OpenAI's IPO delay may cause the most discomfort not on Wall Street, but among employees and early investors who have held shares for years and are waiting to cash out. Private companies can avoid quarterly performance pressure but cannot permanently avoid liquidity issues. Employee compensation includes a large amount of equity, and funds have expiration periods. If the public listing continues to be postponed, the company will need to arrange more share buybacks or secondary market transactions; otherwise, no matter how high the book wealth is, it will be difficult to truly convert it into usable cash. The secondary market can temporarily relieve pressure but will create another pricing system. Different batches of investors may buy and sell at different discounts, and ordinary employees usually have less information, bargaining power, and trading opportunities than large institutions. The longer the IPO is delayed, the harder it is to avoid fairness issues in internal liquidity. Sam Altman says that going public now is not conducive to AI safety, and this reason deserves serious consideration. But choosing not to go public also means the company needs to provide a more complete alternative plan so that employees and early shareholders do not have to endlessly subsidize the company's long-term goals with patience. Safety can require capital to wait, but it cannot require all shareholders to live forever just watching valuation numbers. #OpenAICEO称2026年不会IPO

A low-profile fund with only a few investment staff accumulated about $40 billion in book value from a single SpaceX holding. Vy Capital holds about 3.4% of SpaceX shares, making it one of the company's fifth largest shareholders. This is the most attractive form of concentrated investment: acquiring scarce assets before the market consensus forms, then waiting for the company to grow over more than a decade, ultimately supporting the entire institution's reputation and returns with a single investment. But $40 billion in holdings does not equal $40 billion in cash. SpaceX shares have limited liquidity, and Vy Capital finds it difficult to exit quickly without affecting the price and market expectations. The more successful the holding, the more the fund depends on a single company, Elon Musk personally, and the pace of SpaceX's IPO. What is most worth learning from this investment is not "heavy holdings guarantee wealth," but what is needed before taking a heavy position. Early access rights, long-term capital, board trust, and a capital structure that can endure years without exit are all indispensable. Ordinary people see the multiples of returns; what is truly hard to replicate are the holding conditions. Concentrated wealth creation also turns a company into the fate of the entire institution. #SpaceX股东VyCapital披露约400亿美元持仓

Oracle's $664 billion unfulfilled contracts are quite staggering, but one easily overlooked detail is that some major AI clients have prepaid for GPU funds or directly provided GPUs to Oracle for deployment. This changes how we understand the orders. Traditional cloud providers usually spend their own money to build data centers first, then wait for customers to gradually use them. Customers prepaying or bringing their own hardware means they actively share Oracle's heaviest capital burden, also indicating these clients have a strong commitment to computing power demand. Thus, the orders are not just verbal intentions but come with real money. On the other hand, it's also clear that not many clients are willing to prepay huge sums; orders may be concentrated among a few large AI companies. If any one of them adjusts their model strategy, delays data centers, or faces financing difficulties, the revenue realization pace will be noticeably affected. Therefore, when evaluating Oracle, you can't simply divide $664 billion by several years to calculate revenue. You need to look at the prepayment ratio, number of clients, cancellation terms, and capacity online timing. This contract backlog is solid but not necessarily sufficiently diversified. The huge orders clarify the future but also make the importance of a few clients glaring. #财报观察员：甲骨文AI云收入增121%

Robinhood's crypto trading volume in August increased 61% month-over-month but still dropped 38% year-over-year. Putting these two figures together tells the full story. The sharp month-over-month rise indicates trading sentiment has recovered from a low point, and Robinhood and Bitstamp successfully caught a wave of volatility; the year-over-year decline shows current activity remains far below the previous peak. This is good news for the platform but the valuation cannot be based on only the more favorable half. The most intoxicating aspect of trading platforms is operating leverage. When the market is hot, users trade more frequently, and the company doesn't have to increase costs proportionally, so revenue can quickly amplify. The problem is this process reverses just as fast. When volatility disappears, trading volume, fees, and market enthusiasm can all shrink simultaneously. Therefore, what Robinhood truly needs to prove is that it no longer relies solely on crypto bull markets to survive. In August, stock trading volume grew 68% year-over-year, and event contract numbers increased about 15 times year-over-year; the platform is actively diversifying its revenue sources. The market can get excited about a 61% rebound, but long-term valuation should be based on the full cycle. Only those who can continue to make money during quiet months deserve a valuation as financial infrastructure. #Robinhood加密交易量8月环比增61%

U.S. Treasury yields approach 5%, and the most easily overlooked risk is not stock valuations, but that bonds themselves are becoming a shrinking collateral. Rising yields mean falling prices for old bonds. Banks, insurance companies, pension funds, and leveraged funds holding long-term U.S. Treasuries will face unrealized losses. If these bonds are used for repo financing or derivative margin, continued price declines could trigger margin calls, deleveraging, and forced selling. This is the real danger of the 5% level. It not only signals investors that "risk-free returns are higher," but may also force those who originally did not want to sell to liquidate assets. The 2022 UK pension crisis already demonstrated a similar chain: bond prices fall, margin requirements rise, forced bond sales, then further bond price declines. The U.S. Treasury's expansion of repo operations can improve liquidity for old bonds but cannot prevent all institutions' holdings from being revalued. The market may appear calm on the surface, but that does not mean pressure is absent. Problems in the bond market often hide first on balance sheets, and only when a highly leveraged participant can no longer bear it do people suddenly realize that so-called safe assets can also create liquidity crises. #美债收益率逼近5%，回购难缓长期压力

