Oli.

Oli.

🍓Web3投研 🍑人工智能 🚀《干翻狗庄》系列工具作者

5Following
281followers

Feed

Oli.
Oli.
Lighter has already returned about 5 million USDC to the affected traders, and deposits and withdrawals have also resumed. This is good news. But what concerns me more is a part of the operation in the review. After the attacker obtained the account API key, they used the public pool's share mechanism to transfer value and finally completed the withdrawal. The "trading key" can access which operations was exposed this time. According to the project team's disclosure, the issue involved the unofficial signature software used by the account repeatedly using the nonce, allowing the attacker to deduce the key from the signature. Subsequently, Lighter adjusted the permissions so that transferring into the public pool requires wallet signature; the API key alone can no longer do it. This modification is very specific and also indicates that there was room for improvement in the permission boundaries between different operations before. Many people, when using quantitative tools, are most concerned about bot profits, latency, and how many accounts can run. Only after a real incident do they realize that the signature program, which is almost never looked at, also holds part of the account's security. Even the most familiar strategy, if implemented differently, can carry different risks. The recovery of 5 million this time is commendable, but it also gave me a very direct reminder. Every action allowed by automated programs in the account should be clearly explained. Especially those keys that sound like they are only responsible for placing orders—it's worth rechecking what funds they can actually mobilize. #Lighter遭API密钥泄露,约500万USDC已返还
Oli.
Oli.
Gold has fallen to its lowest point in over a year, and the first thing that made me stop was the phrase "down more than 30% this year." The latest report's figure of about 31.5% refers to the drop from the historical high in January to the recent low. To calculate the year-to-date decline, you still need to compare with the price at the beginning of the year. Those who bought at the peak are indeed hurting, but these two ways of measuring must be distinguished, otherwise there will be a misjudgment about how much the market has actually fallen. In this round of decline, I am more concerned about the next move of the previous big buyers. The Polish central bank has already reached its 700-ton gold reserve target, but the governor still left room for future purchases. What does reaching the target mean? It could continue buying or adjust the pace. The demand that has appeared continuously in the past cannot simply be projected forward at the original speed. This is the most perplexing aspect of gold. The long-term allocation rationale still makes sense, but the current dollar and yields are putting pressure on holdings. Even if central banks do not sell, buying a little less could cause the short-term market to lose some support. I have always been cautious about the phrase "central banks are all buying, prices will come back sooner or later." Central banks arrange reserves on a very long time scale, but ordinary investors' use of funds may change as soon as next month. After the drop, confirming how long you can hold is more practical than repeatedly searching for the name of a big buyer. #黄金跌至一年低点,年内跌幅已超30%
Oli.
Oli.
The new US overseas computing power regulations have been released, and there is one point worth noting. For certain countries, the new advanced semiconductor licensing requirements provide exceptions for companies whose headquarters or ultimate parent companies are in the US and meet the corresponding reporting requirements. This means that within the same park, using similar equipment, the expansion conditions behind different operators may vary. In the past, when looking at overseas data centers, people often compared electricity prices, land, and network. Now, equity structure, ultimate control, and customer background will also be part of the supply chain review. Even if large US cloud providers and local independent computing power providers receive the same electricity commitments, the procedures they must go through to purchase equipment may not be the same. How the rules are specifically applied still needs to be verified case by case according to the company and business. This makes me more worried about those independent projects that have already invested a lot but still lack follow-up financing. Once equipment procurement timing becomes uncertain, banks may ask more questions, and customers may also require clearer delivery guarantees. After project costs increase, the space to compete for customers based on price will narrow. If we see small cloud providers seeking cooperation with large platforms next, I will not only interpret it as looking for more customers. It may also be an attempt to stabilize equipment supply and financing. After policy changes, larger companies have the opportunity to gain advantages more important than procurement discounts. #美商务部新规限制海外AI算力租赁
Oli.
Oli.
