원문 보기
Welcome to the Explainer. Okay, let's get straight to the point, because we're seeing a massive, absolutely pivotal shift in how the market perceives artificial intelligence. For the past few years, we have treated generative AI as a pure algorithmic software revolution happening somewhere in the cloud.
But today we lift the curtain to show what it really is. This is a project that requires huge capital investments in construction and physical infrastructure. We are officially moving from cloud software to a concrete reality. And you know, this reality check comes with a pretty serious warning. BlackRock CEO Larry Fink recently said that a moratorium on data center construction would cause an immediate recession in the United States .
Just think about this for a moment. This isn't some exaggeration from a Silicon Valley tech enthusiast, is it? This is a carefully calculated macroeconomic warning from a leader who manages more than $15 trillion in global assets. He tells us bluntly that stopping data center construction would literally freeze the broader US macroeconomics.
So, here's our plan for this release. We will consider: first, a warning about AI infrastructure. Second, there is a deadlock in the power grid. The third, growing bottleneck for big tech. Fourth, winners and losers in infrastructure. Fifth, two macroeconomic scenarios. And finally, sixth, a guide to action for investors.
Let's start with the first part—a warning about AI infrastructure and unprecedented concentration of capital. Okay, let's talk about the scale of capital involved . 5 trillion dollars. This is the projected total capital expenditure of leading technology companies through 2030. In fact, for comparison, annual spending will approach $1 trillion by 2027 alone.
And the main conclusion here is that this capital no longer funds just lines of code. It goes directly into the physical concrete infrastructure. So, where exactly is all this money going? They go to top-tier design and construction contractors, heavy-duty industrial cooling and power distribution, commercial land acquisitions, high-voltage grid upgrades, and global semiconductor plants.
The absolutely critical point is that this hyper-scale capital spending is currently adding about 40-60 basis points directly to quarterly real US GDP. Therefore, if local permitting restrictions suddenly halt this construction chain, this capital spending will cease instantly. And this shock will spread to the basic materials sector and industrial payrolls.
Deadlock in electricity networks: local obstacles against federal ambitions. Over the past two quarters, more than 100 municipalities and at least 15 US states have called for a temporary moratorium on new hyperscale data centers. Local utility commissions are putting up serious resistance, and it comes down to three main reasons.
The first is the limited capacity of power grids. These new gigawatt campuses require the energy of medium-sized cities, completely putting regional energy resources at risk. Secondly, the negative reaction of retail consumers. Ordinary household consumers are seeing their electricity bills increase by double digits.
And thirdly, the depletion of water resources. Cooling these huge facilities places incredible strain on municipal aquifers. This problem is very real. In the first half of this year alone, the implementation of planned projects worth about $130 billion was suspended . And the time constraints we deal with in the real world are, frankly, mind-boggling.
Imagine. At the first point, you order a project. At the second point, you are waiting 120 to 180 weeks just for the delivery of a large transformer. But the third point is the main “bottleneck ”. In regions like PJM, it takes five to seven years to connect new facilities to the grid. This is complete absurdity, if you think about it .
High-tech servers will arrive years before the physical energy is available to power them. Because of these crazy deadlines, the energy market has completely split. On the one hand, you have local solutions or existing nuclear capacities. They now have a huge valuation, simply because the energy is guaranteed.
On the other hand, there are speculative new projects that do not have guaranteed grid connection agreements. They face serious financial losses . Because if they can't connect to the network, it's effectively frozen capital. A growing problem for big tech companies: the threat to capital efficiency.
When such local permit freezes begin to multiply, the capital efficiency of large tech companies takes a direct, painful hit. And this usually happens in three stages. Stage one— geographic arbitration. Hyperscalers are effectively forced to leave primary nodes and flee to secondary markets or offshore sites, which means higher latency and greater logistical costs.
