The recent earnings calls from the three major cloud providers clearly indicate that enterprise adoption of AI has entered the diffusion phase, which is why I believe we are now in the midfield. As discussed in the recent tweet “The Ice and Fire of AI’s Midfield” https://t.co/RqesZVj4sN: “Enterprise AI is transitioning from ‘buying seats’ to ‘sustained consumption of compute and agents’; the depth of spending and willingness to pay among existing customers have become the core growth drivers, while new customer acquisition continues to rise.”

Apple’s earnings report was indeed mixed, as anticipated in our recent preview: revenue and EPS exceeded expectations, but growth in the China region and services business slowed. More critically, management proactively disclosed forward-looking guidance during the earnings call—high memory costs are expected to intensify over the coming quarters, further pressuring gross margins. CFO Kevan Parekh explicitly stated during the earnings call that memory costs were the primary driver of margin pressure this quarter, and the impact of elevated memory costs will be even more significant next quarter. While Apple previously stocked up on lower-priced components, this inventory is rapidly depleting, forcing the company to procure memory at higher prices going forward. Apple’s gross margin guidance for the December quarter is set between 45% and 46%. The company emphasized that this range “primarily reflects the impact of higher memory costs and certain expenses related to the launch of Siri AI.” This guidance represents a substantial decline of nearly 2 to 3 percentage points from the actual 47.9% gross margin this quarter and is well below prior market expectations, directly triggering an accelerated drop in stock price following the earnings call.

When politics is involved, many people only care about立场, not viewpoints or frameworks. This tweet began by citing prediction market data showing that a Democratic sweep in the midterms is highly likely, while Republicans retaining the Senate is unlikely—this sparked widespread dissatisfaction, with many claiming that polls and prediction markets are part of a left-wing conspiracy. Whether it’s a conspiracy or not, I’m actually not that concerned. In fact, just last week, this link https://t.co/o6v6YrNzrk used the 2022 midterms as an example, cautioning that despite the current strong momentum for Democrats, a sweeping victory isn’t guaranteed. But did people actually read the full tweet? The opening merely presented prediction market data on two possible scenarios—but the core content wasn’t about who would win. Instead, it focused on how different congressional power structures under a Democratic sweep versus Republican retention of the Senate would impact AI and crypto policy, fiscal spending, and market outcomes. More importantly, the tweet emphasized: from a market perspective, “the end of the midterm election may matter more than the result itself.” For investors, which party wins isn’t what matters. What matters is analyzing how different scenarios affect policy trajectories and market outcomes—and preparing appropriate strategies in advance. As for who wins? That wasn’t the focus of this tweet at all. Yet it seems many people only read the beginning and ignored the full content. 🤣

Intel's intraday drop of -30.22% tonight likely set a new record for the largest intraday decline among U.S. publicly traded companies with a $100 billion market cap. Short-term pessimism may have already been largely priced in, but a reversal will likely still take time. https://t.co/5ZPyHaLvGO

