008 crisis. His story ran alongside Steve Eisman’s in The Big Short (check out our interview with Eisman from earlier this year!).

Now Burry is loudly criticizing big tech companies for what looks like a small, boring accounting assumption buried in their financial statements: the “useful life” of their GPUs.

For anyone who doesn’t speak “big tech,” GPUs are the chips that power and train AI models, the backbone of the AI products these companies sell. It sounds trivial, but that one assumption translates into billions of dollars in earnings every year.

Before we get to whether this is the mega-scandal Burry claims, let’s talk about why this is so exciting…it gives us a reason to explain depreciation and (hopefully) HAVE PEOPLE CARE 💃 So, two quick explainers:

1. How does useful life impact earnings? Hint: depreciation

2. Why does depreciation — a non-cash expense — play any role in valuation?

Depreciation: An Overview

When a company makes a big investment, like buying GPUs to train AI models in order to sell products, they need to adhere to rules put forth by “Generally Accepted Accounting Principals" (US “GAAP”). Outside the US, companies must adhere to International Financial Reporting Standards (“IFRS”), which are broadly the same with minor differences.

These rules mean investors are presented with clean, standardized financials in filings so they can make sense of companies’ performance. Contrast that to the dumpster fire that is private companies, whose internal financials are typically a mess.

Cool. So, what do the rules say about how a company must account for huge investments like buying these GPUs? Pretend the company spends $100 million on chips and Management believes these chips will be obsolete after 3 years. After that, these chips won’t be able to be used to support revenue. So the company will then need to buy new chips to power their models.

US GAAP and IFRS require the company recognize that $100mm purchase using something called the “matching principle”, which says you “expense” the asset in the periods you’re actually using it to generate revenue.

Using our imaginary company as an example — let’s call them Meta — here’s what happens:

Year 1 (purchase year):

• Cash Flow Statement: $100 million cash outflow shows up as “CapEx”.

• Balance Sheet: PP&E, an asset, increases by $100 million.

• Income Statement: No impact yet if the chips aren’t being used to generate revenue.

Years 2–4 (chips in use):
Companies usually use straight-line depreciation, so $100mm / 3 years = ~$33 million per year.

Here’s what shows up each year:

• Income Statement: A $33mm depreciation expense which reduces earnings.

• Balance Sheet: PP&E decreases by $33mm from $100 to $67 each year as the chips get utilized.

• Cash Flow Statement: Depreciation is added back, because the cash left the building in Year 1, not Years 2–4. 

How Useful Life Impacts Earnings

This is the heart of Burry’s criticism: increasing the useful life inflates earnings.

If management changes the useful life from three years to five, annual depreciation becomes $100mm / 5 = $20mm. Same GPUs. Same $100mm cash out the door. But depreciation drops by 40%.

The impact on earnings is slightly more complicated because of taxes (we’ll get to that), but the direction is obvious: reported earnings jump simply because management increased the useful life.

A three-year assumption means “these chips won’t produce revenue after Year 3.” A five-year assumption means “they’ll keep generating revenue through Year 5.” That judgment call has real earnings consequences.

Taxes: US GAAP vs. IRS

Now, if you look at this and think wait, a higher useful life means more depreciation now and therefore lower taxes because of a higher tax shield…should’t Meta want that? The answer is: not exactly. Accounting or “book” depreciation has nothing to do with tax depreciation, meaning actual deductions the IRS allows and therefore actual cash flows. They are completely separate systems.

Under the 2025 “One Big Beautiful Bill,” companies can deduct 100% of qualifying equipment in the first year via bonus depreciation. That means for tax purposes, companies typically deduct the entire $100mm upfront, lowering taxable income immediately. Accounting useful life has zero impact on tax books.

For accounting, however, they still spread that same $100mm over the useful life.

These two systems are not required to match – and they don’t. This mismatch creates what accountants politely call a Deferred Tax Liability, which is a topic for another day.

Back to Earnings

So: changing useful life increases reported earnings but does not affect cash taxes. Which means it doesn’t affect cash flow.

