Nearly four years into the artificial intelligence boom, the world's largest technology companies continue to make grand promises about the future, but the problem is they are burning through cash at an alarming rate. According to forecasts from Goldman Sachs, hyperscale tech companies are expected to spend $765 billion on AI this year, with that figure rising to nearly $1.2 trillion by 2027.
Amazon.com raised its full-year capital expenditure forecast to $220 billion on Thursday, the highest among the four major hyperscalers. The company also reported negative free cash flow of $7.6 billion over the past twelve months. Just a day earlier, Meta Platforms, Inc. disclosed that its cash generation ability had dropped 91% compared to the same period last year. Last week, Alphabet reported its first-ever negative free cash flow on record, a startling development for one of the most profitable companies on the planet.
Alphabet's CFO, Anat Ashkenazi, told analysts on an earnings call that free cash flow would continue to face pressure as the company seizes the "AI opportunity."
As tech earnings season largely winds down this week—with NVIDIA scheduled to report on August 26—it is evident that AI investments are distorting balance sheets, even as industry leaders continue to tout the future returns from their massive bets on new data centers and the chips and systems inside them.
A major reason costs are exceeding earlier expectations is a memory shortage, driven by insatiable demand for AI processors that rely on memory from a handful of suppliers.
Tesla Motors CEO Elon Musk described memory pricing as "crazy" on his company's earnings call last week, while Amazon.com CEO Andy Jassy said that "inflated prices" for memory chips were pushing up his company's capital expenditure guidance.
Apple spends far less than its big-tech peers, but it is particularly vulnerable to the memory crisis because the technology is a critical component of every consumer device. Apple has already raised prices for Macs and iPads, and many analysts expect iPhone prices to increase later this year. On Thursday, the company issued a weaker-than-expected outlook due to what CEO Tim Cook called "supply constraints," a problem he does not expect to ease this year.
Cook stated on the earnings call, "Looking ahead beyond September, we see market pricing for memory continuing to rise, which could have an increasing impact on our business." He is set to step down as CEO on September 1. "We are continuously evaluating this situation."
For Apple, memory is a revenue problem as the company faces weak consumer demand due to higher prices. But for hyperscalers, memory is becoming a massive cost hurdle as prices for AI systems purchased from NVIDIA, which have extremely high memory demands, soar.
Musk even thanked memory supplier Micron for "giving the company a very important quota under reasonable terms."
Mixed Reactions
Investor reactions to the earnings reports have varied dramatically.
Tesla Motors and Alphabet saw their stock prices fall last week after reporting negative free cash flow and pointing to accelerating spending. Meta Platforms, Inc. saw its shares plummet on Wednesday after reporting a weak outlook and ongoing uncertainty around its AI monetization strategy. Meanwhile, Microsoft enjoyed its best day in the market since 2008, thanks to better-than-expected results coupled with increased capital expenditure guidance.
Wells Fargo analysts wrote in a note to clients, "There is significant room for valuation re-rating at Microsoft," recommending the stock. The rally trimmed Microsoft's year-to-date losses to around 7%.
Apple's stock declined after its fiscal third-quarter report, as the memory shortage weighed on its outlook, while Amazon.com's surging cloud business growth was the primary catalyst for its stock price surge.
Evercore ISI analyst Mark Mahaney said after the report, "The revenue growth is not only impressive, but profitability is also improving." Mahaney noted that Amazon.com's AWS growth rate had been lagging behind Microsoft Azure and Google Cloud, "and this is the breakthrough the stock needed."
Wedbush analysts stated in a Friday report that Amazon.com's report represented the "cleanest earnings beat" among the hyperscalers they cover, and management provided the clearest explanation of how it would achieve returns on its capital expenditure.
The analysts wrote, "In our view, the clean earnings beat and detailed explanation, compared to what we consider equally strong fundamentals and capex raises for Google and Amazon.com, are the factors leading to the different stock reactions."
However, across the entire mega-cap space, despite healthy revenue growth, no stock—unless you count Micron—is having a breakout year. The market's tepid performance reflects growing investor skepticism about whether this increasingly debt-fueled massive AI buildout will ultimately generate returns.
Then there is the challenge from Chinese peers.
In recent months, a group of Chinese AI labs have released a series of new and updated AI models that are closing the performance lead held by OpenAI and Anthropic at a much lower price point, catering to the broader trend of US corporations becoming more frugal with AI service spending. These so-called "open-weight" models can be downloaded, customized, and hosted on any infrastructure the user chooses.
Because so much of the AI market is built around OpenAI and Anthropic—both valued at nearly $1 trillion in private markets—any potential threat to their business poses a risk to the entire AI trade.
In a report last week, JP Morgan Chase investment strategist Dana Harlap posed a rhetorical question: "Is it all one big AI trade?" Harlap suggested that the market's reaction to Alphabet's report showed that Wall Street is scrutinizing spending.
This is even true when the company beats revenue estimates, which Alphabet did while reporting 82% growth in its cloud business.
Harlap wrote, "We see the market becoming more selective—more discerning—towards hyperscalers as investors try to differentiate AI winners from losers. In the long run, whether hyperscalers can generate acceptable returns on their massive capex investments is likely to be linked to the returns of the AI ecosystem."
Comments