Tech giants are setting records every few months for how much they are spending on artificial intelligence \u2014 but this time, some jitters are also growing.<\/p>\n
On Thursday, Amazon said<\/a> its capital expenditures totaled $53 billion in the second quarter, up 69 percent from a year earlier, as it built A.I. data centers and other infrastructure. It joined a parade of other big tech companies, with Meta<\/a> on Wednesday reporting that its costs had risen 55 percent from last year, while Microsoft said<\/a> its capital expenditures had soared 69 percent. Last week, Google also disclosed<\/a> that its costs had jumped and said it would boost its spending further.<\/p>\n These numbers are only set to skyrocket. Across this year and next, Amazon, Google, Meta and Microsoft are expected to spend a staggering $1.5 trillion building data centers and stuffing them with advanced chips, according to Wall Street estimates compiled by FactSet.<\/p>\n \u201cThe scale of it is nuts,\u201d said Melissa Otto, who leads research at S&P Global\u2019s Visible Alpha division.<\/p>\n Alarms are rising as Wall Street and others question when this spending can be justified. On Thursday, Amazon\u2019s share price rose almost 10 percent in after-hours trading after its cloud computing division experienced its fastest growth since early 2022, and shares of Microsoft, which did not change its spending forecast, jumped more than 15 percent.<\/p>\n But Meta\u2019s stock sank more than 7 percent on Thursday after it revealed its A.I. outlays and that costs were growing faster than its revenue. And last week, for the first time since going public in 2004, Google said it had \u201cnegative free cash flow,\u201d which meant it was spending more on day-to-day operations and building new infrastructure than it was taking in from its businesses. Google\u2019s stock fell more than 6 percent the next day.<\/p>\n Even so, the companies insist they are doing the right thing. \u201cAs long as we see these attractive opportunities to invest, we will continue to invest,\u201d Anat Ashkenazi, Alphabet\u2019s finance chief, told Wall Street analysts last week.<\/p>\n More than a dozen years ago, tech companies became the largest enterprises on the planet, fueled by their software and digital businesses. But A.I. has flipped tech\u2019s \u201casset-light\u201d model on its head. Multibillion-dollar data centers, which the industry likes to call \u201cA.I. factories,\u201d have become critical investments for developing advanced systems and making them available to customers.<\/p>\n From April through June, capital expenditures by the four companies totaled $170 billion, up 72 percent from a year earlier. The figures are ballooning \u2014 and projections keep rising every three months \u2014 as the companies race to open more data centers, and as the prices for necessary components like memory chips jump because of the demand. In April, Microsoft said soaring component prices would add $25 billion to its tab this year.<\/p>\n Despite all the building and spending, the tech giants say they do not have enough computing power to meet demand. In effect, they say, they have been leaving money on the table.<\/p>\n \u201cMore capacity = more sales,\u201d a Bank of America analyst, Justin Post, wrote last week in a research note after Alphabet, Google\u2019s parent company, increased its spending forecast for this year by $15 billion to as much as $205 billion.<\/p>\n The pent-up demand has contributed to a backlog in signed contracts for Amazon, Google and Microsoft. Combined, that backlog reached almost $1.7 trillion, more than doubling from a year earlier.<\/p>\n Much of the growing backlog comes from partnerships with OpenAI and Anthropic, the leading A.I. start-ups. That means the fate of the giants rests in large part on the start-ups\u2019 needing all the computing power they have requested, and their ability to pay for it. That poses a concentration risk, creating \u201ca more circular system that could mask true demand,\u201d according to Moody\u2019s Ratings.<\/p>\n