Amazon waved a wand and created profits from nothing. Between 2020 and 2024, it reported cumulative pre-tax earnings of $155 billion. But it should have reported $95 billion.
Why?
A short accounting lesson.
Companies buy equipment and build facilities to generate profits. All the money spent to buy or build these assets ends up on a company's balance sheet as property, plant, and equipment. Then companies must depreciate the equipment, buildings, and so on, every year down to some estimated worth at the end of each item's useful life.
Amazon buys a lot of servers, routers, cables, and other network equipment. It spends billions buying new equipment and replacing old ones. Since the company's inception, it had always assumed that the useful life for this type of equipment was three years.
Historically, Amazon's annual depreciation expense amounted to roughly 20% (and higher) of the value of its property and equipment.
Then the Artificial Intelligence race began. In the last five years, Amazon bought more servers and network equipment than it had in its past. More equipment means more depreciation. With commensurate profits from AI expected out in the future, reporting higher depreciation without the additional income deteriorates profit margins. Deteriorating margins sink stock prices.
Management can't have that.
So, to offset the increased spending for equipment, Amazon used an old trick to decrease depreciation expenses: it extended the useful life of its servers and network equipment. Remember, from inception Amazon had always assumed this equipment would last three years. Then they increased the useful life from three years to four years in 2020. Then five years in 2022. Then six years in 2024.
When a company assumes a piece of equipment will last longer, it lowers the annual depreciation expense. If you spend $50,000 for a new car, and it lasts 10 years, then each year the car will depreciate $5,000 assuming you sell it for nothing at the end of year 10. But if that car lasts 15 years, then each year the car will depreciate just over $3,000, a lower depreciation rate than the car that lasts 10 years.
This is what Amazon did. By extending the useful life of its servers and network equipment, Amazon reduced its annual depreciation expense to 14% of the value of its property and equipment, a 6% decrease from the 20% it reported in prior years. And from 2020 to 2024, this 6% difference saved it billions, somewhere around $60 billion. This allowed Amazon to report $155 billion in cumulative pre-tax earnings, not the $95 billion it would have with its previous accounting assumptions.
How did the market reward Amazon for its financial wizardry? It boosted its stock price to all time highs, which has enriched management and shareholders alike.
Amazon isn't the only one benefiting from questionable changes to useful life assumptions: Meta saved $16 billion. Alphabet $18 billion. Microsoft (a whopping!) $86 billion. Combined, these four companies saved nearly $200 billion in expenses, propping up apparent profits and stock prices. At some point, these companies will have no choice but to revise their useful life assumptions, likely downward.
(Amazon has started. It reduced useful life for servers from six years back to five years. If you're worried about profits taking a hit, don't worry. Amazon extend the useful life of its heavy equipment now from ten years to thirteen years. Profits preserved!)
Now, what happens when the servers, routers, and cables do not last as long as assumed? Management will have to write down the value of property and equipment. After all, a server that stops working is basically worth nothing, even if your accountant argues otherwise.
When management writes down the value of the equipment, profits will decrease. When profits decrease, so will stock prices. With these four companies accounting for 20% of the S&P 500, what happens to the index?
For now, I gladly sit on the sidelines.