While the global semiconductor sector has already launched a major counterattack in stock markets, the credit market continues to raise financing thresholds for certain AI-related technology companies. On October 23 and 26, the Philadelphia Semiconductor Index rose 0.44% and 2.82% respectively, accumulating a rebound of approximately 3.27% over the two days. The Nasdaq 100 Index had fallen about 6.24% cumulatively over the two weeks ending October 23, meaning that after posting its worst two-week decline since April 2025, the Nasdaq finally rebounded significantly by 1.09% on October 26. However, based on the closing high on October 6, the Philadelphia Semiconductor Index and Nasdaq 100 remain approximately 16.15% and 6.25% lower respectively. The rebound is repairing pessimistic pricing, but there is still considerable room before losses are fully recovered.
The divergence in the global corporate credit bond market is even more striking. Just as Wall Street financial giants including Morgan Stanley and Goldman Sachs continue to take a bullish view on the stock market, recent market conditions have precisely demonstrated the tension of this semiconductor-led technology stock rally alongside the capital fragmentation and divergence displayed in the corporate credit bond market. From September 25 to October 26, the five-year credit default swap spread for CoreWeave surged from 835 to 1,148 basis points, while Meta rose from 98 to 121 basis points, and Microsoft only edged up from 52 to 55 basis points. Under a 40% recovery rate assumption, CoreWeave corresponds to an approximately 62% five-year cumulative implied default probability, though this metric includes a risk premium and cannot be regarded as an actual default forecast.
During intraday trading on October 26, the five-year credit default swap spread for CoreWeave (CRWV.US), the most central target in the "AI computing power selloff storm" and "AI bubble burst theory," stood at approximately 1,148 basis points. Under a 40% recovery rate assumption, the model estimates a corresponding five-year cumulative implied default probability of approximately 62%, though this metric also heavily incorporates credit risk premiums and cannot be directly equated with the company's actual default probability over the next five years. On the cash bond side, data compiled by media as of October 22 showed that CoreWeave's bonds maturing in 2031 yielded 15.76%, SoftBank's five-year CDS rose to 432 basis points on October 20, and its dollar bonds issued in September carried a coupon as high as 9.75%. These quotes across different dates and instruments collectively illustrate that while the stock market can trade earnings recovery in advance, institutional creditors in the credit bond market continue to question whether the capital needed for the AI computing infrastructure expansion boom can be sustained amid geopolitical energy inflation risks and soaring AI financing costs driven by elevated US Treasury yields.
According to the latest "GOAL Kickstart" research report from Wall Street financial giant Goldman Sachs, which focuses on financial market risk appetite and equity-bond allocation stance, the technology-driven risk repricing has significantly reduced market crowding, but credit pressures have not yet been resolved. Earnings support, declining valuations, and reduced positioning are still sufficient to support the Goldman Sachs financial market strategist team's "mildly pro-risk" asset allocation stance, which leans toward optimistic forecasts for both equity and government bond markets. The report places Goldman Sachs' proprietary Risk Appetite Indicator at -0.5, declining credit excess returns, rebounding major government bonds, pressured European assets, and quality and defensive style leadership within a single cross-asset framework, cross-validated through valuation, earnings revisions, fund flows, volatility, liquidity, and recession models. Accordingly, Goldman Sachs maintains an "overweight" on equity assets at both 3-month and 12-month horizons, while being "underweight" on the credit bond market, "overweight" on cash in the short term, and "overweight" on government bonds at the 12-month horizon. These views and allocation stances from Goldman Sachs significantly support "selectively adding equity exposure on dips," though the report does not declare, as it did during the April 2025 US stock market crash, that the market has reached a definitive bottom and is entering a new bull market.
The Goldman Sachs research report shows that semiconductors are leading the counterattack, while the AI credit market is still repricing financing costs and cash flow realization timelines, creating a striking tension in global capital flows where "equities seek repair while credit continues to stratify." The Goldman Sachs Risk Appetite Indicator has fallen into negative territory but has not yet reached the stronger contrarian buying zone seen historically; however, resilient technology earnings expectations, declining valuations, and reduced position crowding have already improved expected equity returns. This is the underlying basis for Goldman Sachs maintaining its overweight on equities at 3-month and 12-month horizons. The opportunity to add positions depends on the relationship between earnings and purchase prices, and whether the market's next leg can stabilize will depend on whether technology giant earnings, AI financing, and central bank policies can improve together.
