Oli.
Oli.
🍓Web3投研 🍑人工智能 🚀《干翻狗庄》系列工具作者
5Following
256followers
Feed
Livestream
Feed
Regarding Robinhood's prediction market growth, I want to first exclude the hype brought by events and elections, then assess how much can be retained.
Previously, when the company expanded its partnerships, it took the football season and the upcoming midterm elections as key scenarios. This approach is very smart: users are already paying attention to the outcomes, and the platform turns discussions into tradable contracts with a low participation threshold.
But events have peaks, and trading volume follows a rhythm. A particularly busy month doesn't mean every month afterward will grow accordingly.
What I care more about is whether users who came in for the first time because of games or elections will continue to use other services afterward. If they are willing to keep funds on the platform and use stocks, options, or other products, the new business could bring longer-lasting value.
If the event ends and both people and money leave, then it’s closer to a successful short-term business, and its valuation must acknowledge this volatility.
There is also the relationship between the number of contracts and revenue. Trading many contracts does not directly indicate that profits grow equally; fees and partnership costs must be considered together.
I am optimistic about Robinhood's ability to find more user participation scenarios, but I don’t want to take all the calendar hype as the company’s own long-term strength. After this round of events and elections passes, seeing how many customers remain will be more convincing than screenshots of peak transactions.
#Robinhood交易业务增长能否延续？
In Anthropic's IPO filing, what I find more concerning is that it simultaneously relies on cloud giants to provide computing power and assist with sales.
According to the confidential draft reviewed by the media, about 47% of sales in 2025 will be completed through Amazon and Google's cloud channels. Note, this does not mean these two companies themselves purchase nearly half of the services; many transactions are still ultimately from end customers.
But channel dependency is still worth noting. Cloud platforms can bring enterprise customers and reduce procurement and deployment resistance; on the other hand, Anthropic has to purchase a large amount of infrastructure services from these partners, and their relationship has already penetrated multiple operational aspects.
This kind of cooperation can drive growth but also makes negotiation terms important. How the channel splits revenue, how computing power is priced, and who controls customer relationships will all affect how much profit the company ultimately retains.
I don't think relying on big platforms is a weakness. Enterprise customers naturally want to procure AI services through familiar cloud providers, and it makes sense for the company to do business according to demand.
However, when facing huge long-term computing power arrangements, investors still need to ask how much of the future revenue growth will remain with Anthropic rather than continue flowing to channels and infrastructure partners.
The more popular the product, the more important it is to clarify this issue. What you buy at IPO is the company's profit rights; you can't just assume that because Claude is easy to use, all the revenue created by the entire industry chain will belong to it.
#AnthropicIPO草案披露巨额算力承诺
Stablecoin reward rules are clearer now, but users still need to consider one question: if interest rates drop in the future, how long can the rewards seen now be sustained?
Many stablecoin-related yields ultimately depend on reserve asset returns, platform revenue sharing, or marketing budgets. Regulatory allowance of certain arrangements does not mean the platform can always provide the same rewards.
This point is easily overshadowed by the excitement of "relaxed rules." Everyone is busy comparing who offers more, but not necessarily concerned about why that money can be given.
If underlying returns decline, the platform may reduce rewards or temporarily subsidize them itself. The former affects user earnings, the latter affects platform profits. The specific choice depends on competition and retention effects.
Therefore, I want to know whether the rewards are supported by long-term operating income or are a short-term customer acquisition cost. Both approaches can have commercial reasons, but users should not mistake short-term discounts for long-term commitments.
For platforms, the real challenge is whether users are willing to stay after rewards decrease. If retaining balances relies solely on continuously increasing subsidies, this business is not as easy as it seems.
I welcome clearer rules and hope platform competition benefits users. But when seeing an annualized figure, it's best to also check its adjustment conditions. How much is written on the page today and how long it can be received in the future are two things that need to be confirmed separately.
#GENIUS法案细则发布，稳定币奖励限制调整
Oil prices are fluctuating at high levels, and what worries me more now is that inventory depletion is making the market increasingly unable to withstand new supply shocks.
In the latest weekly data released in the US, commercial crude oil inventories decreased by about 3.1 million barrels. A single week's decline cannot directly prove a global oil shortage, as changes in imports, exports, and refinery processing can all affect the results, but continuous inventory drawdowns are worth monitoring.
