In economics 101, we learned of something called the “crowded-in” effect.  This is where increased government spending can stimulate economic activity and business confidence, thus prompting companies to invest more capital into structures, equipment and intellectual property. On the flip side, there is another effect that we term “crowding out”, where rising government deficits can push up rates and starve the private sector of financing.

Today, these two opposing dynamic forces remain, but is amplified by a new distortion: AI is concentrating capital expenditure into a tiny cluster of firms and sectors. Instead of broad-based investment cycles, we now have a hyper-concentrated one, with hyperscalers, semiconductor suppliers, and data‑center energy infrastructure absorbing the lion’s share of new investment. Markets once worried about governments hogging the credit card; now the private sector is doing its own hogging—but only one segment of it.

As a refresher, when economists worry about governments running large deficits, the fear is straightforward: public borrowing would “crowd out” private investment by absorbing the available pool of savings and pushing borrowing costs higher. The story makes intuitive sense in a world where capital is finite, productivity is predictable, and corporate capex follows the gentle cycles of economic expansion. Put simply, there are only so many investors with so many dollars to lend to those in need. Governments issue bonds to finance deficits, but someone has to buy those bonds. Yes, there is a not-so-subtle nuance and that is the fact that the Federal Reserve and other central banks also buy those bonds, but the endgame is still the same – too much borrowing vacuums up more dollars that were available for private investment.

As the above chart illustrates, a four-year expansion in the U.S. economy has allowed the level of federal debt as a percentage of the economy (GDP) to fall from its pandemic peak above 120%. Progress was halted in 2022 when interest rates rose alongside tighter monetary policy, and we are now back above 116%. While the yield curve has shifted lower, the US fiscal situation remains a mess. Economic growth is slowing, which means tax revenue growth will also fade – exactly at the same time that a multi trillion-dollar tax cut package is about to go into effect in 2026. As for tariffs covering the bill, the Supreme Court is in the middle of deliberations that could see Trump’s “national security” tariffs rolled back and tariff monies returned to those U.S. businesses that have paid them.

Unlike other countries around the world that have increased deficit spending on capital projects and defense, Washington is watching its deficit remain elevated without the benefit of productivity-enhancing public investment. Recently, Trump has promised cheques to Americans to help them with affordability challenges created by his own tariff policies. Yet, the impact on federal debt and how that growth in net liabilities can still crowd out private investment remains.

Fast-forward to 2025 and this traditional crowding out framework is starting to look a little dated — not because public borrowing has stabilized (it hasn’t), nor because deficits no longer matter (they do, but only when bond markets decide they do). Instead, the traditional crowding-out narrative is being overtaken by something more dynamic, more distorted, and frankly far more consequential: the rise of AI-driven capital concentration. We in turn are entering a world where the biggest constraint on investment isn’t government borrowing — it’s the gravitational pull of artificial intelligence on global capital flows.

Normally, where increased government borrowing crowds out business investment in plant and equipment, we would expect to see this in the data. Growth in the former would subtract from growth in the latter. The problem now is that official statistics on business investment will include the spending on servers, data centers, specialized chips, transmission infrastructure, cooling systems, etc. Case in point, when we look at overall business investment in the U.S., there hasn’t really a major setback. The above chart shows that investment grew in the first half of the year, with almost an 8% annualized pace recorded in the first quarter, slowing to just above 2% in the second. Thanks to the government shutdown, data for the third quarter has been delayed, though we expect it to be released by U.S. Thanksgiving.

Again, when we consider the plethora of deals on the AI front in recent months, it is doubtful that overall business investment growth will show any signs of a major pullback in the quarter and the same may be true for the fourth quarter. In fact, S&P Global recently suggested that data-centre investments and the ancillary segments that feed these investments (energy, servers, software, etc.) could have added half a percent to overall U.S. real GDP growth in the second quarter. Put another way, if these investments were not made, they would not have masked the underlying weaker story in business capital formation outside of AI.

In the above chart, we are looking at quarter-over-quarter growth (or lack thereof) in U.S. non-residential construction. This is a vast category, covering commercial and industrial buildings, healthcare, infrastructure and the equipment and software associated with these projects. Investment peaked at $713.8 billion (annualized) in the fourth quarter of 2023 and it has fallen for six straight quarters. The downward trend is noticeable in general commercial space, multi-merchandise shopping, warehouses and mining. Office investment at around $168 billion in Q2, represented about a third of overall commercial and healthcare construction and data centres were $25.5 billion of that total. It sounds small, but this is a 100% increase over two years. Investment in electrical power has also increased from around $104 billion at the end of 2023 to $110 billion in Q2.

Keep in mind, part of Trump’s policy agenda has been about bringing manufacturing back to America. Biden wanted that too and we did see growth in manufacturing construction from about $113 billion at the start of 2023 to a peak of $151 billion by the third quarter of 2024. Since then, it has slid to $142 billion in Q2. That might show a turnaround in the second half of this year, but we won’t know for sure until well into Q1 of 2026.

It is quite possible that many industries will find themselves starved of investment not because they failed strategically, but because capital allocators are afraid NOT to be in AI. If so, we might end up with a period of time defined by capital scarcity — not due to a lack of money, but a lack of capacity. This is no longer a case of government investment “crowding in” private sector growth, as economists like to say when public spending unlocks private follow-through. We are now watching an inversion: a private-sector megacycle that is at risk of crowding out virtually everything else. At the same time, the direction of fiscal policy and its potential to send yields on an upward trajectory could also create the crowding out effect that also impedes private sector investment.

The macro question isn’t whether AI will pay off. It’s what gets sacrificed along the way. Ally and I have talked with clients over the last few months of the risk that the U.S. and the world, for that matter, enters a situation of overbuild with respect to data centres and power generation linked to AI. We are definitely seeing investors beginning to take account of this risk in recent weeks. To the extent that productivity enhancing capital has been diverted from segments of the economy that could have contributed to economic growth and corporate revenues/earnings, this will ultimately not sit well with market participants.

On behalf of the Pyle Wealth Advisory team, have a wonderful weekend.   

Andrew Pyle