Furthermore, as Meta Muse and OpenAI's GPT-6 Astra fully ignite the AI agent wave, "AI FOMO trading"—referring to investors' fear of missing out on AI-related asset gains, prompting them to chase or rebuild previously reduced technology stock allocations—has made a comeback. From the perspective of the underlying architecture of AI computing infrastructure, Muse and Astra are expected to further expand the scope of AI agent usage. The massive expansion of AI inference workloads they bring will simultaneously increase demand for model inference, tool execution, and state management, while deepening competitive pressure on some software companies. The AI capital expenditure forecast and outlook from JPMorgan's strategist team shows that hyperscale cloud providers' capital expenditures will be approximately $950 billion in 2026, approximately $1.4 trillion in 2027, and approximately $3 trillion in 2030.
For the epicenter of the global AI trade and technology stock selloff—the memory chip sector that has recently suffered heavy losses—bullish sentiment and fundamental optimism remain strong. Recent KeyBanc Asia survey data shows that AI data centers are still absorbing memory and other AI computing resources on a broad scale, pushing price increases through to 2027. The institution predicts that HBM blended prices could rise by 80% to 100% in the first quarter of 2027. On the DRAM memory price front, KeyBanc forecasts a 15% to 20% quarter-over-quarter increase in the fourth quarter of 2026 on top of a strong third-quarter price increase base, followed by another 10% to 15% increase in the first quarter of 2027. For NAND future pricing, it forecasts approximately 20% quarter-over-quarter increase in the fourth quarter of 2026 on a strong third-quarter base, followed by another 10% to 15% increase in the first quarter of 2027.
Risk Appetite Falls Below Zero: Credit Market Optimism Retreats First, Semiconductor Sector Reasserts Grip on Global Pricing
The Goldman Sachs research report shows that Goldman Sachs' proprietary Risk Appetite Indicator (RAI) has fallen to approximately "-0.5," hitting its lowest level since April 2025. Historically, only readings below -1, especially those approaching -2, have formed stronger contrarian signals. However, Goldman Sachs strategists emphasize that the more clearly positive change trajectory currently is that position constraints on risk asset gains have eased compared to late summer. The deterioration in the indicator this time mainly came from the credit sub-component, which also dragged down the "global growth" factor. Rising government bonds in major economies improved the "monetary policy" factor, providing some buffer against the AI stock selloff. Widening French and German government bond spreads also made European-related assets perform weaker than growth and policy factors could explain. The relative strength of quality, defensive, and value styles cushioned the impact of the technology momentum trade retreat on the broader market. The decline in credit excess returns here refers to deteriorating credit performance after stripping out matched government bond returns, and does not mean that all corporate bond prices are simultaneously crashing.
The global technology stock factor, in which the semiconductor sector occupies the vast majority (over 50%), has become the absolute center of global cross-asset pricing—meaning the semiconductor sector has reasserted its grip on the global capital market pricing system, with macro capital flows and cross-asset pricing frameworks all taking their cues from semiconductors. Goldman Sachs has provided very direct chart-based statistical evidence for this trend: the short-term correlation between the RAI and "Nasdaq 100 relative to S&P 500 equal-weight index performance" has recently approached 1. The cross-sectional fit between asset returns and semiconductor correlation over the past three weeks is approximately 54%, while the corresponding fit for oil prices is only about 2%. The US dollar has also recently exhibited stronger pro-cyclical characteristics. The options market is concentrating risk pricing on the next two weeks: technology stocks are more focused on this week's earnings reports from Microsoft, Google parent Alphabet, and Amazon, while long-term US Treasuries and high-yield credit are more focused on next week's US midterm elections. This week's Federal Reserve, Bank of Japan, and European Central Bank decisions are equally critical. Goldman Sachs economists' baseline remains a Fed pause in October and a rate hike in December, so the recent rebound logic in global bond markets, especially long-duration government bond markets, is based more on slightly easing oil prices and incremental capital returning due to persistently surging yields, rather than being directly interpretable as the start of a comprehensive rate cut trade.