The role of inventory is very practical. When transportation is delayed by a few days or some supply is temporarily interrupted, existing stock can be used to cover, so the end users may not immediately feel the pressure. But if inventories keep decreasing, the buffer time for the next problem may shorten.
This is also why I am reluctant to assume oil prices will continue to fall based solely on a single negotiation message. After the channel is restored, it still depends on whether actual supply can cover demand and allow inventories to gradually stabilize.
Of course, one should not blindly be bullish just because inventories are falling. Inventory changes may differ across regions, and crude oil and refined products need to be considered separately; one total figure cannot be used to draw conclusions for all markets.
What I want to observe more is whether this continuous depletion can stop in the coming weeks.
News changes daily, and both bullish and bearish headlines are quite convincing. But if inventories keep going down, it means that in reality, someone is still drawing on previously accumulated oil. This change makes me more uneasy than a single statement like "the situation is expected to ease."
#中东能源风险升温，国际油价维持高位
Both Micron and SanDisk can benefit from AI, but you can't explain them with the same "HBM shortage" logic.
HBM mainly serves high-bandwidth data access needs, while NAND handles persistent storage. Both are in AI infrastructure, but their product uses, supply conditions, and competitive focuses differ.
SanDisk and Kioxia's previously announced new generation QLC flash memory improve read/write performance and energy efficiency. For these products, customers need to compare capacity, cost, and actual workload performance, not just how many GPUs were purchased.
I think this actually makes the storage sector more interesting. AI has expanded the demand for data processing and storage, but where the new demand ultimately flows and who gets the orders need to be studied separately.
Especially when stock prices rebound together, it's easy to think buying any of them is about the same. When earnings reports come out, you find some companies benefit from pricing, some from product upgrades, and others have investments growing faster than revenue.
I have expectations for storage demand but don't want to buy all companies as the same story. Whether new technology can translate into customer purchases and whether costs can be reduced after mass production will impact profits more than the specs at product launch.
If this market rally is to last, I hope to see more companies prove their value through their own business. Being lumped into the "AI concept" and rising and falling together every day is lively but doesn't help much in judging business quality.
#存储芯片股反弹，AI扩产与竞争并存
Seeing ETF capital inflows, people tend to assume that behind it are all institutions with a long-term bullish outlook, but this habit needs to change.
Buying ETFs could mean increasing BTC exposure, portfolio rebalancing, or even hedging with shorts in other markets simultaneously. Funds do flow into the products, but the overall direction of holders cannot be determined from a single flow chart alone.
I'm not saying that all current inflows come from arbitrage—that requires verification with position data and other information. It's just that interpreting every subscription as "institutions preparing to hold for years" lacks sufficient evidence.
Today, we also need to pay attention to the timing. As of the query, some major products for October 22 have not yet reported fully, and among those published, there have been noticeable outflows. We cannot directly apply the previous conclusion of "continuous inflows," nor treat the temporary aggregate as the final result.
This affects our understanding of why the price hasn't strengthened. ETFs are only part of the demand; selling pressure, hedging, and capital withdrawals in other markets may offset their impact.
My attitude toward ETF buying remains positive, but I prefer to treat it as one piece of evidence that needs to be combined with other data.
The easiest mistake is to casually convert inflow amounts into future price gains. The market does not have such a fixed exchange relationship. First, check if the data is fully reported, then look at prices and trading in other markets. Drawing conclusions more slowly is usually more reliable.
#ETF买盘持续回流，BTC为何仍未走强？
Putting SpaceX and Tesla into the same company could make the story richer, but it might also make it harder for shareholders to diversify risk.
This point is easily overshadowed by the imagination of a “super tech conglomerate.” Cars, satellite communications, rockets, and AI seem to belong to different industries and sound naturally complementary, but if they all rely on large-scale investment and continuous financing, changes in capital costs may not be offset by each other.
If the merger really goes ahead, the scope of businesses held by shareholders will expand, and they will also have to accept unified capital allocation. Which business gets funding priority, and which mature business’s profits support others—these decisions will directly affect returns.
I have expectations for technological cooperation, but I am more reserved about the idea that “the more businesses, the lower the risk.” Whether different businesses can form complementary income needs to be proven by operational results, not just by different names.
