INSIGHTS

On October 12, 2026

AAM Viewpoints — The Price of Belief

Authored by: Ryan Severino, CFA, Chief Economist, BGO

Highlights

• Artificial Intelligence (AI) conviction and rivalry drive demand for capital
• Government borrowing intensifies the competition
• Excess demand puts pressure on real yields
• Strategic support can sustain AI investment
• Commercial real estate (CRE) faces costlier capital and constrained supply

Extraordinary belief in AI is creating extraordinary demand for capital. Businesses anticipate enormous profits, while governments see economic leadership and national security at stake. That conviction can border on zealotry when falling behind seems unacceptable. It can make even the high cost of capital seem tolerable. In turn, that AI investment has the potential to drive economic growth. But the consequences of that seemingly unflappable demand for capital spill over into areas beyond growth. Alongside heavy government borrowing, it helps put broad upward pressure on real yields. Add in inflationary pressure pushing nominal financing costs higher, and we have a situation where AI investment and the broader economy can continue expanding despite rising rates, even as more rate-sensitive investment, including commercial real estate, could get crowded out.

Capital Has a Real Price

A nominal Treasury yield incorporates a real, inflation-adjusted return and inflation compensation. When businesses want to invest more than savers want to fund at prevailing rates, the required real return tends to rise. Government deficits intensify that competition by reducing public saving and requiring Treasury borrowing. Domestic saving and foreign capital can respond, causing supply of capital to increase, but that can take time and still not fully balance supply and demand. Therefore, even with unchanged inflation expectations, the real cost of capital can increase.

Today, AI adds a private investment boom to familiar fiscal crowding out. Private investment booms have precedents, but extraordinary belief in AI’s financial potential and strategic importance reinforces demand for chips, data centers, power generation, and transmission. The largest firms bring substantial cash flow and financing access. Retained earnings carry an opportunity cost but reduce dependence on borrowing. And financing expansion from established businesses gives these firms more time for investments to pay off. But even the largest firms’ appetites are increasingly requiring access to credit markets. The spending already underway shows why. One global technology conglomerate spent $80.6 billion on property and equipment in the first half of 2026, compared with $39.6 billion a year earlier. A different global technology leader paid $35.8 billion for property and equipment in the quarter ended June 30, more than double the $17.1 billion a year earlier. These cash expenditures exclude equipment acquired through noncash financing arrangements. Of course, the company-wide totals cover AI and other activities but demonstrate substantial investment proceeding as yields rise.

Borrowing brings that demand directly into bond markets. On August 10, the first mentioned global tech conglomerate completed a $25 billion senior-note offering, including bonds maturing as late as 2066. The financing supports a company with multiple capital needs, so the entire amount cannot be assigned to AI.

Still, its size and long maturities illustrate the funding capacity technology firms can command, even alongside heavy Treasury issuance. Long-term corporate bonds and Treasuries ask investors to absorb overlapping interest-rate risk. If their combined supply grows faster than investor demand, buyers can require a larger term premium, the compensation for holding long-duration debt. Research has identified AI financing as a critical growing source of this duration supply. AI can therefore put pressure on long yields through both economy-wide competition for saving and the additional interest-rate risk bond markets must absorb. Inflation expectations can remain stable while that compensation rises.

The Cost of More Intelligence

Additionally, the nature of AI improvement identifies a key resource constraint. Recursive self-improvement (RSI) means AI helping develop better AI. Several key companies cite it as their explicit goal for AI investment as they seek to achieve artificial general intelligence (AGI). AGI is a hypothetical computer system that can match or exceed human capabilities across any intellectual task. Of late, the progress toward this goal is accelerating. But it appears that sustaining progress toward AGI can require increasing inputs, because research seems to exhibit diminishing marginal benefits (like many things in economics). Faster research may therefore continue to depend on costly computing capacity.

Consequently, if firms (or governments, more on that soon) remain committed to ambitious capability targets, rising resource requirements may prompt larger infrastructure budgets. Of course, a weakening of conviction could instead reduce investment, and breakthroughs could ease resource requirements. But resource intensity creates clear investment needs while beliefs and financing determine how much gets built.

For businesses, a useful comparison is expected investment returns against the risk-adjusted cost of capital. Rates still reduce present values, but extraordinary expectations can keep expected returns above the investment hurdle rate. Revenue growth alone might not establish profitability. Fiscal commitments and AI conviction can make both sources of capital demand relatively insensitive to rates. Total private investment can still grow as higher financing costs squeeze investment elsewhere.

