US debt soars, AI alone: QuantStreet warns of hidden concerns about profit mismatch

According to Woofun AI, the US financial market showed an extremely tense divergence in September 2026: as Harry Mamaysky, founder of QuantStreet Capital, issued a warning in his latest monthly investment letter, the market is facing a rare differentiation between a sharp sell-off of US debt and a strong rise in AI assets. This structural profit mismatch may be eroding the sustainability of AI investment logic in a high-interest environment.

The fixed income market faced extensive valuation pressure during the month, and the change in its core indicators far exceeded historical norms. Specifically, the 10-year US Treasury yield climbed rapidly from 4.75% on August 31 to 5.29% on September 30, with a cumulative increase of 54 basis points in a single month.

This drastic steeper yield curve directly led to a significant drop in bond prices. Some US fixed income assets fell between 2.3% and 5% in September. This sharp retracement is not an isolated event; it is a direct reflection of profound changes in long-term capital pricing mechanisms. When risk-free interest rates rise rapidly, the pressure on long-term assets to lose capital increases nonlinearly, which forces investors to re-examine the cost-effectiveness of asset allocation at current interest rates.

Notably, this sell-off was not triggered by a single liquidity shock, but rather the result of multiple macro-variables resonating, which included concerns about fiscal sustainability and implied repricing of future economic growth paths. For institutional investors who rely on fixed income as ballast stone, this round of adjustments means that long-term exposure strategies established in the past few years are facing severe tests, and the yield has broken through the psychological threshold of 5%, further marking that financing costs have entered a new high range.

In stark contrast to the turmoil in the bond market, extreme structural divisions have occurred within the stock market. The technology sector, particularly the semiconductor industry, continued to perform strongly in September, and the NASDAQ index showed relative resilience against the backdrop of overall market pressure.

At the same time, Bitcoin and momentum strategy products with technology and semiconductor stocks as core holdings have also made outstanding gains. However, this increase is not a general increase; it is highly concentrated in a few leading assets. On the other hand, the equal-weighted S&P 500 index, which represents a wider range of corporate fundamentals, showed weak performance, indicating that the upward momentum in the market did not effectively spread to small and medium capitalization stocks.

Furthermore, industries that are highly sensitive to interest rates, such as real estate trusts (REITs), utilities, and the financial sector, are generally under tremendous downward pressure. What's even more puzzling is that the US dollar and commodities rose simultaneously in September, breaking the inverse correlation that the two usually present.

This “double rise” pattern suggests that the market environment faced by investors is becoming more complex and chaotic. The traditional macro-hedging logic seems to have temporarily failed, and the flow of capital between different asset classes is characterized by irrational congestion, further increasing the degree of market differentiation.

The market has various interpretations of the driving factors behind the sharp rise in US bond yields, but most traditional narratives seem inadequate in the face of empirical data. The first mainstream view is that investors are losing confidence in the US dollar and are therefore selling off dollar assets.

However, Harry Mamaysky has reservations about this judgment because the dollar actually appreciated in September, which is inconsistent with the narrative logic of a complete sell-off of dollar assets. Of course, a stronger dollar does not completely rule out the existence of long-term credit risk, but at least it shows that the market is not experiencing a one-way crisis of confidence in the dollar. The second explanation focuses on the worsening state of the US fiscal situation, with investors demanding higher compensation for long-term risks. The authors agree that the available evidence is insufficient to support this strong conclusion. In particular, he pointed out that the market's implied inflation compensation index (inflation breakevens) remained relatively stable, and there was no change of the same magnitude as the increase in yield. The indicator reflects the yield difference between nominal treasury bonds and inflation-protected treasury bonds, and is a key window for observing investors' pricing on future inflation and related risk compensation.

If nominal yields rise and inflation compensation does not rise sharply at the same time, it is difficult to attribute all the changes to inflation expectations getting out of control. However, this does not mean that fiscal risk can be completely eliminated, because long-term yields are also affected by complex factors such as treasury bond supply, term premiums, real interest rates, and market liquidity. Stable inflation compensation alone does not prove that US debt is not at risk.