Discussing Amazon’s Q3 earnings: “flaws don’t overshadow the strengths.” The results were actually solid, but under current market conditions, some imperfections stand out. First, the flaws: 1) Revenue slightly missed expectations, largely because this year’s Prime Day occurred earlier in Q2 across most of Amazon’s key markets, whereas last year it fell in Q3. Net profit was $20.69 billion, down from $21.19 billion in the same period last year; diluted EPS was $1.89, a 3% year-over-year decline. This drop was primarily due to a high non-recurring base last year—Amazon recorded a $9.5 billion pre-tax gain in Q3 2025 from its investment in Anthropic, with no similar gain this quarter. Advertising revenue underperformed expectations; the company attributed this to timing factors like Prime Day and expects advertising growth to rebound into the high single digits in Q4. If this guidance holds, the slowdown in Q3 advertising may reflect timing noise rather than structural weakening. 2) The market’s biggest concern: AI capital spending became the core pressure on the stock price: the sharp turn to negative free cash flow was the key driver of the market’s negative reaction. Amazon’s Q3 capital expenditures reached $77.26 billion—about 1.2 times last year’s level and over 34% higher than expectations. Free cash flow turned negative for the second consecutive quarter, shifting from an inflow of $14.79 billion last year to an outflow of $42.36 billion, far exceeding the outflow in Q2. The company maintained its full-year 2026 cash capital expenditure guidance at approximately $220 billion, offering no signal of slowing AI investments. 3) Q4 guidance offered no surprises: Amazon’s Q4 guidance remained broadly in line with Wall Street expectations. The company forecast net sales between $240 billion and $247 billion, representing year-over-year growth of 12% to 16%. The midpoint of $243.5 billion was slightly below analyst expectations of $244.52 billion, though still within the guidance range. Operating profit guidance was set at $28.5 billion to $33.5 billion, compared to market expectations of around $33.1 billion. In other words, the upper end of guidance only slightly exceeded consensus, while the midpoint of $31 billion fell below expectations. Now for the highlights: 1) Amazon’s Q3 operating profit was $28.08 billion, a 61% increase from $17.42 billion last year; operating margin rose from 9.7% to 13.9%. 2) The biggest highlight remains AWS. AWS net sales grew 39.3% year-over-year to $45.98 billion—1.3% above analyst estimates—marking its highest quarterly growth rate in nearly five years. Annualized revenue at quarter-end reached $184 billion, up 8.9% from Q2. In summary: Amazon’s Q3 revenue and EPS both slightly missed expectations; advertising revenue was slightly below forecast; Q4 profit guidance midpoint was weak; capital expenditures significantly exceeded expectations, and full-year AI investment plans remain elevated. However, AWS’s strong growth and record-high operating margin were standout positives. Yet under today’s market conditions, investors are more focused on whether AI spending will continue to suppress free cash flow, and whether these high capital expenditures can translate into sufficiently high profits and cash returns within a foreseeable horizon—hence the post-earnings stock decline. How should we view Amazon’s earnings? From my perspective: “flaws don’t overshadow the strengths.” The core issue is growth potential—everyone understands the burning cash, but the market remains unclear about how far AWS and Amazon can scale. However, this Q3 earnings report and earnings call actually revealed a clearer picture of that future vision: 1) CEO Andy Jassy described AWS’s current growth dynamic as a self-reinforcing “flywheel”: leading capabilities and security attract customers to migrate to the cloud; centralized data and applications on the cloud then drive AI inference demand; AI inference increases consumption of core compute, storage, and networking resources, which in turn fuels continued investment in AWS’s foundational infrastructure. In response to a Goldman Sachs question, Jassy suggested that multiple models—each with comparable capabilities suited for different tasks—will emerge, and enterprises will combine them within single applications, switching between them based on evolving price and performance dynamics. He believes Bedrock has the potential to grow into a service scale comparable to EC2. Evidence of mainstream enterprise demand has solidified: Fortune 500 companies’ spending on Bedrock has nearly quadrupled compared to the same period last year, indicating clear expansion in real corporate consumption; Total token usage on Bedrock has grown nearly 17-fold year-over-year; The number of agents deployed in Q3 was roughly double the total deployed across the prior three quarters combined. These figures confirm rapidly accelerating inference demand and broader enterprise adoption of agent deployment. 2) Selling self-developed chips could be a new highlight: In Q3 alone, customers using Amazon’s custom chips (Graviton and Trainium) saved over $2 billion. Revenue commitments for Trainium and Graviton increased more than 20-fold year-over-year. Jassy revealed: Due to such strong demand for Trainium, customers have repeatedly encouraged us to offer it outside AWS—we plan to do so by late 2027. 3) AWS revenue potential over the next two years: Jassy also expects AWS growth rates above 40% in both 2027 and 2028, with potential to reach $1 trillion in annual revenue within five to ten years. Meanwhile, backlog orders reached $578 billion, with year-over-year growth in triple digits. According to Jassy’s figures—AWS annualized revenue at $184 billion in Q3—if growth rates exceed 40% annually through 2027 and 2028 (using the minimum 40% assumption), AWS revenue could exceed $360 billion by 2028. 4) Unwavering commitment to spending: The market worries about Amazon’s capital expenditures—but management expects this spending to “significantly increase” again by 2027. In response to this pace of investment, Jassy’s stance is resolute: he views it as a strategic build-out of a long-term platform—not merely short-term data center leasing. His exact words: “AI is a once-in-a-lifetime opportunity—it’s the greatest technological transformation of our lifetime. If we can sustain over 40% annual growth for two consecutive years on an $18.4 billion business and ultimately reach a trillion-dollar scale, we must invest to maintain leadership.” One line from Jassy’s earnings call captures the magnitude of this gamble: “We have the financial capacity to fund this plan—I believe it will permanently transform Amazon’s scale, revenue, and free cash flow.”

The Fed’s interest rate decision is due tomorrow morning. A hold on rate hikes is highly likely, but the key will be how Walsh responds. Several Fed officials made密集 statements in October, all conveying a consistent signal: tightening will proceed at a “every-other-meeting” pace—i.e., rate hikes in September, a pause in October, and another hike in December. Previously, New York Fed President John Williams explicitly stated there was “no urgent need for action in October” and hinted that only one more hike remains this year. Fed Vice Chair Philip Jefferson suggested policymakers may need more time to assess data. Governor Waller summarized this stance as: rate hikes “need not occur at consecutive meetings, but should be implemented within a reasonable timeframe.” The September FOMC meeting minutes further confirmed this framework: all 19 officials supported the September hike, and “most” deemed one additional hike this year appropriate. Thus, a hold at this FOMC meeting is highly probable. But Walsh’s view on December’s potential move could be critical: 1) How will Walsh explain his assessment of inflation trends? 2) Will he emphasize that inflation will reinforce curve flattening and push front-end rates higher, or will he stress that tighter financial conditions could re-steepen the yield curve? 3) Will he use language such as, “Any further rate adjustments will depend on data trends, evolving outlooks, and the balance of risks,” or will he remain entirely silent on the policy path? If the Fed indeed holds rates this time while leaving December’s trajectory ambiguous, markets will shift focus to current developments. Tomorrow morning also brings earnings reports from Microsoft and Google—if these reports reveal AI commercialization accelerating faster than expected, we may see a rally ahead of Anthropic’s roadshow.