The Stakes of the AI Race

But beyond corporate considerations, geopolitical stakes make this devotion even harder to dislodge. If business leaders and policymakers believe AGI could determine economic and military dominance, then losing to China (the United States’ only true competitor in the space) can look existential. Firms might fear being outcompeted, but the government might fear losing national security and strategic autonomy. Outcomes remain highly uncertain, but those perceptions influence spending. Paying too much for capacity can appear safer than having too little when a rival makes a breakthrough.

And here is where things intersect. Compute capacity is one of the United States’ clearest advantages. A June 2026 BCG report, citing Exponential View, put U.S. computing power optimized for AI at roughly eight times China’s. Epoch AI estimated in April 2026 that five U.S. hyperscalers owned more than 70% of global AI compute. Ownership estimates leave gaps, including illicit chip flows and offshore rentals. Computing capacity alone cannot guarantee leadership. Algorithms, talent, capital, energy, and deployment also matter. Nonetheless, advanced chips and large computing clusters provide a tangible advantage that firms and policymakers have reason to defend through continued investment.

And the current U.S. administration is reinforcing that resolve. Washington’s AI Action Plan explicitly connects winning the AI race to retaining economic and military leadership and prioritizes faster infrastructure development. Even current trade policy shows a bias toward AI investment. January’s 25% tariff on certain advanced chips exempted covered imports for U.S. data centers. That targeted exemption shows a willingness to accommodate the buildout within a restrictive trade agenda.

Keeping Investment Flowing

Geopolitical commitment could also sustain investment if private capital markets pull back. The government values national security, scientific progress (at least where it favors it), and strategic autonomy alongside financial returns. Policymakers who regard AI leadership as essential could support research, chips, power infrastructure, and data-center construction even if commercial returns disappoint. The question would become how much capacity Washington wants to secure and how it finances it.

The administration’s July 2025 data-center executive order directed officials to develop financial support for qualifying projects, potentially including loans, guarantees, grants, tax incentives, and purchase commitments. In March 2026, the Department of Energy announced a $293 million funding opportunity for AI research through its Genesis Mission. The research program demonstrates policy commitment on a modest scale. Replacing a broad retreat in private funding would require a much larger commitment, with funding and terms still to be determined.

The government once thought housing was critical enough to intervene on a massive scale. The current administration is taking equity stakes in firms. If the U.S. truly views losing the AI race as an existential threat, then investing in housing seems like small potatoes by comparison. Government could therefore become a durable customer for compute and help finance new capacity. Long-term purchase commitments could improve revenue visibility and support construction. Guarantees could encourage lenders to finance selected projects, while grants or public financing could sustain research and infrastructure directly. Credible expectations of support could encourage investment before officials act. Intervention would depend on legal authority, appropriations, and political choices, and could take time to organize. Money and guarantees also cannot immediately produce grid connections, transformers, chips, or skilled workers.

Public support could therefore limit the decline in total capital demand following a private-market retreat. The federal government would still rely on capital markets when financing additional spending through Treasury borrowing. Guarantees would create contingent exposure and could draw private lenders back. Either could sustain demand for scarce capital, power, and labor. Supply investments can ease constraints over time, while the initial spending keeps demand firm. More rate-sensitive sectors, including CRE, could continue facing financing pressure even as private enthusiasm for AI cooled.

This connects directly to our tongue-in-cheek weekly about AI from a couple of weeks ago. Why would credible concern about catastrophic AI outcomes fail to stop development? Because competitive incentives can overpower caution. This proves especially thorny when a lack of trust traps rivals in a Prisoner’s Dilemma (when two rational individuals acting in their own self-interest can end up with a worse outcome than if they had cooperated). Participants can perceive existential danger in both uncontrolled development and allowing an adversary to lead. That combination helps explain why acknowledging AI’s risks need not produce collective restraint. Fear can reinforce the same spending decisions as optimism.