Overweight Equities, Underweight Credit Bonds: Goldman Sachs Bets on Technology Earnings Growth Resilience and Valuation Repair
Goldman Sachs' allocation has clear layers: for equity assets, it presents the most optimistic "overweight" stance at both 3-month and 12-month horizons; for credit bonds, it presents an "underweight" stance across the board; cash is overweight in the short term, government bonds are overweight at the 12-month horizon, commodities are overall neutral, and gold is overweight. Within equities, it prefers the US, Japan, and Asia-Pacific excluding Japan, while being underweight Europe. Within credit, it prefers US dollar investment-grade and high-yield bonds while being underweight euro credit, which is not contradictory to the overall credit underweight. Goldman Sachs strategists' S&P 500 index targets at 3, 6, and 12 months are 8,000, 8,300, and 8,700 points respectively, with expected total returns of 5.2%, 9.4%, and 15.3% including dividends. The 12-month expected total return for Asia-Pacific excluding Japan is 29.6%. Cross-asset forecasts also include a 12-month US 10-year Treasury yield of 4.44%, gold at $5,235, and Brent crude at $75. Goldman Sachs' portfolio logic is to continue participating in the semiconductor-led technology stock earnings growth trajectory while retaining cash, government bonds, and gold to hedge against event shocks.
South Korean memory chip stocks and the KOSPI index, where the two memory chip giants SK Hynix and Samsung Electronics carry a 50% weight, provide a prominent example of this expectation gap. Another Goldman Sachs South Korea market strategy report lowered the KOSPI index target from 12,000 to 9,000 points, but as of the report date over the weekend, this still corresponded to over 60% upside potential. The Goldman Sachs Korea strategy team believes that semiconductor valuations implying a 65% earnings decline in 2027 are overly pessimistic, with Samsung Electronics and SK Hynix trading at only about 3 times forward price-to-earnings at the time. The KOSPI has fallen about 40% from its June high, while forward 12-month earnings forecasts rose about 22% over the same period. The Goldman Sachs Korea team noted in its research report that both Samsung Electronics and SK Hynix are currently trading at approximately 3 times forward P/E, the lowest level in history. Based on the 72% goodness of fit (R²) relationship between price-to-book ratio and return on equity revealed by the past 20 years of monthly data compiled by Goldman Sachs, the current 2.6 times price-to-book ratio for the South Korean memory chip sector implies a 2027 ROE of only 23%, while market consensus forecasts an ROE as high as 45%, demonstrating that the market is giving severely discounted valuations to South Korean memory chip giants.
This divergence is equally prominent at the index level: the KOSPI has fallen about 40% from its June high, while forward 12-month earnings forecasts rose about 22% over the same period. The KOSPI's record-low forward P/E of approximately 4 times is at an extreme low among the 18 trillion-dollar major markets covered by relevant statistics. However, the low P/E of the memory industry may also contain concerns about the unsustainability of cyclical peak earnings. The key to Goldman Sachs' bullish judgment on South Korea's two memory chip giants and the KOSPI benchmark index is that as long as the actual earnings downgrade is smaller than the pessimism implied by stock prices, combined with a decline in risk premiums, the South Korean stock market and the two memory chip giants that account for 50% of South Korean market capitalization could achieve strong valuation repair during a period of slowing earnings growth.
Global capital has also begun to re-bet on the South Korean stock market and memory chips. Fund flow data compiled by Goldman Sachs shows that from October 1 to 23, South Korean retail investors net bought 315.4 billion won of leveraged products related to the two companies. As of October 22, South Korean investors net bought approximately $3.548 billion of SOXL (the triple-leveraged US ETF tracking the Philadelphia Semiconductor Index with ticker "SOXL"), accounting for a staggering 73.9% of net US stock purchases during the same period. The main logic behind Goldman Sachs' trading opportunities in the South Korean stock market, memory chips, and semiconductor sector is that actual earnings performance will be better than the pessimistic scenario implied by stock prices, and the return of leveraged capital reflects that this judgment is being converted into positions.
