Editor's note: In September, the U.S. bond market experienced a round of severe selling. The 10-year U.S. Treasury yield rose from 4.75% to 5.29%, up 54 basis points in a single month, and some fixed-income assets fell by 2.3% to 5%. But if you break down the change in yields, you will find a fact that is easily overlooked: during the same period, the 10-year real interest rate rose by about 49 basis points, while the market-implied inflation compensation rose by only about 5 basis points. In other words, this round of bond market selling is not mainly explained by rising inflation expectations; what truly changed substantially was the real interest rate.

This raises a more complex question: why is the market willing to demand a higher real return? Usually, a rise in real interest rates may be related to stronger economic growth expectations, adjustments in the future path of monetary policy, or an increase in the risk compensation for holding bonds for the long term. What is even more unusual is that while U.S. Treasuries fell sharply, AI-related technology stocks still performed strongly, with the semiconductor sector leading gains, while most other stocks came under obvious pressure.

Harry Mamaysky, founder of QuantStreet Capital, proposed an explanation in his latest monthly investment letter: the market may be repricing for stronger future economic growth, as well as the capital demand generated by large technology companies' continued expansion of AI infrastructure. Under this framework, higher interest rates and strong AI stocks are not necessarily contradictory, because investors may believe that future earnings growth will be enough to offset the impact of higher financing costs and discount rates.

But this is still a market explanation that remains to be verified. A rise in real interest rates does not equal a certain improvement in the growth outlook, and changes in the term premium also mean that investors may be demanding higher risk compensation. For both the stock and bond markets, the key is to distinguish: does this round of rising interest rates reflect higher future economic returns, or does it mean that holding long-term assets has become more expensive and dangerous?

The following is a translation of the original text:

In September 2026, a rather unusual set of trends appeared in U.S. financial markets: U.S. Treasuries suffered a sharp selloff, while technology stocks, especially the semiconductor sector, continued to rise.

Performance of major assets in September 2026: technology and momentum strategies were relatively strong, while many types of bonds and rate-sensitive assets came under pressure.

According to U.S. Treasury data, the 10-year U.S. Treasury yield rose from 4.75% at the end of August to 5.29% at the end of September, up 54 basis points in one month. QuantStreet Capital statistics show that some U.S. fixed-income assets fell by 2.3% to 5% that month.

But the stock market did not experience a synchronized, broad decline. Bitcoin and momentum strategies with important holdings in semiconductor and technology companies performed prominently, and the Nasdaq index remained up. At the same time, U.S. small- and mid-cap stocks, the equal-weighted S&P 500 index, and rate-sensitive sectors such as real estate investment trusts, utilities, and financials were sold off.

Usually, a sharp rise in long-term interest rates raises corporate financing costs and lowers the discounted value of future profits, which is especially unfavorable for highly valued stocks. But the divergence shown by the market this time suggests that investors may be pricing the future returns of different assets very differently.

Harry Mamaysky, founder of QuantStreet, believes that to understand this round of market action, one must first answer a question: what exactly is the bond market trading?

I. U.S. Treasury yields soared in September; what truly rose was the real interest rate

From the perspective of the yield structure, an important feature of the September U.S. Treasury selloff was that the rise in nominal interest rates mainly corresponded to a rise in real interest rates, rather than a simultaneous surge in inflation compensation.

To understand this, one must first distinguish three concepts.

Nominal Yield is the Treasury yield that investors usually see, which includes compensation for future inflation as well as required real returns and other risk factors.

Real Yield can be understood as the yield after deducting inflation compensation. In the U.S. Treasury market, the yield on Treasury Inflation-Protected Securities (TIPS) is usually used to observe the market's pricing of real returns.

The difference between the two is the Breakeven Inflation Rate, often used as a reference indicator for the market's long-term inflation expectations. However, it also includes factors such as inflation risk and liquidity, and is not equivalent to a pure inflation forecast.

The change in U.S. Treasury yields in September can be further broken down. The 10-year U.S. Treasury nominal yield rose from 4.75% on August 31 to 5.29% on September 30, up 54 basis points. Among this, the real yield rose from 2.44% to 2.93%, up 49 basis points; implied inflation compensation rose only from 2.31% to 2.36%, up 5 basis points.

Performance of major U.S. bond ETFs in September: long-term Treasuries and some credit bonds came under obvious selling pressure.

This means that of the increase in the 10-year U.S. Treasury yield in September, more than 90% corresponded to a rise in real interest rates, rather than an increase in inflation compensation. At least from the breakdown of this market indicator, the main change in the September bond selloff was not in inflation compensation, but in the real interest rate.

This is especially important because rises in real interest rates and inflation compensation often correspond to different economic explanations.

If the rise in yields mainly comes from inflation compensation, it may mean that investors are worried about future price increases and a decline in the purchasing power of money, and therefore demand higher nominal returns. But if the rise mainly comes from real interest rates, then one must further consider economic growth expectations, the future real policy rate, and the risk compensation investors require for holding long-term bonds.

