AI Trading Faces Setbacks Again as Investors Rediscover the Safe-Haven Value of US Treasuries

According to Odaily English Edition, BlackRock's president stated that artificial intelligence (AI) is not in a bubble. U.S. Treasury Secretary Janet Yellen also sees no “systemic issues” in the current wave of prosperity. Nevertheless, regardless of how Wall Street executives try to suppress market concerns about this sector, the market refuses to settle down.

The Nasdaq 100 Index registered its worst two-week performance since last April, while indicators measuring credit risk and market turmoil surged sharply. A basket of AI-related stocks plunged 4.9%, prompting investors to flock into U.S. Treasuries, which pushed the 10-year Treasury yield below 5% for the first time in a month. Safe-haven trades, which had vanished over the summer, have resurfaced.

Overall, these market movements indicate increasing investor skepticism about a trading logic that was almost unanimously accepted by Wall Street: going long large-cap tech stocks while shorting defensive markets pressured by concerns that AI-related spending could reignite inflation. Although previous rounds of market volatility eventually proved manageable, the uptick in U.S. Treasuries this week serves as evidence to Neuberger Berman’s Jeffrey Blazek that a more significant shift might be underway.

Blazek—co-chief investment officer for multi-assets at the $630 billion asset manager—said: “This time, things are moving fast—we’re seeing funds rush toward high-quality assets, a pattern completely opposite to AI trading, as investors buy U.S. Treasuries for safety. This is a full plot reversal.” About a month ago, Neuberger’s Asset Allocation Committee began reducing stock exposure due to worries about overheated AI trades and rising interest rates.

Market Dynamic Shifts

During the latest wave of AI sell-offs, U.S. Treasuries regained their status as safe havens, in stark contrast to the situation in July. This week’s stock market decline saw the S&P 500 fall 0.5%. The tech-heavy Nasdaq 100 dropped 0.8%, a 7.3% retreat from its record highs earlier this month. Meanwhile, the largest ETF tracking long-term bonds rose for a third consecutive week and is on track for its best monthly performance since last February.

This shift in capital is also reflected in ETF flows. According to Baird Strategas data, the gap between demand for equity and fixed-income assets has dropped to the lowest 10% range in history, showing investors’ growing preference for bond exposure.

Much of investors’ unease centers on the vast sums that companies are borrowing and investing in AI, as well as whether this technology can generate returns quickly enough to justify these investments. This week, reports indicated that Anthropic—one of the leading AI developers—may spend up to $750 billion on additional computing power in the coming years.

Meta Platforms (META.US) announced that it will spend about $825 billion over the coming years renting and contracting for computing power and AI infrastructure. The company’s share price dropped over 9% this week, its biggest weekly fall since May; after announcing the plan, a metric measuring Meta’s credit risk hit a record high. Other hyperscale cloud firms and chip manufacturers also saw a sharp rise in debt protection costs.

Meanwhile, global corporate bond yield spreads continue expanding. A key U.S. corporate debt market risk indicator—the CDX Investment Grade Index—rose by 1.5 basis points to 64.4 on Friday, the highest intraday level since March. The index rises when credit risk concerns among investors intensify.

As investors start focusing more keenly on AI’s impact on the economy, a basket of AI stocks has moved in the same direction as 10-year Treasury yields for nine consecutive weeks—the longest synchrony since OpenAI launched ChatGPT. This starkly contrasts with July’s market pattern, when both AI trades and Treasuries lost momentum simultaneously.

Cameron Dawson, chief investment officer at NewEdge Wealth, commented: “As formerly leading stocks waver and rotate toward more defensive sectors, while credit indices send stress signals, investors now have a reason to seek the relative safety of bonds.”

However, even as the bond market continues to emit signs of caution, some well-known figures refute concerns that AI trading poses broader threats. According to the Financial Times, Yellen told investors at a closed-door meeting that she does not see systemic issues in the AI boom. Jon Gray of Blackstone shared similar views, suggesting investors shouldn’t worry about AI trades forming a bubble.

JPMorgan strategists, including Nikolaos Panigirtzoglou, attribute the sharp selloff in AI stocks to so-called VaR shocks. This term refers to asset managers’ risk models triggering forced liquidations.

According to their analysis, investors who built large positions during recent calm months are now being forced to unwind them as volatility suddenly spikes. The team estimates that chip manufacturers account for about one-fifth of global equity market risk capital.

Panigirtzoglou and his colleagues wrote in a report: “VaR shocks do not necessarily require fundamental triggers, as they often reflect position adjustments. Given the recent decline in semiconductor stocks, we suspect the previously excessive long exposure in semiconductors has largely resolved the outstanding risk.”

Nevertheless, the shift toward defensive posturing is apparent in the U.S. ETF market, where demand for fixed-income funds is rising. Data show that so far this month, investors have poured about $58 billion into bond ETFs, accounting for roughly 38% of all ETF flows. This marks the second consecutive month that fixed-income assets have made up over 35% of overall inflows and is the first time since early last year that this has happened two months in a row.

Strategist Athanasios Psarofagis commented: “Bond fund inflows over the past two months have significantly exceeded their usual share. Part of this likely reflects a shift toward defensive assets among investors, but current yields are also attractive enough that some investors may simply wish to lock in income while rates remain high.”