In contrast, Mamaysky favors the third explanation: the market is repricing future stronger economic growth and the huge capital requirements brought about by AI infrastructure construction. Major cloud service providers and technology companies continue to expand capital expenditure (capex) to build data centers, procure chips, and invest in supporting power infrastructure. These huge expenses mean that companies need to take up more capital, and may also increase external financing requirements.

If investors also expect AI to bring higher productivity and future profits, then rising long-term interest rates may not only be understood as worsening economic risks, but may also partly reflect changes in growth expectations and capital demand. This shift in perspective is critical because it links the sell-off in the bond market to the investment boom in the real economy. In other words, the rise in US bond yields may be the market's' vote 'for AI-driven economic growth, even though this vote is accompanied by high financing costs.

This logic explains to some extent why some tech stocks are still able to rise when the bond market is sold off — because they both share the same macro-narrative: that AI will reshape the global economic landscape, and current interest rates are the 'entry fee' to pay for this grand vision.

Data compiled by Woofun AI shows that at the stock valuation level, this logical game is reflected in a fierce confrontation between rising discount rates and AI's future profit expectations. According to the traditional valuation model, when other conditions remain the same, the higher the market interest rate and the higher the return required by investors, the lower the future profit of the enterprise will be converted into today's value. This is why high-valued growth stocks are generally more sensitive to rising interest rates.

However, the September market didn't exactly follow this linear logic. Mamaysky's explanation is that investors may think that AI will create strong enough future profit growth to offset the valuation pressure brought about by rising discount rates. Judging from the valuation formula, this is equivalent to two forces competing with each other: one share is an increase in the discount rate, depressing the present value of future profits; the other is an expected increase in profit, increasing the intrinsic value of the stock.

If the latter increases sufficiently, stock prices may continue to rise even in the face of higher interest rates. Companies such as AMD (AMD.US), Micron (MU.US), Intel (INTC.US), Cisco (CSCO.US), and Applied Materials (AMAT.US), as core suppliers of AI infrastructure, appear in the main positions of the Momentum ETF that the author is concerned about. As long as the market believes AI capital expenditure will remain high, upstream suppliers' profit expectations are likely to continue to be supported.

This logic of 'profit growth offsetting interest rate pressure' provides a theoretical basis for the strength of technology stocks, but it also lays a huge risk of uncertainty.

However, there is a core question in this logical chain: Why is ROCS (Rest of the Corporate Sector, Other Corporate Sectors) not benefiting at the same time? In September, the semiconductor industry showed strong performance, but the S&P 500 index performed weakly. Compared to a market capitalization-weighted index, the equal-weighted S&P 500 gives each constituent stock roughly the same weight, so it is better to observe whether the rise in the market spreads widely rather than being mainly driven by a few large technology companies. The author refers to companies other than semiconductors as ROCS. In his opinion, AI investment has a logical closed loop that takes time to verify. Businesses are now buying chips, servers, and software because these technologies are expected to increase productivity in the future. The market is willing to fund this construction process in advance, also because it believes there will be unrealized profits in the future. Therefore, it is not surprising that profits in all industries are not growing at the same time at this stage. The problem is that the stock market itself is forward-looking.

If investors are convinced that AI will significantly improve the future profitability of other companies, in theory, this part of the expectation should also gradually be reflected in the relevant company's stock price. However, there was no such broad rise in September. Semiconductor stocks continued to strengthen, while other companies did not receive similar valuation support. This raised the author's question: If the final buyers of chips don't get enough new profits, how can they bear the ever-increasing costs of AI procurement and construction over the long term?

This issue touches on the profit distribution mechanism in the AI investment chain and is also the root cause of current market differences.