This does not mean that inflation risk has disappeared. The inflation level itself remains elevated, and oil prices, fiscal policy, and the Federal Reserve's interest rate path will also affect market expectations. But from the yield changes in September, explaining this round of U.S. Treasury selling solely by "intensifying inflation concerns" is clearly insufficient.

II. It is not simply inflation worries; the market may be repricing economic growth and capital demand

Why did real interest rates rise significantly? Mamaysky discussed several market explanations in his investment letter.

The first is that dollar credit is being questioned. But the dollar actually appreciated in September, which does not fit the narrative of a broad crisis of confidence in dollar assets.

The second is that investors have begun to worry about the U.S. government's ability to repay its debt. The author believes that if the market's concerns about U.S. fiscal credit were mainly reflected through future inflation risk, then long-term inflation compensation should have risen more noticeably, and the September data did not show this characteristic.

However, this does not rule out fiscal risk. An increase in Treasury supply and a rise in the risk compensation for holding bonds may also push up long-term yields while inflation compensation remains relatively stable.

By contrast, Mamaysky prefers to focus on another explanation: the market may be expecting stronger economic growth while repricing the growing capital demands of large technology companies.

The key variable here is AI. As investment in artificial intelligence infrastructure continues to expand, large cloud service providers (Hyperscalers) are investing enormous amounts of money to build data centers, purchase GPUs, expand computing power, and deploy power and network infrastructure.

These expenditures first mean demand for capital.

From a macro perspective, if companies want to expand investment at the same time, while the long-term funds available for allocation do not increase in tandem, the price of funds may face upward pressure. At the same time, if investors believe that AI will raise future economic productivity and create more corporate profits, they may also raise their required long-term real rate of return accordingly.

Both forces may be related to the rise in real interest rates, but the mechanisms are not exactly the same: the former emphasizes capital demand and financing conditions, while the latter emphasizes expectations of future economic returns.

Recent research by ING also points in a similar direction, arguing that AI's impact on bond yields comes not only from technology companies borrowing to finance themselves, but may also be reflected in real interest rates through productivity and long-term economic growth expectations.

However, this type of judgment still belongs to market analysis, rather than a confirmed causal relationship. A rise in real interest rates itself cannot prove that AI is driving faster growth in the U.S. economy, nor can it prove that AI financing demand is the dominant factor behind the Treasury selloff.

For Mamaysky, the appeal of this explanation mainly comes from the reaction of the stock market.

If the sharp rise in Treasury yields fully reflected a deterioration in the economic outlook, then the stock market would usually also come under broader pressure. But in September, AI-related stocks such as semiconductors remained strong, showing that investors at least remain optimistic about the long-term growth of some technology companies.

This leads to a more interesting relationship between the bond market and the stock market: higher real interest rates may be being priced by the market at the same time as higher future earnings expectations.

III. Why can AI stocks still rise when real interest rates are higher?

From the perspective of traditional valuation logic, a rise in long-term real interest rates is usually not good news for growth stocks.

Stock prices essentially depend on the value of future cash flows after discounting. The higher the return rate the market demands, the lower the value of a company's future profits converted into today's terms. For growth companies whose profits are mainly concentrated in the future, this effect is usually even more pronounced.

But the performance of AI stocks in September shows that the market may be betting on another force.

In September, the semiconductor ETF (SMH) rose about 9.4%, while the equal-weighted S&P 500 index (SPW) fell about 4.8%, showing a clear divergence between the technology sector and the broader market.

Mamaysky understands this as a "battle between numerator and denominator" in valuation: a rise in the discount rate pushes up the denominator, putting pressure on valuations; but expected future profit growth pushes up the numerator, which may offset part or even all of the negative impact. In other words, the market is not necessarily ignoring high interest rates, but may believe that the future earnings growth brought by AI is enough to cover higher capital costs.

In September, momentum ETFs with important holdings in companies such as AMD, Micron, Intel, Cisco, and Applied Materials performed strongly, and semiconductors continued to be an important force driving the market higher.

This trend has a certain fundamental logic. AI infrastructure construction first requires chips, servers, and related equipment, so upstream suppliers in the industrial chain can obtain orders and revenue earlier.

But this also leads to the issue the author is most worried about. Semiconductor stocks continued to rise, while the equal-weighted S&P 500 index performed weakly. This shows that there is a clear divergence between capital's expectations for the profit prospects of AI infrastructure suppliers and its expectations for the broader corporate sector.

Mamaysky calls the broad group of companies outside semiconductors ROCS (Rest of the Corporate Sector), that is, the rest of the corporate sector.

In his view, the reason companies buying AI chips are willing to invest large amounts of money is that they believe they will be able to obtain economic returns in the future through productivity improvements. The market is also willing to provide financing in advance for these profits that have not yet been realized.

Therefore, it is not surprising that not all industries are growing in sync at this stage. What is truly puzzling is that the stock market itself is forward-looking. If investors were already convinced that AI would significantly improve the future profitability of other companies, then these expectations should also gradually be reflected in the valuations of the relevant companies.

