As the narrative surrounding AI investment returns shows signs of strain, investors are beginning to retreat, turning instead to embrace US Treasury bonds. According to trading desk sources, a research report released on February 13 by Morgan Stanley's US interest rate strategy team indicates that market sentiment has shifted. Confronted with the swelling AI investment frenzy and lofty market valuations, investors are pulling back from risk assets and buying US Treasuries as a safe haven.

In the report, while Morgan Stanley raised its 2026 US real GDP growth forecast from 2.4% to 2.6%, citing capital expenditure from "hyperscale" cloud providers as a growth driver, its team of economists also issued a stern risk warning. Morgan Stanley pointed out that this growth does not come without a cost. The logic is very clear: "The more the AI-related capital expenditure cycle boosts economic activity, the greater the risk from an over-investment cycle if the return on investment ultimately fails to materialize."

The market has clearly understood this logic. Investors are growing increasingly sensitive to the negative spillover effects of the AI investment cycle and are no longer blindly chasing high valuations. As Morgan Stanley strategists stated: "Investors are growing weary of new narratives attempting to justify extremely high valuations for risk assets. This is leading to a divergence in market sentiment and a shift towards US Treasuries, pushing yields lower so far this year."

Cracks are appearing in the market. Although the S&P 500 index had been hitting new highs, stocks of companies whose business models are being 'disrupted' by the wave of AI application have already begun to decline sharply. Morgan Stanley constructed a basket of 108 stocks impacted by AI. Data shows that these stocks decoupled from the broader market as early as the end of last year and continued to trend lower even as the main indices reached new peaks.

This divergence sends a dangerous signal: the narrative of optimism surrounding AI may have peaked. Beyond the directly impacted stocks, the Software-as-a-Service (SaaS) sector is also under significant pressure, while a more hidden risk lies in private credit. Morgan Stanley warns that due to the opaque and lagging nature of fundamental indicators in private credit portfolios, the stock performance of alternative asset managers with private credit exposure in the public market serves as a real-time risk barometer. Data shows a basket of seven key alternative asset managers faced substantial downward pressure at the start of the year. The market is voting with its feet, avoiding potential credit default risks stemming from AI over-investment.

Weakness in the stock market is transmitting to confidence in the real economy, particularly among high-income groups. Morgan Stanley observed that upper-income households (annual income over $100,000), who are the main drivers of consumption, have shown a significant shift in their perception of the current economic situation since the beginning of the year. "Although starting from a better level than at the end of 2021, this shift in sentiment looks very similar to early 2022—a period after which the US economy subsequently experienced two quarters of negative growth."

Why is high-income confidence falling? The reason points directly to asset price volatility. Morgan Stanley economists believe that "the bursting of an asset investment bubble would pose a greater risk to the economy." When affluent households start tightening their belts, it is often a reliable leading indicator of an economic downturn. In this environment, US Treasury bonds have once again become the premier asset class for hedging recession risk. Morgan Stanley's strategy team stated plainly: "We believe that with market-implied policy rates still pricing in almost no downside risk premium, US Treasuries as a hedge look attractive."

If concerns about an AI bubble provide the long-term rationale for buying US Treasuries, the latest inflation data served as a direct catalyst. January's CPI figures came in lower than expected across the board, dealing a blow to the "stubborn inflation" narrative. Headline CPI increased 0.17% month-over-month, below the 0.27% expected by economists. Core CPI rose 0.30% month-over-month, also slightly below the 0.31% expectation. Underlying data, which excludes extreme volatility, is even more noteworthy. The Cleveland Fed's Trimmed Mean CPI and Median CPI both increased by only 0.19% in January. "These readings are the lowest for the comparable period since 2021. Given that January is typically the strongest month for sequential data in the calendar year, we believe this downside surprise will carry significant weight as investors consider full-year inflation figures."

The weak inflation data directly reshaped expectations for Federal Reserve policy. The market reacted swiftly, with traders beginning to price in a lower terminal rate. Current market pricing suggests 21 basis points of rate cuts by the June meeting, accumulating to 62 basis points of cuts by the end of 2026. Morgan Stanley believes that receding inflation expectations will allow the Fed to ease policy further to prevent real interest rates from becoming overly restrictive. This implies that US Treasury yields have further room to decline. Simultaneously, the Fed's bill purchase operations are providing liquidity support to the market. Over the past 14 operations, the Fed has purchased $109.2 billion in bills, with each operation reaching maximum capacity (100% uptake rate), maintaining a loose environment in funding markets and further benefiting short-term US Treasuries.