The US and Iran have clashed again, but oil prices haven't surged all the way up. This kind of market situation most easily makes people suspect whether the market has become numb. After reading the latest news, I think traders are calculating two accounts at the same time. Attacks in the Strait will disrupt transportation, while some Middle East exports resuming might increase market supply. The risk is rising, but the available crude oil might also increase; these two forces are colliding. A report on November 17 mentioned that the US struck an Iranian missile launch site, and mediators are still in contact. For the oil market, the next thing to watch is whether military actions hit facilities that are harder to restore. If a clash has limited impact, prices might quickly retract the initial gains. If ports, storage, or other energy facilities are damaged, repairs will take time, and the supply situation will change noticeably. I am now a bit cautious about the reasoning that "firing should push oil prices up." The market has already priced in a lot of risk premium for this conflict; new news has to be worse than previously feared to keep pushing prices higher. What is truly worth worrying about is damaged production capacity that takes a long time to recover, while spot buyers start paying premiums for alternative sources. Only then will the impulses in the futures market more easily turn into continuous buying over several days. #美伊再度交火,原油供应风险会否上升?
Oli.
Oli.
What makes this downturn most uncomfortable for me is that "cheap" is becoming increasingly hard to judge. BTC has fallen from its highs, looking like a discount. The stock prices of companies holding coins have dropped even deeper, appearing to be an even bigger discount. But with the same amount of money, buying coins directly and buying a financing coin-holding company involve quite different risks down the line. Strategy spent about $33.4 million buying coins in the latest week, while repurchasing STRC cost about $228 million. This arrangement quite well illustrates the current situation. The company wants to maintain BTC exposure but also has to take care of its financing instruments. As coin prices fall, assets shrink, yet cash dividends and financing arrangements still need to be managed. Shareholders are waiting for BTC to rebound, but the company must first get through this period well. So after $800 billion in market value evaporated, I think it's worth re-examining exactly what you hold. Those directly holding BTC can choose to wait. Those holding financing coin companies or leveraged products need to consider issuance, dividends, refinancing, and even how daily settlements change the path to break-even. The most frustrating scenario is when BTC really does rise back, but your position hasn’t recovered. Doing your homework now still gives you a chance to calculate slowly; when prices suddenly rebound, emotions easily overshadow these details. #加密市场缩水8000亿美元,这轮下跌有何不同?
Oli.
Oli.
NVIDIA's CEO search has finally made concrete progress. The chairman confirmed at the earnings call that they have hired Heidrick & Struggles to assist in finding a successor and clearly stated that Jensen Huang will not be rehired. After days of speculation about "whether he will return," at least now the focus can shift forward a bit. Rather than continuing to guess names, I paid more attention to the arrangement of this call. Interim CEO Colette Kress was responsible for discussing operations, and Ian Buck, who oversees hyperscale computing and high-performance computing businesses, also participated. This doesn't directly indicate who will succeed, but at least it gives outsiders a chance to see who, besides the founder, can clearly articulate the technical roadmap and customer needs. The real challenge for the new CEO after taking office is earning this kind of trust. Customers sign multi-year procurement plans and must believe that the next-generation products will be delivered on time and that the team can handle any issues that arise. It's hard for one person to immediately replicate the relationships Jensen Huang built over decades; it can only be done through the management team repeatedly fulfilling promises. I hope the ongoing search process allows the company enough time. The quarterly revenue of $115.8 billion already shows the business is still running. Going forward, whoever can keep the core R&D team and customers willing to follow will have a better chance of securing this position. The names the market guesses today may not be the ones the board ultimately needs. #英伟达下一任CEO人选会是谁?
Oli.
Oli.