Why Yields Are Rising

And of course, all of this is contributing to the rise in Treasury yields. The nominal 10-year yield rose from 4.18% at year-end 2025 to 5.18% on September 24. Over the same period, the 10- year inflation-protected yield increased from 1.93% to 2.85%. Of the 100-basis-point nominal increase, 92 basis points came from the real yield, while the nominal-real spread widened from 2.25% to 2.33%. Since August 31 alone, nominal yields rose 43 basis points and real yields rose 41. Those figures establish where the increase appears in market pricing. But they cannot isolate how much AI caused. Real Treasury yields reflect expected real policy rates and risk premia as well as investment and saving. The nominal-real spread, often called breakeven inflation, includes inflation expectations, inflation risk compensation, and differences in market liquidity. Fiscal borrowing and changing expectations for monetary policy also influence the outcome. Therefore, the evidence supports examining capital demand while leaving its precise contribution unresolved.

Inflation, of course, adds to these financing pressures. Energy costs, tariffs, and supply constraints can raise expected inflation, while uncertainty can increase the premium investors require for inflation risk. Both lift nominal yields. And persistent inflation can also keep expected Fed rates elevated. On September 16, the Fed raised rates while describing capital investment as robust (as we have discussed above). AI’s anticipated returns and strategic importance help explain why investment can proceed despite these additional pressures.

Timing helps reconcile these forces. Construction and equipment purchases increase demand today. Broader productivity gains depend on capacity coming online and businesses using AI effectively. Those gains could eventually lower production costs and inflation even as better investment opportunities keep real rates elevated. But the net effect on nominal yields would depend on both components. My longer-run anchor near 4% for the 10-year Treasury remains a useful benchmark. For most of the 14 years since the Fed formally established its 2% inflation target, the 10-year Treasury yield has been below 4%. But now forces are pushing in the other direction. But that should not be construed as a permanent reset of the anchor, just as it wasn’t on the downside. A permanent reset would require evidence of durable changes in factor inputs, productivity growth, saving behavior, or risk premia, which we do not foresee soon.

An Economy Still Spending

Last week, yields seemed to jump on the belated realization (at least among non-readers of this weekly) that the economy was becoming more interest-rate insensitive. September’s S&P Global flash PMI seemed to cause a bit of an awakening to this fact. The composite output index climbed to 58.4 from August’s 56.0, its strongest reading since July 2021. Hiring accelerated to its fastest pace since June 2022, while input costs rose at their quickest pace in almost four years. Even in the face of higher rates, growth was proceeding and poised to accelerate.

AI optimism also reaches households and supports the economy via financial markets. Gains in AI-linked equities increase wealth among their owners and can support spending beyond the technology sector. The physical buildout generates work for construction crews, electricians, engineers, and suppliers, while additional consumer spending supports other jobs. These channels help reconcile anxiety about future automation with current labor demand. Their benefits remain highly uneven. Additionally, elevated short-term rates also support income from money market funds and other interest-bearing accounts, reinforcing consumption alongside the equity wealth effect. Their returns follow short-term rates, so a higher 10-year yield alone does not raise them. Savers can benefit while borrowers face higher bills and existing bondholders suffer valuation losses.

CRE Implications

AI investment and resilient household spending support demand in selected locations and property types, while competition for funding raises financing hurdles. Higher debt service reduces borrowing capacity and increases the equity required at refinancing. Development also competes for construction labor, power, and equipment. Limited construction can support occupancy and rents where demand holds. Treasury yields and cap rates need not move together because income expectations and property risk also matter. Data-center underwriting must examine ultimate customers, power availability, equipment obsolescence, guarantees, and replacement tenant costs. A compelling national strategy cannot substitute for sound property economics.

Credit conditions still affect the pace and distribution of investment. Lenders could tighten terms if AI revenues disappoint, capacity payments outrun cash flow, or refinancing becomes harder. Wider spreads could raise project financing costs even if Treasury yields fell. Diversified firms with substantial cash flow have more flexibility than leveraged developers and compute providers. Continuing construction can coexist with defaults, restructurings, and credit losses. Falling AI equity prices could also weaken the wealth effect and household spending. Public support’s economic impact would depend on its scale and timing relative to these other drags, even if it preserved strategically important investment. The same conviction that sustains AI development despite concern about catastrophic risks can also sustain investment despite higher financing costs. Alongside heavy government borrowing, that demand for capital can push real yields higher even as AI investment supports economic growth. Government and the strongest AI borrowers can keep raising capital while more rate-sensitive investment, including CRE, could get crowded out.

CRN: 2026-1002-13836 R

The opinions and views of this commentary are those of BGO and are not necessarily those of Advisors Asset Management. Any forecasts or opinions expressed herein are BGO’s own as of October 7, 2026, and subject to change without notice.