Invisible risk: how expensive AI investments could hit Alphabet, Meta and Amazon

The record investments of Alphabet, Amazon and Meta in artificial intelligence hide new financial risks, according to Bank of America. As data centers and graphics processors are put into operation, the accelerated growth of depreciation may exceed the rate of AI revenue and begin to put pressure on the companies' profitability.

Huge investments by big tech companies in artificial intelligence have hidden costs that many on Wall Street overlook, MarketWatch writes.

Thus, Alphabet, Amazon and Meta Platforms are expected to collectively spend $274 billion on capital investments by the end of the year, in an effort to increase the infrastructure for AI. And in 2026, according to Bank of America's forecast, the total amount will grow by another 33% to reach $333 billion. At the same time, such costs are not written off immediately in the financial statements: companies spread them over the years through depreciation and amortization of assets - for example, data centers or graphics processors. If these costs start to grow faster than AI revenue, or if the lifespan of the equipment turns out to be shorter than expected, this could start to weigh on the stock price, the publication explains.

At this point, says BofA analyst Justin Post, market projections don't fully reflect the magnitude of the AI investment cycle. "Wall Street is still lagging behind in terms of [accounting for] rising depreciation costs," he said in a Sept. 16 note cited by MarketWatch. Underestimating these costs creates the illusion of higher profitability at tech giants than there actually is, which could fuel a bubble around AI, the analyst warned.

BofA itself recently raised Meta's depreciation expense expectations for 2026 and 2027. According to Post, for Meta, Alphabet and Amazon, depreciation "will accelerate in 2026 as new infrastructure comes online, putting additional pressure on operating costs." In 2027, Alphabet could have a $7 billion gap between Bank of America's forecast and Wall Street consensus, MarketWatch points out. For Amazon and Meta, that gap could be $5.9 billion and $3.5 billion, respectively.

The bulk of the capital investment goes toward buying technical infrastructure, including Nvidia's GPUs. According to Post, GPUs may have a shorter lifespan than traditional chips due to high workloads and the rapid pace of AI innovation, which makes the hardware obsolete faster.

Over the past five years, Amazon, Meta and Alphabet, by contrast, have been extending the lifespan of server and networking hardware to stretch costs over a longer period. But that trend is starting to change, Post observes. Earlier this year, for example, Amazon shortened the lifespan of some of its server and network assets from six to five years.

Shares of Alphabet and Meta have risen by about a third since the beginning of the year, Amazon's quotes have added just under 6%. For each of the three companies, the vast majority of Wall Street recommendations are advice to buy their securities. The most optimistic analysts look at the prospects of Amazon, they offer to buy 93% of experts.

This article was AI-translated and verified by a human editor