But in September, the market did not show this broad-based rise. Chip suppliers are already making money, but the companies buying chips have not yet generally achieved corresponding profit improvements. This means that the current AI trade has a commercial loop that needs to be verified: the revenue obtained by upstream companies ultimately needs to be supported by economic value continuously created by downstream companies. If AI cannot create enough profits for the broader corporate sector, then continuously increasing chip procurement, data center construction, and financing costs may gradually erode investment returns.

Mamaysky does not believe that AI has already formed a bubble. He still believes in the long-term economic value of AI, but thinks the market needs to see more evidence that the benefits of AI are spreading from the technology industry to other companies.

Labor productivity data released by the U.S. Bureau of Labor Statistics has already shown some positive signs, with productivity growth in recent years higher than the long-term average since 2010. But this improvement cannot all be attributed to AI, and even less can it directly prove that companies have already obtained additional profits sufficient to cover investment costs. From this perspective, the bond market and the stock market are actually waiting for the same answer: can future economic growth deliver the returns currently priced in advance?

Historical changes in labor productivity in the U.S. nonfarm business sector.

IV. Rising real interest rates are not necessarily a positive, and term premium is another risk

Understanding the rise in Treasury yields as the market becoming more optimistic about economic growth can indeed explain part of the performance of asset prices. But this explanation still has an important limitation: a rise in real interest rates does not fully equal an improvement in future economic growth expectations.

Long-term government bond yields not only reflect investors' expectations for future short-term interest rates, but also include the term premium, that is, the additional compensation investors demand for bearing risks such as long-term bond price volatility.

The term premium can reflect interest rate uncertainty, fiscal supply, market supply and demand, and other risk factors. Even if inflation compensation has not risen significantly, long-term yields may still rise as long as investors are unwilling to lock up funds for the long term and demand higher risk compensation.

This distinction is especially important in the current market.

A Reuters market analysis on October 7 pointed out that the term premium on the U.S. 10-year Treasury has risen to about a 12-year high. This shows that the rise in long-term yields may not only include economic growth expectations, but may also reflect investors' reassessment of fiscal policy, monetary policy, and the risks of holding bonds for the long term.

It should be emphasized that real interest rates and term premium are not two independent indicators that can simply be added together. The TIPS real yield itself may also include a real term premium. Therefore, the roughly 49 basis point rise in real interest rates in September does not mean that all 49 basis points came from stronger growth expectations.

Two different drivers may have different effects on asset markets. If the rise in real interest rates mainly reflects improved economic growth expectations, then companies' future profits may rise in tandem, and some stocks can withstand higher discount rates.

But if the rise in real interest rates and long-term yields comes more from term premium, then companies may face continuously rising financing costs without a corresponding improvement in future earnings. At that point, high interest rates will put more direct pressure on stock valuations, bond prices, and corporate investment.

This is also why one cannot conclude solely from the rise in AI stocks that this round of Treasury selling is necessarily a positive signal for economic growth. For QuantStreet, the current market does not yet have enough evidence to support a full shift into any one asset.

The institution remains relatively overweight in value stocks and low-volatility stocks, hoping to retain exposure to the broader corporate sector, while continuing to hold some technology stocks in portfolios with higher risk tolerance.

Bond allocation has also begun to see slight adjustments. Mamaysky believes that when the 10-year Treasury yield reaches about 5.25%, the potential attractiveness of bonds has already increased somewhat. Therefore, QuantStreet has begun to moderately extend duration in low-risk portfolios, that is, increasing bond allocations that are more sensitive to interest rate changes.

But this does not mean that the institution has fully turned bullish on long-term bonds. Its model still does not favor longer-duration assets, and overall bond duration remains below the benchmark, only the underweight has narrowed somewhat.

The author also mentions that for suitable investors, some alternative assets such as certain evergreen private equity funds may provide some diversification, but the liquidity and valuation risks of the related products still need to be considered separately.

These adjustments reflect a cautious attitude: long-term yields have begun to have some appeal, but there is no clear answer yet as to whether the forces causing yields to rise have faded.

Next, the market needs to watch three types of signals: first, how long-term real interest rates and term premiums change, in order to distinguish between growth expectations and risk compensation; second, whether AI investment is beginning to genuinely improve the productivity, profit margins, and cash flow of non-tech companies; and finally, whether Federal Reserve policy expectations, fiscal financing needs, and long-term Treasury supply continue to exert upward pressure on yields.

If economic growth and corporate earnings continue to improve, high real interest rates and a strong stock market may coexist for a period of time. But if term premiums keep rising while the returns on AI investment fail to materialize for a long time, tech stocks that currently appear able to withstand high interest rates will also face a more severe valuation test.

The most important signal from the September U.S. Treasury selloff is not that inflation expectations are once again out of control, but that the long-term real returns investors demand have clearly risen.

The truly unresolved question is: does this higher return requirement come from confidence that the future economy can generate more profits, or is the risk that must be borne to hold long-term assets increasing?

Both explanations could push up U.S. Treasury yields, but they imply completely different market prospects for stocks, bonds, and the future of the AI investment cycle.