Oli.
Oli.
🍓Web3投研 🍑人工智能 🚀《干翻狗庄》系列工具作者
5Following
260followers
Feed
Livestream
Feed
Samsung's Q3 operating profit is expected to increase nearly 8 times, and the numbers are indeed shocking. But before the full financial report is released, what I most want to resist is equating this directly to "HBM has completely won."
Samsung is a conglomerate with many businesses. Storage, wafer foundry, mobile phone businesses, etc., face different demands and costs. A significant increase in group profit can prove strong overall performance but cannot automatically endorse every business line.
Especially regarding the rise in storage prices, the implications may differ across departments. The supply side benefits from the price increase, while end products must consider component costs. Internal transactions also require consolidation and elimination, so the benefits of each department cannot be simply added together.
So this time I want to see which businesses mainly contributed to the profit and which still need continued investment.
If storage profitability improves while other businesses also improve, the market's understanding of growth will be more complete; if profits are highly concentrated in one area, then we need to ask how long that segment's prosperity can be maintained.
Such a large year-on-year increase inevitably creates the impulse of "if you don't buy now, it will be too late." I also think the results are impressive, but the earnings forecast has already told everyone the total and breakdown. The value of the full financial report lies in seeing how each subject performed. The upcoming expectation gap is likely hidden there.
#三星Q3利润预增近8倍，完整财报即将揭晓
On October 26, the US BTC spot ETFs recorded a net inflow of approximately $110.3 million, with funds turning positive again, bringing some relief to the market after continuous outflows.
However, I want to clarify a common misunderstanding: ETF trading on the exchange does not mean the fund simultaneously receives the same amount of new capital.
When investors buy ETF shares on the secondary market, they may simply be acquiring them from another holder. Only when fund share subscriptions or redemptions occur does it affect the capital flow we usually discuss. Therefore, a high trading volume cannot be directly translated as "this much institutional money has bought BTC."
This also explains why sometimes ETF trading is very active, but the coin price does not respond as expected. Trading activity and new buying demand need to be observed separately.
Of course, I hope to see net inflows continue to improve. But rather than being kept awake by a huge trading volume, it’s better to align actual capital flow with price performance.
If new funds keep coming in and pullbacks are gradually absorbed, then there is more basis for judgment. This current inflow is worth noting but not enough to guarantee anything in the coming days.
BTC already causes enough emotional ups and downs; there’s no need to misinterpret "someone changing hands" as "new money rushing in," adding an extra layer of expectation.
#BTC现货ETF资金反复，市场买盘能否企稳？
Amazon's earnings report is coming soon. Besides AWS growth, I also want to hear about Trainium.
Amazon disclosed in Q2 that the annualized revenue scale of its AI business and chip business each exceeds $25 billion. It is important to note that the annualized scale is calculated based on the current business pace and should not be considered as completed full-year revenue, nor should the two be simply added together as independent new revenue.
What interests me is whether self-developed chips can help AWS retain customers.
If customers migrate workloads over and achieve suitable performance and cost, AWS has the opportunity to secure longer-term usage through the combination of chips, software, and cloud services. Such customer relationships are more worth observing than a one-time equipment sale.
However, from announcement to large-scale use, adaptation is needed. Whether developers can migrate easily and whether operations are stable will affect how many tasks customers ultimately put on it. Signing a partnership is far from completing all the work.
I hope the earnings report can provide more progress on actual usage and less hype supported only by promised amounts. After all, customers' willingness to pay continuously indicates the product has entered daily operations.
AWS is already very large, and the market's willingness to expect higher growth naturally demands more concrete evidence. If self-developed chips can continuously bring customers and profits, I think this achievement is more exciting than simply buying another batch of GPUs.
#AWS算力需求旺盛，亚马逊Q3财报有哪些看点？
Yesterday there was still debate about whether the financing was completed, but today there is a company announcement, which at least moves the discussion forward.
SpaceX has reached a binding commitment with an entity managed by Apollo to acquire GPUs and related equipment through a financing lease. The formal lease still has delivery conditions, and the company expects to complete it by the end of the year, with equipment delivery and lease arrangements expected to start in the first half of 2027.