Supreme Leader Mojtaba met with Iran’s top security officials, reiterating the conditions for ending the war with the U.S. and discussing the management of the Strait of Hormuz. It appears that yesterday’s high-level military delegation from Saudi Arabia, Turkey, and Pakistan, which traveled to Iran to mediate, had some effect. Iran places significant importance on this meeting among the Mecca Agreement member states. For years, Iran has worked to establish a regional order that isolates Israel and excludes the United States. Today’s meeting represents a step—albeit a small one—in that direction, though currently form may still outweigh substance. Nevertheless, differences remain. According to The New York Times yesterday, Qatari mediators are pushing for a phased agreement that envisions both the U.S. and Iran making a series of reciprocal concessions over five days, including Iran reopening commercial shipping through the Strait of Hormuz and the U.S. lifting its maritime blockade on Iranian ports. The main obstacle lies in the precise sequence of these concessions. Iran has previously expressed pessimism about reaching an agreement, fearing it may still face new military strikes after making concessions. The key next focus will be the outcome of the upcoming Israeli election, particularly regarding Netanyahu’s future.

“Strike while the iron is hot.” If earlier, Musk demonstrated to us the extraordinary vision and sense of mission of a tech entrepreneur, now he reveals the cold ruthlessness and impeccable timing of a business titan in the face of competition. He lays bare the brutal reality of corporate warfare: highlighting genuine funding pressures to make rivals harder to finance and less likely to survive, while simultaneously showcasing his own advantages—non-AI revenue streams and investment-grade credit. This isn’t just harsh—it’s also likely to accelerate industry consolidation. His timing is flawless: CoreWeave and other leading new-cloud providers have recently seen their CDS spreads surge, bond prices plummet, and stock values sharply decline, as the market has grown deeply skeptical of business models reliant solely on borrowing to buy GPUs and then leasing them out. Moreover, Musk isn’t just FUDding the new-cloud vendors—he’s scaring off their customers. Who would entrust their business to a cloud provider that appears to be on its last legs? Why is Musk so ruthless—and why is his timing so precise? Everyone knows these new-cloud firms carry heavy debt. But if you dig deeper into their backlog of orders, revenue, operating cash flow, and repayment schedules, you’ll find that as long as they maintain their current growth trajectory, they’re likely to weather the storm. We discussed this before here: https://t.co/KiHjYb7glR—but what’s the Achilles’ heel of these new-cloud firms? It’s continued customer orders and sustained revenue growth. Musk has struck directly at the core: his FUD may cause many customers to pause or withdraw, cutting off or severely reducing cash flow—the most lethal blow possible. It’s a classic case of cutting off the fuel supply. Finally, it’s impossible not to reflect: when the most influential tech magnate launches a FUD campaign, most people follow the emotional tide—and suddenly, all discussions about backlog, operating cash flow, or repayment schedules become irrelevant. At this point, the new-cloud companies are surely hoping Huang will speak up.

Today, Bitcoin experienced a sharp breakdown. From my perspective, liquidity continues to tighten, with the lowest point expected at the end of October. For the full reasoning, see today’s tweet and the two links within it. The exact wording from today’s tweet: “The buy point for Bitcoin I mentioned—possibly in the second half or late October—may finally be arriving; we’ll see what next week brings.”

As of this past Wednesday, the latest Fed H4.1 report shows reserves declined to $2.96 trillion, while the Treasury’s TGA account stands at over $940 billion. Reviewing the Treasury’s debt auction schedule, multiple bond issuances—including bills and coupons—are still pending through the end of October, with a net funding requirement of $132 billion ($106.1 billion from bills and $26 billion from coupons). This implies that by month-end, the TGA is likely to rise again to its target range of approximately $1.05 trillion—the same range the Treasury set for late October in early August. Consequently, reserves could fall further to around $2.85 trillion, aligning with our earlier prediction of a liquidity low point by late October. According to the Treasury’s prior plan, TGA funds will begin to be gradually released starting in November, leading to a rebound in reserves. This suggests that the previously discussed optimal buying opportunity for Bitcoin—likely in the second half of October or early November—may indeed materialize. We’ll know more next week. Of course, the Treasury’s debt issuance schedule for November through January 2027 will be announced soon; we’ll then assess the precise timing and pace of upcoming auctions. For a detailed breakdown of the logic behind this Bitcoin correction, see here: https://t.co/huSfI3Scit For a comprehensive analysis of the relationship between liquidity, Bitcoin/crypto, and U.S. equities, see this detailed breakdown by Grok: https://t.co/eVTr1w5szw