The Zhitong Finance App learned that the US stock market is entering a very decisive week. On the face of it, this is a dense financial reporting window; in essence, it is a simultaneous stress test of the roots of the artificial intelligence (AI) bull market narrative.
Last week, after experiencing the most intense single-day sell-off since June, the Nasdaq Composite Index barely stabilized with a slight decline on Friday. The disappointing performance of Meta (META.US), one of the hyperscale cloud service providers that invested heavily in the construction of data centers, and semiconductor leader Fanlin Group (LRCX.US), which had impressive gains in the current AI market, has made market anxiety public.
Obviously, current investors are no longer willing to unconditionally pay for “AI capital expenses,” and the three largest cloud service providers to be launched this week — Alphabet (GOOOGL.US), Microsoft (MST.US), and Amazon (AMZN.US) — financial reports, the Federal Reserve interest rate meeting, the performance guidelines of the Korean memory chip duo, and PCE inflation data will jointly determine whether the current round of rebound is a “healthy correction” or a “prelude to deeper revisions.”
It's almost a foregone conclusion that the Federal Reserve remains on hold in October, but Walsh's “silent signal” is the real source of risk
This week, the US, Japan, and the European Central Bank will hold successive policy formulation meetings, while the Federal Open Market Committee (FOMC) will announce interest rate decisions on Wednesday local time. As of press release, the CME FedWatch tool shows that the probability that the market price will raise interest rates by 25 basis points in October is only about one-tenth, and the probability that the federal funds rate will remain in the 3.75% to 4.00% range. In an economists' survey conducted from October 16 to 20, 94 out of 100 respondents expected to stay on hold this month. This expectation is highly consistent with market pricing.
However, the October “moratorium” is almost certainly not the end of the austerity cycle. In the same survey, all 100 economists expected interest rate hikes in December, yet less than half held this expectation in last month's survey. According to the FedWatch tool, the market expects the probability of interest rate hikes in December to be about 72%, reflecting that the market already regards an austerity before the end of the year as the benchmark scenario.
David Mericle, chief US economist at Goldman Sachs, wrote in a review: Most Federal Reserve officials tend to “raise interest rates by 25 basis points every other meeting,” which means that December and March of next year will each be a plus. The September bitmap shows that 16 out of 18 officials expect to raise interest rates at least once more during the year.
What is more noteworthy about the ratio resolution itself is Chairman Kevin Walsh's communication attitude at the post-meeting press conference. Since taking office in May 2026, Walsh has been known for refusing to provide “forward-looking guidance” to financial markets, which clearly broke with the previous chairman's communication paradigm.
A team led by Deutsche Bank's chief American economist Matthew Luzzetti said in a research report that the core message that Walsh is likely to convey is that recent data has hardly allayed concerns about continued inflation, and tightening financial conditions cannot replace further interest rate hikes. Luzzetti specifically pointed out that one key observation point is whether Walsh continues to describe overall financial conditions as “loose” — if he avoids repeating the words at the July press conference that triggered the market to question the Fed's will to act, it can be interpreted as an implicit endorsement of the December interest rate hike.
The dilemma facing Walsh is a split in his dual mission: inflation has been above target levels for the fifth year in a row, while unemployment is still hovering near historic lows. If interest rate hikes are too slow, there is a risk that inflation will be undervalued at the beginning of the pandemic; raising interest rates too fast may trigger unnecessary financial tightening in a market already highly sensitive to AI capital expenditure.
The rising energy costs fueled by the Middle East conflict further complicate this trade-off — the Federal Reserve needed time to assess the extent of transmission of the fuel price shock, which was the tactical reason for “waiting” in October, but it was far from sufficient to form a strategic basis for stopping austerity.
Greenwood Capital's chief investment officer Walter Todd said he would pay attention to any hint from Walsh about the interest rate path at the post-decision press conference. “Whether he will give some clues about December because current expectations are that interest rates will rise,” Todd said.
Alphabet, Microsoft, Amazon earnings report: The question of the market has changed from “how much to spend” to “when to earn back”
Despite adverse factors such as soaring US Treasury yields and rising energy costs due to the Middle East conflict, strong US profit growth supported this year's stock market gains. According to LSEG IBES data, up to now, more than 100 companies have announced their results, and the profit of the S&P 500 index in the third quarter is expected to increase by more than 32% year-on-year. And next week will be the busiest week of the US stock earnings season.