This means that the $40 billion should not be understood as having been fully received today or turning into immediately available computing power tomorrow.
There is an actual delivery process involved: the equipment arrives, then needs to be deployed, and only afterward can it serve customers. Financing solves the funding arrangement, but revenue realization still depends on business operations starting. Compressing these time points into the same day easily leads to overly high expectations for performance.
I am quite interested in this type of equipment financing. It provides companies with an additional funding option during expansion but also brings the responsibility of ongoing payments. Equipment updates quickly, so whether customer demand and utilization can match the lease cycle is worth watching.
Management mentioned that the return on existing AI infrastructure investments is very fast, which is certainly attractive. However, the performance of historical projects also needs to be verified after the new batch of equipment is put into use.
This announcement is much more valuable than just saying "financing succeeded." At least we can finally judge which upcoming developments are worth looking forward to based on the equipment delivery timeline.
#SpaceX达成400亿美元GPU融资租赁协议
Meta launches approximately $35 billion financing; at first glance, it seems AI infrastructure has more funds, but looking further, shareholders have more to consider.
This arrangement includes the issuance of common stock and mandatory convertible preferred stock. The company also proposed a subsequent ATM issuance plan of about $60 billion, but the specific implementation terms have not yet been determined, so both parts cannot be counted as already received.
What I care about more is the return per share.
The company obtaining new funds can build more infrastructure and potentially create larger business. But if the number of shares increases with financing, how much the old shareholders ultimately get depends on whether the returns from the new investment can cover the corresponding dilution effect.
Therefore, "the company earning more in the future" is not enough to solve this problem. Shareholders want to see improvements in the earnings and cash generation ability per share.
I support investments that yield returns. Meta already has mature business and conditions to do long-term construction, which is commendable. But as the financing amount grows larger, it is also very normal for the market to question investment efficiency.
Getting the money is progress; spending it for returns is the achievement. Facing a figure like $35 billion, I would be excited, but after the excitement, I still want to know: after this round of expansion, how much more valuable will each share held by old shareholders become?
#Meta启动350亿美元融资，支持AI基建
Anthropic's official IPO filing is indeed exciting. An AI company that has been discussed for so long is finally starting to undergo more detailed scrutiny from the public market.
However, there is one sentence in the official announcement that is particularly worth noting: the number of shares issued and the price range have not yet been determined.
In other words, the "potential record valuation" is still just an expectation at this point. The company is strong, and buying in at any price is a different matter. Popular IPOs tend to mix these two ideas, creating the illusion that if you miss the first day, you will never have another chance.
I actually think the research only begins after the prospectus is made public.
How revenue is recognized, why customers continue to pay, how inference costs change—these issues will all affect valuation later. Leading model capabilities are certainly exciting, but the public market will ultimately ask: how much profit can the increased usage retain?
I am willing to seriously consider this company and admit that I am attracted by the scale of this listing. But the more I like the product, the more I have to remind myself to separate the investment price for evaluation.
Users of Claude and holders of Anthropic stock do not have exactly the same expectations. The former hope for stronger capabilities and lower prices, while the latter also care about how these changes affect returns. Waiting for the issuance terms to come out before deciding whether it is worth participating is, in my opinion, not late at all.
#Anthropic正式申请IPO，或创上市估值纪录
Waiting for Robinhood's earnings report tonight, I want to focus on the structure of trading revenue.
The company's Q2 options revenue was $342 million, and cryptocurrency revenue was $100 million. This gap is very straightforward: when we talk about Robinhood, we often follow the crypto price, but its trading business is far more than just cryptocurrency.
The issue is that different trading categories have different monetization methods and revenue levels. Stock trading volume, options contracts, and cryptocurrency trading volume cannot be lumped together to conclude that "the busier, the more profitable."
If trading volume increases but revenue doesn't keep up, we need to see whether the business structure has changed or if the revenue per unit of trading has changed. Conversely, even if total trading volume isn't particularly impressive, growth in high-revenue-contributing businesses can still lead to results exceeding expectations.
I am optimistic about the platform enriching its products, but just looking at a screenshot of trading volume and shouting about explosive performance really skips the process.
