AI trading "ignored risks": in case, a vast amount of capital expenditure "cannot be spent"

The story of AI is evolving from "software eats the world" to "hardware gets stuck in the world". In the extremely polarized political environment of the United States, it is almost impossible to find issues that can bring together far-left Senator Bernie Sanders and far-right Governor Ron DeSantis - except for "restraining data centers". This is not only a political spectacle in Washington, but also a cold "physical correction" that Wall Street is facing. When Silicon Valley giants wave checkbooks even more expensive than the "Apollo lunar landing program" in an attempt to sustain AI prosperity through stacking computing power, they are hitting a wall built by political and physical power grid limits. Legislators in New York have also introduced legislation to impose a minimum three-year moratorium on the construction and operation of new data centers. New York is at least the sixth state considering pausing the construction of new data centers. In short, from community protests in Florida to regulatory brakes on the Texas power grid, a market-ignored risk is rapidly escalating: if the physical power grid is not connected, and the political environment does not approve, the billions of dollars in capital expenditures that were originally included in valuation models may simply be unable to be spent. When Sanders and DeSantis "collude" Sanders and DeSantis are at odds on most issues, but they have reached a rare consensus on the issue of the rapid increase in data centers: the brakes must be applied. This bipartisan "union" animosity is rooted in the acute pain of the American people over the "side effects of AI". Across the country, the continuous low-frequency noise from data centers is unbearable for neighboring communities, the huge cooling demand is causing local water resources to be in crisis, and residents and small businesses are experiencing soaring electricity bills. Public protests are growing louder and louder. The sharp change in attitude of Florida Governor DeSantis is the most significant manifestation of this political trend. Just last June (2025), he signed a major tax relief bill extending the tax exemption period for data centers from 2027 to 2037. However, faced with the increasingly vocal public protests, DeSantis quickly changed his stance. "We don't want to subsidize technologies that will replace the human experience," DeSantis said at a recent roundtable meeting. He called for the enactment of an "Artificial Intelligence Rights Act" and supported legislation requiring data centers to pay their full water and electricity costs. He emphasized that local communities should not foot the bill for the expansion of these "richest companies in human history," saying, "You shouldn't pay a penny more for this." This rhetoric is in line with Sanders' stance. Sanders had previously released a report warning that if decisions are made by billionaires on boards who only care about short-term profits, technology will not improve the lives of workers. He explicitly called on Congress to pass legislation to pause the construction of new data centers: "I think we need to slow this process down." Politically savvy legislators are following suit. Arizona, Georgia, Virginia, and other states are pushing for legislation to cancel tax incentives or prohibit the signing of non-disclosure agreements (NDAs) that withhold details from the public; and in Georgia, Oklahoma, and Vermont, legislators are even proposing to suspend the construction of new projects directly, as Sanders suggested. For tech giants, the era of "red carpet" investment attraction has come to an end. Can the massive capital expenditures be spent? If political resistance is a "soft constraint," then the bottleneck of the physical power grid is a more deadly "hard wall." Wall Street is currently facing a troubling logic paradox: does the market truly believe that the expected $600 billion in capital expenditures set to appear in 2026 can be implemented? According to the latest data, the AI infrastructure spending plans of just four tech giants - Microsoft, Meta, Amazon, and Google - this year amount to a staggering $670 billion. In terms of the proportion of the US GDP, this scale exceeds the "Apollo moon landing program" of the 1960s and the "Interstate Highway System" of the 1970s, ranking second only to the "Louisiana Purchase" of 1803. Amazon alone plans to increase capital expenditures by nearly 60% to $200 billion this year. Most of these massive funds will be used to build data centers, which require massive amounts of energy. According to BloombergNEF's forecast, by 2035, the energy demand of data centers will double, soaring from 34.7 gigawatts (GW) in 2024 to 106GW, equivalent to the electricity consumption of 80 million households. The problem is that the current US power grid is simply unable to meet this demand. This physical constraint has evolved into a regulatory crisis in Texas. As the second largest hub of data centers in the United States, with only Virginia surpassing it, ERCOT (Electric Reliability Council of Texas), the Texas grid operator, is applying unprecedented "brakes" on projects. ERCOT has proposed a review of projects consuming approximately 8.2 gigawatts of electricity - equivalent to the power output of 8 traditional nuclear reactors. It is worth noting that many of these projects had already been approved in the past. Currently, ERCOT has launched a review mechanism called "Batch Zero" to review projects in batches to assess their overall impact on the grid. Katie Bell, Meta's energy project manager, admitted that some projects have been submitted for 18 months and still do not meet the standards of "Batch Zero". This uncertainty is undermining the expansion plans of tech giants: if the grid is not connected, data centers cannot be built; if data centers cannot be built, the $670 billion budget cannot be spent; if the money cannot be spent, the anticipated AI computing power supercycle will turn into a bubble. Wall Street's most crowded trade: "physical correction" When the risk of "having money but unable to spend it" begins to be priced, the financial market's reaction is severe. Recently, the US stock market experienced the fourth largest single-day sell-off of "momentum stocks" in the past decade. It is noteworthy that even independent power producers (IPPs) and nuclear power concept stocks, which were originally seen as benefiting from AI prosperity, were not spared. The previous market logic was "AI electricity shortage is good for power stocks," but the current logic has evolved to: if the connection to the grid is hindered, new electricity demand cannot be realized. UBS analysts stated that due to concerns that new loads cannot be contracted with existing power generation capacity, giant IPPs such as Constellation Energy saw a significant decline in their stock price, with a decline of 27% year-to-date. The market has realized that without the physical expansion of the grid, mere power generation capacity is meaningless. This panic has led to the rise of "anti-AI trading". Funds are flowing out of high-beta tech stocks and into defensive sectors such as chemicals and regional banks. This is a typical "de-leveraging" drop, driven by both quantitative funds and active management funds. At present, the market must address this annoying paradox: either believe in the miracle expansion of the grid to accommodate the $600 billion Capex, or accept that we have reached a physical bottleneck. If it is the latter, it means no grid construction, no capital expenditure implementation, no chip demand - ultimately, the valuation bubble of the AI supercycle will face a burst. Currently, with Sanders and DeSantis colluding at the political level and ERCOT closing the door at the physical level, it seems that Wall Street is being forced to accept the second possibility. This article is translated from "Wall Street View"; GMTEight editor: Chen Siyu.