Hyperscale cloud service providers with a combined capital expenditure of close to $800 billion a year will continue to hand over their report cards this week. But now, the nature of this AI investment boom has changed markedly: from 2024 to 2025, the market rewards companies that announce large-scale AI investments; since the second half of 2026, the penalty threshold has rapidly declined.
Nick Giorgi, chief stock strategist at Alpine Macro, commented: “There are many uncertainties currently intertwined. More tech companies will release earnings reports this week, and the market is currently taking a cautious look at all AI related spending. The market will not give unconditional trust to companies that spend huge sums of money to build AI infrastructure. In our opinion, this is actually very healthy, but it still indicates pain in the short term.”
Alphabet will announce the results after the market on Wednesday EST. The company raised its 2026 capital expenditure guidance by US$15 billion in the previous quarter to US$1950 billion to US$205 billion, and warned that short-term profit margins and free cash flow would be under pressure due to the adoption of third-party computing power transition strategies in the third quarter. In other words, Google is sacrificing current free cash flow in exchange for catching up with AI computing power supply. The market's tolerance depends on whether Google Cloud can accelerate revenue quickly enough to hedge against depreciation and the erosion of rental costs.
Bank of America Securities analyst Justin Post expects Google Cloud's revenue growth rate to reach 93% in the third quarter, which is significantly higher than the 89% market consensus. The impetus comes from incremental production capacity investment and the start of TPU external sales. Google Cloud's $514 billion order reserve provides the bottom line for revenue visibility, but the real valuation risk is that the gap between capital expenditure and free cash flow is widening, and the AI infrastructure depreciation cycle has only just begun. One marginal variable less discussed is the priority of computing power allocation—there are reports that Google Cloud is prioritizing limited computing capacity to internal DeepMind models, which may raise concerns about some commercial customers migrating to competitors.
Microsoft will also announce the results after the market on Wednesday EST. Microsoft's narrative is relatively stable, but the absolute magnitude of capital expenditure constitutes an independent source of pressure. For the previous quarter ending June 30, 2026, Microsoft reported revenue of $90 billion, an increase of 18% over the previous year. Revenue from Azure and other cloud services grew 43%, reflecting continued market demand for cloud computing and AI infrastructure. Management indicated that Azure revenue for the September quarter increased by approximately 44% to 45% at a fixed exchange rate.
However, Microsoft also expects its capital expenditure to exceed $50 billion in the first quarter of the fiscal year, including the impact of lease reclassification after updating asset life assumptions. This means that investors will examine whether Azure's growth, enterprise AI adoption, and profitability can keep up with these huge investments. Any slowdown in cloud business growth or weaker-than-expected guidance may cause the market to worry about future profit margins. The company's capital expenditure plan of about 190 billion US dollars for the natural year 2026 (including the impact of component price increases of about 25 billion US dollars, which was later lowered to 175 billion US dollars due to accounting changes, and the actual investment scale did not decrease) increased by about 61% compared to 2025.
Amazon, on the other hand, will release earnings after the market on Thursday EST. In an outlook released by Deutsche Bank, it raised the AWS revenue growth forecast for the third quarter to 41%, further to 44% for the fourth quarter, and set the operating profit margin forecast for the third quarter at 37.1%, up from 34.6% in the same period last year. The acceleration of AWS will be the most directly beneficial variable in this financial report — it responds to the question of whether enterprise-grade AI workloads are moving from piloting to large-scale consumption. However, the same research report also raised Amazon's capital expenditure forecast for 2027 to about 330 billion US dollars, which means that spending expansion for the current fiscal year has not peaked, and the free cash flow repair schedule is being pushed back. For investors holding Amazon, the tension between the rise in AWS profit margins and the trend of free cash flow at the consolidated level will determine the direction of stock prices after financial reporting.
SK Hynix and Samsung: AI hardware chain profits are being realized, but “after cashing out” problems have surfaced
If the financial reports of the US cloud giants test the intensity and sustainability of AI demand, then the performance of South Korea's storage duo tests the extent to which profits are realized in the AI hardware supply chain — and the answer is already quite clear.
