The scale of capital expenditure in the technology sector has reached an unprecedented level, with major players like Alphabet, Amazon, Meta, and Microsoft collectively projecting over $630 billion in spending. This figure, disclosed through recent earnings calls, represents a dramatic increase in investment, primarily directed towards scaling artificial intelligence compute capabilities. These sums are not merely large; they are comparable to the annual gross domestic product of entire nations, such as Sweden or Israel, highlighting the immense resources being poured into AI infrastructure.
Alphabet, Google’s parent company, announced plans to nearly double its capital expenditure to an estimated $185 billion by 2026. Amazon followed suit, revealing a staggering $200 billion allocation, significantly surpassing prior Wall Street expectations. Meta, weeks earlier, had indicated its full-year capital expenditure would climb to as much as $135 billion. This coordinated surge in spending signals a focused strategic pivot, with much of this capital earmarked for the physical infrastructure – data centers, advanced servers, and robust power systems – essential for the burgeoning AI landscape.
Gil Luria, managing director and head of technology research at D.A. Davidson, articulated the sentiment within the industry, noting that such extensive investment is unparalleled. He acknowledged, however, that the technology driving this buildout also possesses unprecedented promise. These new data centers are not small undertakings; some are envisioned to be the size of multiple football fields, or even dwarf landmarks like Central Park in Manhattan, demanding extraordinary energy and resources for their construction and ongoing operation. The physical footprint of this expansion is so significant that some experts, like Jefferies analyst Brent Thill, anticipate that even existing commercial spaces, such as large shopping malls, could be repurposed to house these massive computing facilities.
Despite the monumental investment, Luria suggests this is not merely speculative spending but a response to substantial existing demand. He points to record-high backlogs for companies like Amazon, Meta, and Microsoft. Microsoft, for instance, has seen its order backlog double to $625 million, largely attributed to its collaboration with OpenAI. Thill elaborates that this buildout directly addresses a critical bottleneck in the AI industry. What began as a chip shortage, then evolved into an energy constraint, has now shifted to a scarcity of suitable physical infrastructure, or “physical shells,” to house the necessary hardware.
This aggressive capital outlay, however, has not been without its repercussions in the financial markets. A notable wariness regarding software valuations has emerged, triggering a significant week-long selloff across tech stocks and cryptocurrency. Investors are grappling with the uncertainty surrounding AI’s immediate profitability, even as firms express long-term bullishness. This uncertainty, compounded by weaker jobs data, contributed to nearly $1 trillion being wiped from software and services stocks. Investors in these tech giants are reportedly growing anxious, according to Luria, as companies appear to be committing a substantial portion of their available capital to these infrastructure projects. Shareholders, he notes, are seeking returns on investment rather than continuous, massive capital deployment.
Nevertheless, Big Tech remains confident in the high return on investment potential of AI. Thill characterizes the current competitive landscape as a “game of leapfrog,” with several major public vendors vying for dominance. He suggests that the primary concern among tech firms isn’t the risk of overbuilding, but rather the risk of not investing enough. Even in a scenario of overcapacity, Thill believes there would be ample demand to absorb any surplus, with various entities, including state, local, and federal governments, eager to procure additional AI infrastructure.
