Everyone knew, but no one looked: New cycle of tech spending and its risks

In the world of capital, there are times when warnings don't come in the form of blaring headlines, but rather in the figures that quietly slip into quarterly reports.

In the style of Chronicle of a Death Foretold, where everyone knew what was coming, but no one acted, the tech market seems headed toward a point of tension that, although evident in the numbers, is fading into collective euphoria.

Big tech is spending like never before. Combined capex and R&D spending for the top Nasdaq companies reached $420 billion over the past 12 months, a 19% year-over-year increase and the highest level ever.

Of that total, nearly $160 billion corresponds to investment in physical infrastructure—data centers, computing networks, electrical installations—most of which is geared toward deploying or scaling generative artificial intelligence capabilities.

This massive spending is concentrated in a few hands: Microsoft, Amazon, Alphabet, Meta, and Apple account for 76% of total capex, while NVIDIA, Broadcom, and AMD lead R&D spending within the semiconductor sector, accumulating more than $40 billion in the last 12 months.

The reasons are well-known: maintaining a competitive advantage, absorbing the explosion in demand for accelerated computing, and, above all, capitalizing on the momentum of artificial intelligence before it dissipates.

But what until recently was interpreted as unquestionable strength—the aggressive reinvestment of leaders—is beginning to generate concern. Not so much because of the absolute amount, but because of the way it is sustained: based on expectations of future revenues that, if not met, would leave a cost structure that is difficult to reverse.

A critical part of the problem is that the business ecosystem is increasingly interconnected and leveraged around a common narrative. According to Bloomberg estimates, the flows between OpenAI—valued at $500 billion—and key players such as Microsoft ($3.9 trillion), NVIDIA ($4.5 trillion), AMD, Oracle, xAI, and CoreWeave form a web of cross-investment that multiplies the risks. NVIDIA plans to invest up to $100 billion in OpenAI; Oracle signed a $300 billion cloud deal, and AMD granted them an option to acquire 160 million shares, equivalent to about $28 billion.

These commitments imply not only expectations of exponential growth, but also a mutual reinforcement of valuations: OpenAI depends on the infrastructure of NVIDIA and Oracle, which in turn are betting that OpenAI's growth will justify their own investment levels. This creates a circular architecture, where each link validates the next without any obvious anchor in tangible results.

The situation is reminiscent, in a more sober tone, of the movie Don't Look Up, where experts warn of a threat, but the rest of the world is too distracted or engaged to react.

The threat here isn't an asteroid, but rather the risk that AI revenue projections won't materialize or that margins will begin to shrink under the weight of competition, regulation, or energy costs.

If that happens, the adjustment would be systemic: it would affect not only startups that thrive on narrative, but also the giants that finance them, supply them with chips, provide them with cloud computing, or integrate them into their balance sheets.

The parallel with Icarus is not lost on us: in his ambition to reach the sun, the protagonist forgets that his wings are made of wax. Tech companies, by building data centers that consume more energy than small countries and spending tens of billions on future capacity, risk forgetting that cycles of technological euphoria are not eternal, and that the market—no matter how efficient it may seem—is also driven by faith.

For now, the results continue to support the narrative: revenues are growing, margins are still holding up, and investors continue to buy into the dream. But the first cracks are already showing: negative margin surprises from companies like AMD and Intel, weak guidance from some enterprise software companies, and the delay in monetizing some foundational models could be signs that vertigo is beginning to overtake enthusiasm.

History doesn't predict the future, but it does offer warnings. As the quote attributed to Mark Twain goes, "History doesn't always repeat itself, but sometimes it rhymes." In this case, the characters are already in place, the relationships are established, and the incentives are aligned.

The question is not whether a collapse is imminent, but whether someone, this time, will decide to look up.

Big tech is spending like never before. Combined capex and R&D figures for the top Nasdaq companies reached $420 billion.