According to preliminary guidelines released by Samsung Electronics earlier this month, operating profit for the third quarter reached 107.4 trillion won, surging 782.5% from 12.17 trillion won in the same period last year, surpassing 100 trillion won in a single quarter for the first time in the company's history. SK Hynix (SKHY.US) will announce its official results after the market on Tuesday EST. The 10 South Korean brokers' consensus expectations are revenue of 96.66 trillion won and operating profit of 75.97 trillion won for the third quarter, with an operating profit margin of over 75%. The combined operating profit of the two is expected to be around 185 trillion won (about 139 billion US dollars), setting the highest quarterly profit record in the history of the global storage industry.
The driving factor behind the explosion in performance is undoubtedly the commercialization volume of HBM4. The two companies are supplying sixth-generation high-bandwidth memory for Nvidia's next-generation Vera Rubin accelerator architecture. The HBM4 doubles the data bandwidth and reduces power consumption by more than 40% compared to the previous generation HBM3E. SK Hynix accounted for about 50% of global HBM revenue in the second quarter, 33% for Samsung Electronics, and 18% for Micron Technology (MU.US) — Samsung is rapidly catching up after solving the packaging bottlenecks during the HBM3E period, and indicates that HBM4 will account for more than 60% of its HBM revenue in the second half of the year. Meanwhile, the simultaneous rise in standard DRAM contract prices (server memory contract prices rose 13% to 18% in the third quarter) provided a second layer of support for profits.
However, the “perfect financial report” of memory chips alone is not enough to eliminate the overall anxiety of AI transactions. The reason is that there are two fault lines that will soon be examined by the market.
The first is the high concentration of profits. Almost all of Samsung Electronics' operating profit comes from the Device Solutions semiconductor division. The foundry business and mobile division recorded losses during the same period, which means that any marginal loosening in storage prices will impact profits with a disproportionate leverage effect.
The second is the distribution signal for return on capital. Samsung's total shareholder return plan for 2026 is between 90 trillion and 110 trillion won, while SK Hynix promised that at least 50% of the cumulative free cash flow from 2025 to 2027 will be used to return shareholders, and has completed 40 trillion won in treasury stock repurchases and cancellation. In a context where AI capital expenditure bills are still expanding, storage vendors are choosing to use large-scale cash returns rather than fully reinvest in production capacity, which may raise some questions in the market: they may be more cautious about the duration of this HBM overcycle than buyers assume.
In addition, other high-profile companies will also release earnings next week, including chip companies Intel (INTC.US) and SanDisk (SNDK.US), pharmaceutical companies Lilly (LLY.US), oil giants ExxonMobil (XOM.US) and Chevron (CVX.US), and industrial equipment giant Caterpillar (CAT.US).
Resolve the next lowest inflation reading to reprice the denominator for AI valuations
In terms of economic data, the US personal consumption expenditure price index for September will be implemented on October 29 — the day after the announcement of the Federal Reserve's decision — within the same week as the GDP data for the third quarter. This schedule is expected to amplify the impact of the data on the market: investors will be forced to independently reprice the inflation trajectory less than 24 hours after digesting the perfect Federal Reserve decision.
Currently known anchors are: overall PCE rose 3.4% year on year in August, and core PCE rose 3.0% year on year, all marginally flat compared to July, but overall inflation is still significantly higher than the Federal Reserve's 2% target. Deutsche Bank expects the core PCE to rise to 3.1% to 3.2% year on year in September. The CPI widely exceeded expectations as the basis for this judgment. More importantly, GDP for the third quarter is expected to continue to show steady growth and stable labor market conditions — this means that the stickiness of inflation is not a product of demand disruption, but rather a result of demand continuing to exceed supply potential. For Walsh, the composition of this data set means that in the absence of a substantial rift in growth, there is no reason to regard the current policy interest rate of 3.75%-4.00% as restrictive enough.
The impact of PCE data on AI transactions is that the valuation of long-term growth stocks is essentially a discount on forward cash flow, and the actual interest rate path is the core denominator component of this discount process. While the molecular side (the return cycle of AI capital expenditure) is being examined by the market from a more strict perspective, any improvement on the denominator side (interest rate expectations) will have a multiplier-level valuation compression effect. What is unique about this week is that the testing of the numerators (the financial reports of the three major cloud giants and storage champions) and the recalibration of the denominator (Federal Reserve Decision and PCE) took place within the same five-day window.