A confluence of surging AI-related corporate bond issuance, intensifying debate over a potential "AI credit bubble burst," and climbing yields on U.S. Treasuries with maturities of 10 years or longer is triggering a classic "discount rate shock coupled with shrinking risk appetite" across global technology stocks centered on the AI computing trade.

Data compiled by Wall Street financial giant Goldman Sachs reveals that the information technology sector—home to the majority of tech stocks—was the most heavily net-sold industry in the U.S. stock market last week, registering its largest gross exposure reduction in over two years. Crucially, the selling originated almost entirely from rapid institutional long-position reductions rather than an expansion of short positions, signaling that Wall Street institutions are actively locking in profits, reducing portfolio concentration in AI, and rotating toward classic safe-haven sectors to mitigate the risks posed by surging yields.

This positioning shift more closely resembles a phase of deleveraging and crowding clearance within the AI trade rather than a wholesale rejection of the long-term demand narrative for AI computing power and AI application software. However, should long-dated Treasury yields and credit spreads for technology companies continue accelerating higher, AI-related concept stocks—which depend on external financing, rely on massive distant future cash flows, feature lengthy capital return cycles, and remain highly sensitive to fluctuations in the risk-free rate on the denominator side—will continue to face pressure.

Conversely, leaders in the AI computing supply chain possessing robust free cash flow, clear visibility into computing orders, and formidable technical barriers that are difficult to replicate, along with key cloud-based AI inference computing providers, may be first in line for capital reallocation once institutional crowded positions and leverage are cleared. The market is thus transitioning from a broad-based rally in AI computing themes toward fundamental stock selection.

Not Short-Seller Aggression but Long-Seller Exit: AI Conviction Questioned by Bond Yields, Tech Stocks Become the Largest Net-Selling Sector

Goldman Sachs data indicates that the information technology sector was the dominant theme in terms of net selling volume across the U.S. stock market last week, and this substantial selling was driven almost entirely by long-side selling pressure. Goldman Sachs senior analyst Robbie Stancard noted that the sector's total allocation exposure also recorded its largest percentage decline in over two years, highlighting investors' significant reduction in risk allocation to the sector.

The chart compiled by the firm shows that the total exposure of the information technology sector stands at approximately 20% of total U.S. market exposure, notably lower than the peak of nearly 24% recorded earlier this year. The net exposure metric is around 19%; during the recent selloff, this gauge fell from a peak of approximately 26% in 2026 to about 16%. Earlier, both indicators had climbed substantially during the first half of the year. Total exposure rose from roughly 18% at the start of 2026 to near 24%, while net exposure once surged from about 17% to 26%, before reversing sharply following the June pullback in South Korean equities and the major selloff in global semiconductors in July.

The analyst team led by Stancard stated that the latest round of selling was driven almost entirely by long-side selling rather than an increase in short positions, indicating that investors are trimming their bullish positioning in popular tech sectors tied to the AI computing theme. In their research report, the Goldman Sachs team noted that the market's investment direction has shifted—but more precisely, it is not a case of capital "completely abandoning the AI computing theme." Rather, it is a transition from a "single high-beta trade in AI computing hardware" toward "deleveraging/reducing crowded positions, rebalancing toward high-cash-flow stocks with historically low concentration, and cross-asset diversification."

Goldman Sachs data shows that global hedge funds, after a period of consecutive bargain-hunting in late July, suddenly reversed to net selling of U.S. equities, with the Net Leverage indicator dropping to a one-year low of 48.3%. Combined with the fact that information technology became the largest net-selling sector, with the selloff almost entirely stemming from long-position reductions, total tech sector exposure has retreated from this year's peak of nearly 24% to approximately 20%, while net exposure once fell from roughly 26% to 16%. This indicates that institutions are compressing crowded positions and portfolio volatility rather than aggressively establishing new short bets against the AI fundamental outlook.

Surge in 10-Year-and-Beyond Long-Dated Yields Combined with Bond Issuance Pressure Reshapes Wall Street Strategy and Investment Direction

The Goldman Sachs analyst team attributed the core trigger for the style rotation to the dual and sustained upward pressure from "long-dated discount rates plus AI financing costs": the 30-year Treasury yield briefly touched approximately 5.33%, while the 10-year approached 4.7%, simultaneously elevating the discounted value of tech companies' future cash flows, the financing costs for data center projects, and the substitutive appeal of Treasuries relative to equities. Since 2026, AI-related bond issuance has reached approximately $220 billion, far exceeding last year's $125 billion. Technology credit spreads have widened to roughly 89 basis points, approximately 9 basis points wider than the overall investment-grade bond universe, indicating that investors are now demanding a higher "AI financing risk premium."

However, AI bond issuance primarily acts as an amplifier of long-end term premium rather than the sole root cause of rising yields—U.S. fiscal deficits, inflation persistence, and Federal Reserve credibility remain the deeper driving forces. The so-called "AI momentum deceleration" referenced by Goldman Sachs in its report primarily refers to a wholesale reshuffling of price momentum factors, which is not equivalent to a peak in AI orders, token demand, or computing capital expenditure. Software has now become the largest weight in the three-month momentum long book, while semiconductors and the composite AI complex have slipped into the short side, reflecting a shift in the primary investment narrative from "which AI supply chain participants capture the largest share of AI capex" toward "who can consistently convert computing power and tokens into robust revenue, earnings per share (EPS), and free cash flow."

Goldman Sachs continues to anticipate that enterprise AI computing infrastructure deployment is accelerating, and that increasingly complex and massive inference processes will keep AI computing constraints tight for cloud giants. Consequently, the AI computing trade is far from over; it is simply bidding farewell to indiscriminate valuation expansion. Cash-rich platform-based cloud leaders, AI application software, and storage chip giants and core data center infrastructure suppliers whose stock prices clearly diverge from their EPS trajectories still present opportunities. Meanwhile, new-age cloud providers that rely heavily on debt, project financing, and continuous refinancing will face a more stringent balance sheet assessment.

On the cross-asset front, the Goldman Sachs analyst team noted that capital appears to be accelerating the construction of a classic "barbell" portfolio combining "high-quality AI cash flows plus banks plus hard assets." European and Japanese banks benefit from steepening yield curves, "higher-for-longer" rate expectations, and improved net interest income (NII). Physical gold and gold miners capture fiscal credit dilution and dollar-weakening trades, while copper miners benefit from physical supply constraints in the most critical energy infrastructure layer amid grid expansion, data center demand, geopolitical tensions, and superpower competition between East and West.

Goldman Sachs added, however, that investors should not treat all real assets as winners—utilities and real estate continue to face selling pressure due to their bond-substitute characteristics. Should long-end term premia, credit spreads, and AI-related financing supply continue to rise substantially, European and Japanese bank stocks, gold, copper miners, AI application platforms with current cash flows, and high-quality fundamental stocks with historically low position concentration are more likely to assume market leadership.