Mexico is debating whether retirement savings should be used to finance roads, ports, and other national infrastructure projects. However, there is a much larger infrastructure project already being funded with little attention, and some of that money is Mexican. The construction is not in Sonora or on the Isthmus. Combined, those four companies are on track to spend more in a single year on this deployment than the entire economy of Chile. None of this makes the venture fantastical. It is a bet that a giant cycle of fixed investment will generate returns quickly enough to satisfy public markets. This matters to Mexico because the country is not watching from the sidelines. In practice, Mexican retirement savings are exposed to the valuations of the U.S. stock market at a time when those valuations are being pulled, unusually, by a narrow group of technology companies. The most useful question is who is carrying the risk while the answer remains undefined. How much of Mexican workers' retirement savings is ultimately tied to the Magnificent Seven, either directly or through the construction of the indices themselves? And what proportion of that exposure has been treated as broad market risk when it is, in fact, something much more specific? By the time the political class decides whether pensions should finance infrastructure at home, it might be worth recognizing that the same savings are already financing it abroad. Technology can be transformative, and the eventual economic gains could justify much of what is being spent. Mexico is stuck in a political debate over whether pension money should be channeled into national infrastructure. The country hesitates over a port in Sonora while its pension system, in effect, helps pay for another data center in Virginia. The risk is not necessarily that this wave of spending will collapse. Meanwhile, through listed markets, that money is already helping to finance the physical expansion of U.S. digital infrastructure. Some of the capital behind this deployment comes, indirectly, from the pensions of Mexican workers. It is one of the least commented features of the current tech boom. For a time, the focus was on the models, the chatbots, and the competitive drama around each new launch. But times matter. How severe would the fall be if enthusiasm for the spending cycle were to cool suddenly? It is too early, and probably too simplistic, to force the choice in those terms. These investments are going to appear in the portfolios of institutions that were never really described as betting in a concentrated way on the infrastructure layer of the U.S. tech sector. The Afores among them. So the question is not whether AI is a bubble or a revolution. Lately, the story looks less like software and more like heavy capital formation. OpenAI, reportedly, operates on around $25 billion in annualized revenue. Amazon seems to be heading toward $200 billion. It could work. The revenues are substantial. If the returns come late, or turn out to be more uneven than expected, the consequences are not going to be contained to the venture investors or the founders of the companies. Markets are not just being asked to believe in the utility of AI in the long run. Nvidia closed its fiscal 2026 with $215.9 billion in revenue. They rather resemble an industrial cycle: it depends on land, transformers, substations, cooling equipment, semiconductor supply, and access to energy. Meta says it expects to spend between $115 and $135 billion on AI-related capex this year. But when you put those numbers up against the cost of the underlying infrastructure, the picture changes. What is being bet on is not simply better software margins. But anyone looking at the concentration in the index can see that a growing portion of global equity exposure already amounts to a fairly concentrated view on a handful of large U.S. companies and their ability to turn extraordinary spending into durable profits. And that is where the domestic debate starts to look a little odd. At the end of February, the Afores managed 8.67 trillion pesos, about a quarter of GDP. They are being asked to believe that unprecedented capital spending today will translate into profits on a timeline that sustains current valuations. That is a more fragile proposition than the general enthusiasm suggests. Microsoft and Alphabet operate on a similar scale. About 13.5% of that amount, more than $60 billion, was invested in international equities, much of it through vehicles linked to indices like the S&P 500 and the MSCI World. In theory, prudent diversification. The figures involved don't much resemble the old Silicon Valley playbook. Anthropic is approaching $19 billion. It's in northern Virginia, in Texas, and in Ohio, where data centers are going up at a notable pace. The MSCI World still carries the label of a global index.
