MW Google, Meta and Amazon are facing an expensive AI risk that nobody is paying attention to

By Christine Ji

As Big Tech companies ramp up their capital expenditures, rising depreciation costs are threatening their margins, and Wall Street seems to be vastly underestimating the impact

Alphabet, Amazon and Meta are spending billions on networking and server equipment for AI workloads.

Big Tech's artificial-intelligence spending frenzy has a hidden cost that many on Wall Street are overlooking.

Alphabet Inc. $(GOOGL)$ $(GOOG)$, Amazon.com Inc. (AMZN) and Meta Platforms Inc. (META) collectively are expected to spend $274 billion on capital expenditures by year's end as they race to build out AI infrastructure. However, the bill for these hyperscalers could come due sooner than expected - as early as next year - as escalating depreciation costs eat into profit margins, according to Bank of America analyst Justin Post.

When companies spend money on long-term assets such as data centers or graphics processing units, they don't count the full capex cost as an expense upfront. Instead, they recognize incremental depreciation and amortization expenses over a specific timeframe determined by the useful life of the assets. If depreciation and amortization expenses accelerate faster than the pace of AI revenue generation, or if the useful life of AI assets is revised down, then these companies' high capex levels could pose a risk to stock prices.

Post believes consensus estimates haven't taken account of the full scale of the AI capex cycle, writing in a note Wednesday that "the Street is still playing catch-up on growing depreciation expense." Underestimating these costs makes these Big Tech players look more profitable than they actually are, potentially fueling an AI bubble.

The chart below displays Bank of America's depreciation and amortization projections compared against Wall Street consensus estimates. Bank of America's forecasts are higher than consensus for nearly every measure, with the exception of Amazon's 2025 estimate.

   Company                   Ticker           2025E          2026E          2027E 
   Alphabet Inc. Class A     GOOGL    $21.1 billion  $31.0 billion  $43.5 billion 
     Street  $20.9 billion  $27.9 billion  $36.5 billion 
   Meta Platforms Inc.        META    $18.4 billion  $28.3 billion  $40.7 billion 
     Street  $18.1 billion  $28.1 billion  $37.2 billion 
   Amazon.com Inc.            AMZN    $63.4 billion  $78.4 billion  $95.9 billion 
     Street  $63.5 billion  $76.3 billion  $88.6 billion 
                                                          Source: Bank of America 

Bank of America recently lifted its projections for Meta's 2026 and 2027 depreciation and amortization expense. Depreciation and amortization expenses "will accelerate in 2026 as recently deployed assets come online, adding incremental pressure to operating expenses," Post wrote of Meta, Google and Amazon.

Also read: This is the critical detail that could unravel the AI trade: Nobody is paying for it.

The disparity could only compound going forward: In 2027, Alphabet is slated to have a $7 billion gap between the Bank of America and Street estimates. The 2027 gap is projected to be $5.9 billion and $3.5 billion, respectively, for Amazon and Meta.

The highest amount of capex spending is going toward purchasing technical infrastructure, including Nvidia Corp.'s (NVDA) graphics processing units. Post believes GPUs could have shorter lifespans than traditional assets due to high workloads and the rapid pace of AI innovation rendering assets obsolete faster.

Amazon, Meta and Google have actually revised the useful lives of server and network equipment higher over the past five years, allowing them to spread costs over longer periods of time. But Post is beginning to see a reversal of that trend. Earlier this year, Amazon changed the useful lives of some server and networking assets to five years from six years as a result of the increased pace of technology development brought on by AI.

There's also the risk of overbuilding AI infrastructure, causing supply to outpace demand. In this scenario, hyperscalers could enter into a pricing war to boost sales, commoditizing AI products and decreasing profitability.

Against this backdrop, it's even more important for AI technologies to ramp up monetization. Although advertising and cloud computing have been early areas of monetization for Big Tech companies, especially Google and Meta, depreciation and amortization expenses are growing at a faster pace. Bank of America estimates that Alphabet, Meta and Amazon combined will grow revenue at a 13% year-over-year rate in 2026, and 12% in 2027, while combined depreciation and amortization expenses are expected to grow at 33% and 30% over the same time periods, respectively.

Big Tech is still betting big on AI - Bank of America estimates that combined capex for Google, Meta, and Google will increase 22% to $333 billion in 2026. But the clock is ticking on monetizing this historic capex cycle.

Read on: The AI trade will make or break your stock portfolio. Here's how to win in 2025.

-Christine Ji

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

September 17, 2025 15:36 ET (19:36 GMT)

Copyright (c) 2025 Dow Jones & Company, Inc.