BTC's daily new supply is inherently limited, and the ETF outflow of nearly $450 million over three days brings marginal selling pressure that is far more significant than the numbers suggest. ETFs don't need to control all BTC to potentially impact short-term prices. Market prices are determined by the last batch of buyers and sellers at the margin, not by wallets that remain inactive for a decade. When ETFs continuously redeem, market makers need to reduce exposure, and spot and futures hedges adjust accordingly, causing selling pressure to propagate across multiple markets. This also explains why, even though ETFs only account for a portion of BTC supply, fund flows can frequently become the core of market movements. The vast majority of coins are not traded; the truly liquid coins involved in daily pricing are far less than the total supply. Hundreds of millions of dollars flowing out may not be huge relative to the entire BTC market cap, but it is entirely different when considered against the daily tradable coins. However, I will not turn long-term bearish just because of three days of outflows. What needs to be confirmed is whether this redemption continues beyond a macro event window and whether miners and long-term holders are simultaneously increasing sales. ETFs make BTC more accessible to global capital but also allow traditional market panic to arrive faster. Institutionalization has never been a one-way positive. #BTC现货ETF三日流出近4.5亿美元

What the Strait of Hormuz truly lost is not just the ability for ships to pass, but also everyone's confidence that negotiations can resolve the issues. An Iranian cargo ship was attacked again, and the regional meeting originally scheduled to discuss temporary shipping arrangements was subsequently postponed. Commercial vessels fear this situation the most: the route is not officially closed, yet no one dares to guarantee the safety of the next ship passing through. Shipowners, insurance companies, and cargo owners will not wait for diplomatic statements; they will first raise premiums, suspend voyages, or choose longer alternative routes. This means that even if the actual daily loss of crude oil quantity does not continue to expand, the risk cost may still rise. Oil price trading is no longer just about supply gaps but about how long the crisis will last. What’s more troublesome is that the Red Sea and Saudi Arabia’s east-west oil pipelines are also under threat. Routes that could bypass Hormuz in the past are now unsafe themselves. So-called alternative solutions are failing one by one. The market always likes to expect that a meeting or a statement can bring oil prices down. But when the negotiation table cannot even be set up on time, shipping companies can only prepare for the worst-case scenario. What now affects prices is not the next attack itself, but that everyone begins to believe attacks will continue to happen. #霍尔木兹船只再遇袭，地区会谈推迟

Trump accepts the new version of the crypto ethics clause, and the toughest hurdle for CLARITY has finally loosened. But this does not mean the bill is secure yet. The new plan requires the president, members of Congress, and related personnel to handle significant crypto conflicts of interest, possibly requiring them to sell holdings or place them in blind trusts. It touches on an issue the industry has long avoided: can those who make the rules also profit from the assets covered by those rules? I support writing ethical restrictions into the bill. For crypto regulation to gain long-term legitimacy, the public must not feel that the rules are tailored for a few power holders and family projects. No matter how professional other parts of the bill are, if conflicts of interest are unclear, it will ultimately become a target for political attacks. But including the clause is only the first step. Who is responsible for investigation, what counts as a "significant interest," how family and related entities are calculated, and whether there are real penalties for violations—these details determine whether it is a firewall or just decoration. CLARITY needs not only regulatory clarity but also clarity regarding legislators' interests. If the industry only cares about who regulates the tokens but not who profits from the legislation, it will sooner or later pay the price for this selective blindness. #特朗普接受新版伦理条款，CLARITY投票临近

Anthropic chooses Nasdaq, while OpenAI decides to postpone its IPO. The two most watched AI companies are giving the capital market two completely different answers. Anthropic calls for slowing down frontier model development while continuing to push for a 2026 IPO. It seems contradictory, but it is very realistic: safety research, computing power procurement, and talent competition all require huge funds. The more worried about technology getting out of control, the more money is needed to build testing, auditing, and protection systems. But after going public, another kind of pressure will also appear. Quarterly revenue, valuation, and stock price will force the company to continuously release stronger models. When the safety team says "wait," the capital market may ask "why is growth slowing down." What Anthropic needs to prove is not just how much Claude can earn, but whether the public market can accept an AI company that actively puts the brakes on itself. I actually look forward to its IPO. AI companies constantly talk about impacting all humanity, yet for a long time only disclose their operations to a few private equity shareholders, which is unhealthy. The public market is noisy, but auditing, governance, and continuous disclosure at least allow outsiders to see who bears the cost. #Anthropic拟赴纳斯达克IPO

If the market has already priced in a 25 basis point rate hike, the real driver of volatility will no longer be "whether to hike or not," but how far the hikes will go after that. In a Reuters survey, 86 out of 101 economists expect the Federal Reserve to raise rates this week, pushing the rate to 3.75% to 4.00%. Such crowded expectations mean the rate hike itself may only trigger a brief reaction. The market is more concerned about whether the dot plot will continue to shift upward, whether the statement emphasizes energy inflation, and how many more actions might occur before March next year. The ones most likely to lose out are those who only bet on the meeting outcome. Even if the Fed hikes rates, if it signals a "wait and see" approach, U.S. Treasury yields and risk assets could reverse and recover; even if it holds steady, as long as Powell emphasizes continued tightening ahead, the market will still struggle. This meeting is like a movie with a spoiler revealing the ending; the real value lies in the last ten minutes. BTC, gold, and U.S. stocks are not trading on that 25 basis points, but on the entire path of funding costs over the coming months. Don't just focus on the news headlines. At the moment the result lands, the market has often already started trading the next meeting. #本周FOMC揭晓，加息能否落地？