That distinction matters a lot when we talk about valuation.

Why Does This Matter for Valuation?

There are two primary valuation methodologies:

1. Comps

2. DCF (Discounted Cash Flow)

Method 1: Comps

When you run a comps analysis, you must choose a multiple and earnings metric. There are lots! Sales, EBIT, EBITDA, Net Income, EPS…the list goes on. The multiple — and therefore the metric you use — is based on the industry norm. Big tech companies typically trade off a PE ratio — price to earnings (EPS) — so EPS matters a great deal. Stretching useful life boosts EPS, thereby inflating valuation. How?

Stock price = PE x EPS

If EPS is higher, all else being equal, your stock price is higher 

BUT this shouldn’t be used in a vacuum. 

Many investors also look at secondary multiples and metrics, and arguably the second most commonly used is: EBITDA (Earnings Before Interest, Taxes, Depreciation & Amortization) which is a proxy for operating cash flow. Depreciation has no impact on this number, as it is excluded in this metric. Therefore, a tweak in the useful life has zero impact on earnings or valuation.

The way this metric makes its way to the stock price is a tad more complicated: 

Enterprise Value = Multiple x EBITDA 
Equity value = Enterprise Value - Debt - Preferred - NCI + Cash
Stock price = Equity value / diluted shares

None of these are impacted by depreciation. Therefore, useful life has no impact on the valuation or stock price.

Method 2: Discounted Cash Flow (“DCF”)

Additionally, the gold standard methodology is the DCF analysis, which projects cash flows into the future and discounts them back to today. Depreciation only matters indirectly in two ways:

First as a tool for estimating cash taxes – but we just established that accounting depreciation has nothing to do with tax depreciation.

Second, as it indirectly relates to the assumption an investor chooses for projected CapEx. Now, this point actually matters. If these chips burn out in three years instead of five, the company has to reinvest sooner, which means more cash out the door, faster. But the accounting useful-life number Meta chooses is arguably irrelevant. What ultimately matters is the real replacement cycle and the business’s future investment needs.

You can’t look at the company’s accounting choice in isolation. Investors need to do their own work and decide what they believe about long-term capex requirements, not blindly accept the depreciation timeline in the 10-K.

So…is Burry right?!!?!? Should we assume the company needs to upgrade chips after 3 years instead of 5?

As non-tech experts, we can’t pretend to know the exact economic lifespan of a GPU. That’s the whole point: investors have to decide for themselves what they think is reasonable. Maybe you side with Burry and think Meta (and the rest of Big Tech) are overstating earnings and should be trading lower. Or maybe you think Burry is full of sh*t and the current valuations make perfect sense.

All I can offer is a real-world analogy from The Wall Street Skinny which has exactly one meaningful piece of PP&E: our computers. My former employer assumed a three-year useful life for laptops. Meanwhile, I’m still typing this on the same machine I bought 3 years ago and going strong. We hope to squeeze at least another two years out of them. When they die, we replace them. At the outset of buying the computer I can’t even predict when exactly my computer will completely crash and I’ll have to buy a new one.

Whether the same logic applies to GPUs powering massive AI workloads is up for debate. If Burry is right and today’s GAAP earnings are overstated, tomorrow’s CapEx may be understated. Regardless, this is the beauty of markets.

Investors should be looking at more than just EPS when valuing these companies – and this is why understanding accounting matters. As Warren Buffett likes to remind people, accounting is the language of business: imperfect, sometimes misleading, but still essential if you want to understand what’s going on.

And honestly? It’s fascinating to watch accounting in the headlines, especially on the heels of the recent wave of data-center financing structures designed specifically to keep debt off the balance sheet. The world’s biggest companies are ramping up CapEx at an insane pace to keep up with AI demand, and some of Big Tech’s accounting treatments have gotten…creative.

Whether using a five-year useful life instead of three is truly “creative,” or simply a rational assumption based on the hardware’s real economic life, is still up for debate.