Users open the app, complete trades, and the company ultimately retains profit, with clearing, service, and operating costs in between. The faster the expansion, the more these details cannot be ignored.
Robinhood has already shown the market many growth stories. Next, what I want to see is whether these new businesses actually make each unit of revenue more solid or just add more items to the financial report.
#Robinhood交易业务增长能否延续？
The US and Iran discuss a plan to open the Strait of Hormuz, and oil prices fall; this reaction is easy to understand. As long as the possibility of easing the conflict increases, the market will first remove some risk premiums.
But when I see this kind of news, I increasingly care about whether the plan includes an arrangement of "who acts first."
Iran hopes to lift the blockade, while the US demands free passage through the channel. Both sides are willing to talk, but there is a big difference between both sides accepting the same execution order. Especially when each step requires the other side to act first, negotiations can easily stall, while prices have already moved ahead.
For oil prices, a very frustrating trend may follow: prices drop a bit when there is progress, then rise back when execution gets stuck. You think you are judging supply and demand, but in reality, you are guessing the next diplomatic news every day.
I don’t want to directly write "discussion of the plan" as the end of the supply crisis, nor do I think oil prices must rebound just because they fall. What’s more worth tracking is whether both sides have a clear timetable and which actions can be independently verified.
The most annoying part of this market phase is that the viewpoint may be correct, but the timing of entry is completely off. Peace expectations are welcome, but the market can trade in advance, while reality needs to be fulfilled step by step. The time gap between the two is enough to torment those chasing rallies and selling off multiple times.
#美伊讨论霍尔木兹开放方案，油价回落
Tomorrow we will look at SK Hynix's earnings report, and the conclusion I most want to avoid is a lazy one: profit growth is all thanks to HBM4.
The company has already disclosed that HBM4 began mass shipments in Q2 and will continue to ramp up in the second half of the year. But within the same results, rising prices of traditional DRAM and NAND, as well as increased sales of high-value products, are also driving profitability. Several forces combined ultimately present a nice figure, which should not be casually attributed to the hottest concept.
This time I will pay special attention to whether management clearly explains the contributions of product mix and pricing.
If profit improvement mainly comes from a higher proportion of more competitive products, I will be more confident; if it mainly relies on industry-wide price increases, then we need to continue to assess how long this price level can hold. Both types of growth can be profitable, but their sustainability requires different evidence.
There is also an easily overlooked issue: after suppliers increase shipments, are customers accelerating usage or stocking up in advance? This will affect the rhythm of subsequent orders and cannot be directly extrapolated just because this quarter's deliveries are lively.
I have expectations for HBM4 and am also willing to acknowledge the contribution of traditional memory. When spending money to buy stocks, there is no need to be a fan of a certain technical term. The better the earnings report looks, the more we should break down the reasons for making money clearly, so as not to end up betting on the right industry direction but misjudging the source of profits.
#SK海力士Q3财报在即，HBM4能否提振盈利？
For this rate decision, I plan to pay less attention to the debate over whether "holding steady is good news" and focus more on U.S. Treasuries.
The reason is simple: the Federal Reserve's announced policy rate remaining unchanged does not mean that market financing conditions haven't changed. If the press conference makes investors feel that high interest rates will be maintained longer, bond yields could still rise, and valuation pressure on risk assets would increase accordingly.
So, rushing to buy BTC just because the "rate is unchanged" seems a bit hasty to me. What you're really trading is whether money will become more expensive in the coming months. Focusing only on whether there was an adjustment this time easily misses the latter part.
After the decision is announced in the early hours of October 29 Beijing time, I will observe BTC alongside the two-year U.S. Treasury yield. If BTC rallies first but yields continue to rise afterward, whether this rally can hold is questionable. Of course, the two won't move in a fixed relationship every minute.
I understand everyone is hoping for a breather. After such a long period of volatility, everyone wants a clear direction.
But there is a gap between "results meeting expectations" and "worth rising afterward." Rather than rushing to label the meeting as dovish or hawkish, I care more about which direction the market re-prices future interest rates after hearing the news. This calculation will ultimately be reflected in the price.
#本周将公布10月FOMC利率决议
