This week, Microsoft, Alphabet, and Amazon need to answer the question of returns on AI investment. ExxonMobil and Chevron offer another reference point: when profits are realized as cash, and that cash can be converted into dividends, buybacks, and debt reduction, what price is the market willing to pay for such assets?

S&P 500 buy-and-hold delivered a cumulative return of 392% over the past 10 years, while Bitcoin trend-following strategy averaged 25% annual returns over the past 5 years. Different markets require different investment strategies.

A preliminary quarterly guidance — released before full financial results — has already told the demand-side story. The answers on gross margin, cash flow, and customer structure will only come with the full earnings report on November 3. This piece lays out the numbers that can already be verified, then works through the profit math the market is currently trading on.

A testable interpretation is this: U.S. AI stocks are earning money from the compute shortage, while Hong Kong AI stocks are earning money from traffic and scenarios. These two kinds of money move at different speeds, so their valuations should not be the same.

Whether this divergence can persist depends on two sets of numbers after tonight: Meta's capital expenditures and free cash flow, and whether oil prices can hold below $100.

"After roughly ten months in cash, the on-chain market state model has maintained a 67% net long allocation for a second consecutive month, with $74,000 and $77,000 becoming the key levels for whether the bull market can continue"

Bitcoin's long-term indicators still point to a bull market, but short-term signals are cooling, inflation is picking up again, and the Fed is expected to tighten policy further. This article lays out the reasons for caution and the reasons for optimism, and starting from the age-old question of "hold or trade," explores how to respond next.

For the rebound in early trading on October 16 to continue, we need to see a narrowing of net foreign selling or new corporate actions; otherwise, the pricing pressure from the buyback vacuum may persist.

What South Korea encountered was the simultaneous ebb of three types of buying flows, with the impact amplified by high concentration; what Hong Kong stocks encountered was on-market rotation by allocation-type capital, with the impact absorbed through structural diversification. The sevenfold difference in declines can be half explained by asset structure, and the other half depends on the nature of the buying flows. This framework is inductive rather than a causal proof; whether it holds can be verified item by item using the observation points below.

In a week when AI stocks had just tumbled sharply and credit market alarms were ringing frequently, what this briefing delivered was not just a set of revenue figures, but a public test of the AI commercialization narrative.

The structure of compute consumption may be shifting from GPU-dominated to CPU, memory, and GPU collaboration. On the hardware side, demand increments are relatively more verifiable, while the impact on the software side depends on the pace of agent substitution; both threads are driven by the same product cycle. To what extent this shift has already been priced in by the market remains to be seen.

KOSPI has fallen for five consecutive sessions, with foreign investors selling net continuously. The performance of Samsung and SK Hynix serves as a direct read on sentiment across the memory supply chain. How much of the repricing has already been priced in by the market remains to be seen.

On the market side, the 10-year U.S. Treasury yield closed near 5.33% last Friday, having briefly dipped before rebounding more than 10 basis points on the day the nonfarm payrolls data was released, and touched an intraday high of 5.357% on October 8; this Monday is Columbus Day, and the U.S. bond market is closed.

Taiwan stocks closed down 1.97% at NT$2,490 on Tuesday. A week before the earnings report, the Philadelphia Semiconductor Index had just experienced four consecutive declines, and Broadcom's CEO once again put the warning that "capital is the biggest constraint" on the table, giving this report card an extra layer of interpretation compared with previous years.

Based on the overall market picture and public information, one possible scenario is that the headline reading comes in high while the core meets or approaches expectations, allowing the pricing for the Fed to hold steady in October to persist. If the core lands on the 0.3% side, the consensus on the pace of adjustment would need to be reassessed.

Bank of America and Morgan Stanley take the baton tonight, and two judgments need to be verified: whether the rebound in provisions is an industry-wide trend or an individual difference, and whether the strength in equity trading revenue can extend to the second firm. The answers to these two questions will determine sector pricing as the big five banks' earnings week comes to a close.