To verify whether AI is actually bringing productivity gains, Mamaysky began paying more attention to macro data. According to revised data released by the US Bureau of Labor Statistics (BLS) on September 3, 2026, labor productivity in the non-farm sector increased 1.4% month-on-month and 2.2% year-on-year in the second quarter. Looking at the longer cycle, from the fourth quarter of 2019 to the second quarter of 2026, labor productivity in the US non-farm business sector grew at an average annual rate of about 2.1%, higher than the level of about 1.5% in the previous business cycle. This is consistent with the trend of productivity improvement observed by the authors.

What needs to be distinguished, however, is that the increase in macroeconomic labor productivity is not equal to the contribution of AI has been confirmed. Capital investment, labor allocation, technological advancements, and cyclical factors may all influence this indicator, and it is currently impossible to directly calculate how much profit AI has created for the enterprise. As a result, Mamaysky proposed a further standard of verification: productivity improvements ultimately need to be reflected in corporate profits outside of the technology industry. A closed commercial loop of AI investment can only be formed if non-tech companies can prove that they have achieved cost savings or revenue growth through AI applications. Otherwise, the current boom may just stop at the stage of “selling shovels,” and “gold diggers” have yet to find enough gold to pay for shovels.

This cautious approach to the causal relationships behind the data reflects the calm thinking of professional institutions in a frenzied market.

Based on this judgment, QuantStreet fine-tuned its investment strategy in an attempt to find a more balanced risk return in an uncertain environment. Despite the poor performance of value stocks and low volatility stocks in the previous quarter, the agency maintained a relative overallocation of these two types of stocks, hoping to retain exposure to the wider corporate sector to seize potential opportunities for a return to the mean.

At the same time, in portfolios with high risk tolerance, some technology stock investments will continue to be retained to share the dividends of AI infrastructure construction. In the fixed income market, the agency has also begun to adjust the duration (Duration). Long-term is used to measure how sensitive bond prices are to changes in yield. The author believes that when the 10-year US Treasury yield reached about 5.25%, the potential investment appeal of bonds has begun to improve. As a result, QuantStreet slightly increased the tenure of bonds in the low-risk portfolio, which is the agency's most obvious change in direction over the past year.

However, this does not mean that the agency is already broadly bullish on long-term debt. Its model is still not optimistic about high-term fixed income assets. The overall long-term term of the portfolio is still below the benchmark; it's just that the level of low allocations has narrowed. Furthermore, for suitable investors, the author also mentioned the decentralized allocation effect of alternative assets such as Evergreen Private Equity Funds (Evergreen Private Equity Funds). According to the performance of some of the products listed, related funds rose by about 0.5% to 0.75% in September, providing a certain portfolio diversification effect during the month when most stocks were under pressure.

However, such products still have limitations such as valuation frequency, liquidity, and underlying asset risk, and monthly returns do not prove their long-term defense capability.

What we really need to watch for in the future are three sets of key signals that will determine whether the AI investment narrative can continue. First, is whether AI infrastructure spending is sustainable, and whether upstream vendor revenue growth is still supported by strong enough demand. Second, is AI starting to improve the profitability of non-tech businesses. Productivity data can provide early clues, but corporate profit margins, cost savings, and new revenue are more direct evidence that return on investment can be realized. Finally, is the rise in long-term US bond yields more reflective of economic growth expectations, or is it inflation, fiscal supply, and maturity risk compensation.

If growth does not improve as expected and financing costs continue to be high, corporate return on investment will be under greater pressure. The question that AI transactions now really need to answer is not only how long the demand for chips and computing power can grow, but how much additional profit these investments will eventually generate for the entire economy. Every dollar bond issued by the US Treasury corresponds to a financing act in the real economy. If these funds cannot be converted into widespread productivity increases, then the accumulation of debt under high interest rates will become the sword of Damocles hanging over the top of the market. Only when AI's productivity gains are gradually transformed into real profits outside of the tech industry can the market obtain more complete evidence to support current large-scale investments. Otherwise, this growth spree driven by capital expenditure will eventually face a severe test in the reality of profit mismatch.