The December rate hike has almost been priced in by the market, with the probability increasing by one or two points daily; there is actually very little new information to gain. What I care more about now is the interest rate expectations for next spring. One rate hike in December followed by a pause to observe, versus continuing to raise rates after December, represent two completely different funding environments for risk assets. Especially in the crypto space, where it’s common to trade “bad news priced in” in advance. If a 25 basis point hike does happen in December, but the statement and subsequent remarks lead the market to continue raising next year’s rate expectations, the relief might not even have time to set in before positions need to be repriced again. The cost of borrowing being a bit higher affects how long positions can be held and how much projects can still raise. So when I see numbers close to 97%, I’m not particularly excited. It indicates that this upcoming meeting already has very crowded expectations, and we still need to see which way those expectations move next. If the upcoming inflation data only meets forecasts, it may not be enough to roll back bets on rate hikes next year. For capital to be willing to extend holding periods again, the market needs to see inflationary pressures easing continuously. A single decision meeting that meets expectations can bring a few hours of rebound, but turning those few hours into weeks still requires follow-up data. #美联储12月加息概率近97%,加息预期延伸至明春
Oli.
Oli.
Anthropic's subscription is nearly 7 times the issuance size, and there are rumors that the pricing may be higher than the original $70–78 range. What I find most subtle is that there is also a $26 billion placement alongside. Participants include companies like Nvidia, Microsoft, and Google, while Anthropic itself is a major customer of chips and cloud computing power. This, of course, has business logic. Suppliers want major customers to have enough money to continue expanding, and model companies also need long-term partners. But from the perspective of IPO investors, I want to look more closely at the subsequent revenue. How much money enterprise customers pay for Claude, and how long the funds from partners can support computing power expenses—both of these will affect its future cash burn. If growth is rapid after listing, and computing power commitments keep increasing, the company may have to continuously seek the next round of funding. At that time, it is worth calculating how long today's oversubscription can provide a buffer. I understand that the market is willing to pay for scarcity now. But every increase in the issue price means the profits public shareholders wait for will be further away. When real money is subscribed, everyone is enthusiastic. When the financial report needs to explain tens of billions of dollars in expenditures, patience begins to be tested. #AnthropicIPO进展:认购订单接近发行规模7倍
Oli.
Oli.
NVIDIA delivered $115.8 billion in quarterly revenue this time, a year-over-year increase of 103%, with next quarter's revenue guidance midpoint at $130 billion. Yesterday we were still discussing whether it could surpass $100 billion, and today the figure is significantly higher. Demand is so strong that there's really nothing to argue about. What caught my eye even more was the gross margin. The adjusted gross margin for Q3 was 74.9%, but the guidance for next quarter drops to 71%–72%. Revenue continues to climb, but how much gross profit is retained per dollar of sales is starting to change. How much it will cost to scale the new platform and whether those costs can be recouped in the following quarters will affect how high valuations can go. This is also why I felt a bit excited yet couldn't fully relax after reading the earnings report. Vera Rubin has already started shipping; the business hasn't stalled amid Jensen Huang's departure news. But the more customers buy, the more complex production and delivery become during the company's transition. Moving forward, even if revenue continues to exceed expectations, if gross margin doesn't recover soon, analysts will still have to recalculate profits. The $100 billion revenue milestone has been achieved, but the pressure ahead grows accordingly. Buying this stock now means accepting that the market will start measuring it with a stricter yardstick. #英伟达Q3营收首破千亿美元超预期
Oli.
Oli.
Strategy bought another 485 BTC, and Strive bought 251. Even with such a drop, companies are still adding to their positions, which is indeed encouraging. But when reviewing the recent holdings announcements, I noticed a timing discrepancy. The company's treasury disclosed recent purchases, while the newly released institutional 13F filings record positions as of the end of September. Putting these two types of information in the same news easily gives the impression that "everyone is bottom-fishing together." For example, Harvard held about 3.04 million shares of IBIT at the end of Q3, the same number as before. The increase in holding value mainly came from price appreciation during the quarter. As for whether it bought after the November drop, this document does not answer that. On Strategy's side, it's clear: they spent $33.4 million on buying coins during the same period and about $228 million on repurchasing STRC, showing the priority in their capital allocation. I still think it's good that companies are willing to keep holding coins continuously, but now the bigger concern is how much cash they still have on hand and how much they can continue to buy after financing. Companies are willing to bear volatility, but ordinary shareholders still have to endure the impact of financing and dividend arrangements. It's normal to get excited for a few seconds when seeing big institutional names, but if you really want to place orders based on this, at least first confirm when they bought and exactly how much more they bought this time. #BTC持续下跌,财库与机构仍在配置