Stage two— decreasing return on invested capital (ROIC). Listen, if you buy $5 worth of cutting-edge chips and your data center is delayed by 18 months, those extremely expensive chips are just sitting in storage, depreciating in value and completely wiping out your investment income. And this leads to the third stage— the inhibition of free cash flow.
If capital expenditures remain at these record levels, but facilities cannot be connected to energy to generate revenue, shareholder value is under serious threat. So, chapter four, winners and losers in infrastructure, tracing the supply chain. Money flows in a very specific way here. It all starts with the capital expenditure of hyperscalers in the first stage, moves to industrial equipment suppliers in the second, and finally reaches semiconductor manufacturers in the third stage.
But this physical "bottleneck " of the power grid that we talked about acts as a strict selection mechanism between these last two stages. For example, industrial equipment suppliers are incredibly well protected. If a major project in New York is put on hold, it's okay. They simply reroute that liquid cooling unit or distribution equipment to a facility in Texas or Ohio.
They have orders for years ahead and very strong pricing power. But the semiconductor supply chain faces extraordinary timing risks. As we noted earlier, hyperscalers simply cannot deploy advanced chips without powered physical racks. A prolonged permit freeze means these chips are piling up in warehouses, which could significantly lower semiconductor valuations to their historical averages.
You might ask , how is all this financed without completely destroying the balance sheets of big tech? Well, a $1.5 trillion private lending strategy is coming onto the scene . Traditional corporate balance sheets simply cannot finance a $5 trillion expansion without a credit rating downgrade. Therefore, alternative investment managers are intervening using tools such as securitization of computing assets, infrastructure joint ventures, and off-balance sheet structures funded by sovereign wealth funds and private pension capital.
That's why investment managers are now so fiercely defending data centers. This is literally the largest platform for attracting private lending in a decade. Two macroeconomic scenarios for the next 24 months. Looking to the future, we are effectively at a crossroads. Two very different paths for the next 12–24 months.
Scenario A— federal priority regulation, to which Source analytics gives a 60 % probability. This is an optimistic scenario where the federal government steps in, recognizes AI as critical national infrastructure, and accelerates the construction of regional power lines to bypass local zoning moratoriums.
Then we have scenario B, regulatory " contagion" with a probability of 40%. In this pessimistic scenario, local political opposition escalates into a complete ban at the state level. Hyperscalers are forced to freeze additional infrastructure orders, and boom—the same recessionary environment that Larry Fink warned about—is becoming a reality.
Let's wrap it all up in chapter six: an investor's guide to positioning capital in the AI era. Based on everything we've covered today, here's an institutional-level asset allocation strategy . The tactical position is to increase the share of independent electricity producers, mainly because they have fantastic pricing power over existing assets “behind the meter”.
It is also worth increasing the share in networking and cooling equipment due to their huge order portfolios for many years. For cloud hyperscalers, the position is neutral or selective. Yes, they are very resilient, but you definitely have to watch for free cash flow compression if capital expenditures remain high.
We are considering undervaluing the share in commercial speculative real estate that does not have guaranteed agreements with utilities. There are too many risks of loss. And finally, private credit and infrastructure are given the upper hand. Thanks to the exceptional opportunities to receive commissions from off-balance sheet financing.
To summarize this action plan, here are some concrete steps you can take right now. First, check your infrastructure assets, including the status of their connection queue. Second, watch for margin discrepancies. Watch closely to see if capital spending starts to erode the quarterly operating margins of these tech giants.
And third, start treating state utility commission hearings as first-level macroeconomic events. Seriously, they are now as important as regional inflation or production data. So, AI trading has fundamentally changed. This is no longer just an abstract algorithmic race in the cloud. This is a fierce battle for physical infrastructure, fought for concrete, copper, and kilowatt-hours .
Investors who truly understand this transition are closely monitoring municipal permits and debt structures financing development. So I'll leave you with this final thought while you review your own assets. When the digital world collides with the physical, will your portfolio be powered by the web, or left in the dark?
Comments 0