The latest assessment from Morgan Stanley's Chief US Equity Strategist Wilson supplements the time dimension of the rebound: AI capital availability and costs, AI safety and data center construction obstacles, midterm elections, and Federal Reserve liquidity are three groups of factors that need to be digested in the coming weeks. He maintains his optimistic S&P year-end target of 8,000 points, emphasizing that "the internal correction in the US stock market that began in June has entered its final tail phase, but the index may still experience one last pullback," while retaining a tactical path of first pulling back to 7,100-7,200 points, viewing liquidity as both the biggest risk and a potential catalyst for forming the final low. Goldman Sachs focuses on "improved expected returns for equities after the correction," while Wilson focuses on "which events can end the correction." Clearly, both support investors preparing equity exposure for year-end market conditions, but Goldman Sachs' "8,000 in three months" and Wilson's "8,000 by year-end" are different forecast horizons.
Why Global Financial Markets Are All "Taking Their Cues from the Semiconductor Sector": AI Connects Earnings, Capital Demand, and Financing Costs
Understanding through the industrial mechanism, technology stocks have become the market's "anchor" because AI simultaneously affects corporate earnings, real investment, and capital supply and demand. After agent applications expand, GPUs and application-specific integrated circuits (ASICs) handle model computation, CPUs run tools and task orchestration, high-bandwidth memory (HBM) and server DRAM support models and concurrent environments, enterprise SSDs preserve persistent states, and the entire AI computing infrastructure chain—including networking, power, data center optical interconnect systems, high-performance CPUs, and data center underlying power equipment and cooling systems—collectively ensures the delivery of computing resource services. Muse and Astra, along with Meta's latest open-source Muse Spark Open, are expected to collectively further expand the scope of AI agent usage. The massive expansion of AI inference workloads they bring will simultaneously increase demand for model inference, tool execution, and state management, while deepening competitive pressure on some software companies. GPUs, TPUs, and other accelerators handle model computation; high-performance server CPUs run browsers, virtual machines, product searches, database queries, and transaction orchestration; HBM and server DRAM carry model data, context, and concurrent work environments; enterprise SSDs store product indices, task records, and persistent states; and high-speed networking and optical interconnects support distributed data exchange.
Technology giants' procurement budgets therefore simultaneously correspond to revenue for chip, memory, equipment, and energy companies. Expenditures are realized through cash flow, bonds, equity, and lease financing, which in turn affect credit spreads and capital costs. When the market downgrades AI return expectations, the shock propagates along the customer financing, equipment orders, supplier earnings, and Asian export chains, as well as through global equity, bond, and currency markets and commodity markets. When AI financing costs and commercialization prospects improve significantly, this chain also possesses reverse repair capability. JPMorgan's asset management division similarly points out that current financing disputes revolve more around capital costs and delivery returns, with low-rated neoclouds and high-rated cloud giants having markedly different capacities to withstand pressure.
Goldman Sachs remains willing to take on equity risk, ultimately supported by three data points: the S&P 500's forward 12-month P/E has declined to just 18.6 times, below the ten-year average of 19.2 times; Asia-Pacific excluding Japan is at only 9.5 times; and Goldman Sachs' estimate of US recession probability over the next year is 15%. Technology earnings forecasts are still being revised upward, although the breadth of upward revisions has cooled. In comparison, US investment-grade and high-yield bond spreads stand at 89 and 337 basis points respectively, still below their respective ten-year averages of 109 and 370 basis points. Higher absolute bond yields do not necessarily mean that credit risk compensation is already sufficient. From this perspective, the underlying basis for Goldman Sachs' support of buying on dips is that strong earnings growth resilience, declining purchase valuations, and significantly reduced position crowding have collectively improved expected equity market returns. The risk appetite falling below zero is only background, not a sufficient condition on its own to trigger buying.
Comments