There is also a very practical question: Are those who originally bought Starlink willing to bear the volatility of the automotive business at the same time? Are those optimistic about Tesla willing to accept more investment in aerospace and computing power?
Currently, it is still in the evaluation stage, with no formal transaction plan, so there is no need to answer on behalf of all shareholders prematurely.
I hope that if there really is merger material in the future, the capital arrangements for each business will be clearly explained. Large scale is certainly eye-catching, but what shareholders ultimately care about is why they should hold these businesses together.
#SpaceX业绩大增，与特斯拉合并能否推进？
There is one thing in Meta's latest earnings report that shareholders should pay more attention to than the increase in expenses: the company is preparing to establish Class C common stock without voting rights.
According to the announced arrangement, existing Class A and Class B shareholders will conditionally receive one Class C share for each share they hold. This proportional distribution itself will not change shareholders' voting rights, nor should it be understood as gaining an additional asset out of thin air.
What is truly worth pondering is what comes next. The new class of shares can be used for financing, employee compensation, and potential acquisitions, giving the company more fundraising tools while maintaining the founder-led governance structure.
I can understand why Meta needs financing flexibility. With the scale of AI investment growing, having an additional stock instrument besides cash payments is indeed useful. But from the perspective of ordinary shareholders, they should also be concerned about the future issuance scale, usage, and how their economic interests will change.
Operating and legal expenses have already complicated the financial statements, and if a new class of shares is added, investors will have even more to scrutinize.
I can be confident in Zuckerberg's judgment; however, trusting management does not mean every expense in the future requires no explanation.
If the company wants to continue large-scale investments, shareholders have the right to know where the money comes from and what the new shares are exchanged for. Especially when buying shares without voting rights, one must carefully assess how much they are willing to pay.
#MetaQ3营收超预期，法律费用与AI支出承压
In Tesla's latest earnings report, what I am more willing to continue following is the software subscription, especially whether the company can keep receiving money after the car is sold.
The company disclosed that active FSD subscriptions have reached about 1.74 million. This number makes me think that when discussing Tesla's profit recovery, we should pay more attention to the vehicles already delivered, not just how many new cars can be sold next quarter.
If existing owners continue subscribing, the company has the opportunity to generate revenue from the current fleet without relying on new deliveries to drive growth each time. If this path goes smoothly, it will help the revenue structure of the automotive business.
But the most important thing not to overlook here is subscription quality. After the active number increases, how many people stay after the trial ends, whether paying customers cancel, and the pricing and usage in different regions all affect the final revenue.
Simply multiplying the subscription number by a US monthly fee and treating it as guaranteed company revenue sounds good but can have a large margin of error.
I am somewhat optimistic about this business but would like to see disclosures on renewals and actual revenue. Users willing to try and those willing to pay year after year are separated by the test of experience.
The rebound in automotive gross margin has already given some confidence. If software revenue can also be steadily solidified, profit recovery will have another support. Rather than continuing to stuff distant robot income into the valuation, I prefer to first see whether car owners will still pay next month.
#特斯拉Q3汽车毛利率回升，自由现金流转正
After the Federal Reserve enters the quiet period, the things most likely to stir up excitement are actually those that have already been said.
A speech from last week, reappearing with a different title, can easily make people think there is a new policy signal. Now that expectations for a pause in rate hikes are quite high, the market is especially sensitive to wording, and this timing difference in information can easily affect judgment.
My attitude toward this rate decision is: don’t take “no new dissenting voices” as confirmation that the pause is certain. Officials naturally reduce public comments during the quiet period; no one stepping forward to correct the market does not mean the Fed endorses every interpretation.
The data to be released next also can’t be simply interpreted as “good means rise, bad means fall.” If the economy is too strong, it might raise concerns about continued rate hikes; if too weak, it might trigger worries about profits and demand. How prices react also depends on how much expectation has already been priced in.
The most frustrating thing is that once positions are taken, any news afterward can be interpreted as favorable to oneself. When the meeting actually happens, it turns out the market didn’t follow one’s logic at all.
These days, I prefer to first verify the original date of the news before judging whether it changes anything. Truly new information may not be much, but the emotions stirred by repeated dissemination are enough to wear people out.
#10月FOMC临近，美联储暂停加息还有哪些变数？
