Institutionalization of ZEC has not eliminated volatility; instead, it has equipped volatility with a bigger engine. After the Grayscale spot Zcash ETF started trading, traditional accounts gained easier access to ZEC. Institutional funds, privacy narratives, and short-covering collectively drove the market, with ZEC briefly breaking above $1000. Approximately $36.6 million in leveraged positions were liquidated in a single day, the vast majority of which were shorts. However, after the peak, long leverage also began to be cleared. This indicates that the so-called "institutional entry" never means someone is backing the price floor. Institutions bring not only long-term capital but also arbitrage, hedging, ETF creations and redemptions, and more complex derivative positions. Liquidity deepens, and the liquidation chain also lengthens. Privacy demand, ETF channels, and improved regulatory environment have indeed given ZEC a pricing foundation it lacked before. But when the price rises far faster than real usage growth, it is easily driven by perpetual contracts and liquidations. That kind of rally feels great but most easily misleads people into thinking they understand the fundamentals. Institutionalization means ZEC has entered a bigger casino; it does not mean the house suddenly became a philanthropist. What should be most watched at the top is whether the spot market can absorb the leverage being liquidated. #ZEC机构资金入场，高位杠杆开始出清

SpaceX is increasingly like a super infrastructure company disguised in aerospace clothing. The CFO disclosed a new AI computing power hosting agreement that can bring in about $1.11 billion in monthly revenue starting in December, and reiterated the goal of hitting $100 billion in annual recurring revenue by the end of the year. The market sees growth, but what I see is the company's identity rapidly changing. Rockets, satellite internet, ground data centers, chips, and future orbital computing are all capital-consuming businesses. They can synergize with each other, but may also compete simultaneously for power, equipment, engineering teams, and the balance sheet. AI hosting revenue is tempting, but computing power construction requires continuous investment. If catching up with demand slows down rocket or Starlink projects, the short-term high revenue could instead harm the deepest moat. What makes SpaceX truly powerful is combining launch, communication, and computing into an infrastructure network. What makes it truly dangerous is the ambition so vast that each part requires astronomical funding. $100 billion ARR sounds like the finish line, but it is actually the starting point of another capital consumption battle. What we need to watch next is whether the AI business can provide blood transfusion to the aerospace main business, rather than dragging the main business into endless expansion. #SpaceXCFO称有信心实现1000亿美元ARR

OpenAI has secretly filed for an IPO but decided not to go public in 2026. Sam Altman’s reason is that AI safety work is not yet complete. I am willing to believe that safety issues do indeed exist, but postponing the IPO also has another consequence: the public will continue to be unable to fully see the company’s financial status, related-party transactions, and capital consumption. Going public brings quarterly performance pressure, which may force the company to accelerate the release of immature models for revenue; but going public also means stricter disclosure, auditing, and governance. Staying in the private market allows management to avoid short-term stock price noise and also avoid the most direct scrutiny of the public market. This is exactly the hardest contradiction for OpenAI to handle. It says its technology could impact all humanity, yet development speed is still mainly decided by a few managers and large capital. Safety cannot rely solely on the CEO’s conscience, and commercial oversight cannot rely solely on stock prices. Pausing the IPO may be a responsible brake or it may be a way to buy time when valuation, losses, and market conditions are not ideal. It is too early to draw conclusions now, but since the company proactively cites “safety” as the reason, it should publicly disclose more specific safety goals and verification standards in the future. #OpenAICEO称2026年不会IPO

Chip suppliers are preparing to become anchor investors worth billions of dollars in their customers' IPOs, which is more than just "being optimistic about AI." According to reports, Nvidia is discussing investing up to $10 billion in Anthropic's potential IPO. Anthropic hopes to raise huge funds through the listing while continuously purchasing computing power driven by Nvidia GPUs. Nvidia invests capital, Anthropic gains expansion funds, and the money may then flow back into the Nvidia ecosystem through cloud services and GPU procurement. This arrangement does not necessarily represent false demand, but it does blur the price discovery function that an IPO originally carries. When the largest supplier also becomes a major investor, does the issue price truly reflect independent market demand, or is it a growth expectation jointly maintained by members within the ecosystem? Nvidia certainly has ample reasons to do this. Supporting Anthropic can expand CUDA usage and also prevent the AI model market from being monopolized by a single company. But investors need to see clearly that it is transforming from a chip-selling company into a bank, underwriter, and risk co-bearer of the AI industry. When winning, it can capture profits from the entire industry chain; if the customer's financing ability weakens, the risk will also return along the same chain. #Anthropic拟赴纳斯达克IPO

Robinhood's crypto trading volume increased by 61% month-over-month in August, but the number of funded accounts only grew by about 120,000 during the same period. This combination is quite interesting. Trading volume surged sharply, but the user count did not explode in tandem, indicating that this growth is more likely driven by existing users increasing their trading frequency and institutional and overseas traffic brought in by Bitstamp, rather than a large influx of new retail investors suddenly entering the market. The group's crypto trading volume reached $17.5 billion, with Bitstamp contributing $10.1 billion and the Robinhood App contributing $7.4 billion. Mergers and acquisitions are helping Robinhood scale up, but they also make the "Robinhood trading volume" metric more complex. Going forward, when looking at this company, one cannot just look at a single total figure; it is necessary to distinguish whether growth comes from the App, institutional clients, or consolidated acquired assets. I actually like this change. Relying solely on U.S. retail sentiment means revenue quickly shrinks when the market cools; after integrating Bitstamp, the company now has a broader customer base and liquidity sources. But the real test is just beginning. Trading volume can be rapidly amplified by volatility, but the real challenge is converting these active users into stable assets, subscriptions, and interest income. Buzz is valuable, but retention determines valuation. #Robinhood加密交易量8月环比增61%

When the 10-year US Treasury yield approaches 5%, even if the Federal Reserve does not raise interest rates, the financial environment is already tightening on its own. Corporate loans, mortgages, and asset valuations do not only reference policy rates; a large amount of long-term financing will be repriced around Treasury yields. With US Treasuries near 5%, it means companies wanting to issue bonds for expansion need to pay higher interest; investors buying stocks will also demand higher returns to compensate for risk. This is the most conflicted aspect of the current market. Everyone is watching whether the Fed will raise by 25 basis points, but the bond market may have already completed a larger scale tightening on its behalf. The Treasury's expansion of long-term bond repurchases can only make trading smoother; it cannot eliminate energy inflation, fiscal deficits, or new bond supply. For the crypto market, a near 5% risk-free yield is especially brutal. Funds can just sit in Treasuries to earn interest and will not easily pay for distant stories and high volatility. BTC must prove it offers more than just upside potential, and tech companies must deliver real cash flow. What truly burdens the market may not be a single rate hike, but the inability to lower long-term funding costs. #美债收益率逼近5%，回购难缓长期压力

The average diesel price in the United States has surpassed $6 per gallon for the first time, and many people who don't drive diesel vehicles think it doesn't concern them. Soon, they will realize that even though diesel isn't in their gas tanks, it is hidden in almost every bill. Truck transportation, agricultural machinery, construction equipment, cold chain logistics, and express delivery all heavily rely on diesel. After fuel prices rise, transportation companies won't absorb the costs out of thin air; instead, they will pass them on to consumers layer by layer through fuel surcharges, price increases on goods, and reduced service frequency. This is especially true for meat, vegetables, and frequently restocked items, whose prices often react faster than official inflation data. Currently, the average diesel price nationwide is about $6.06, compared to approximately $3.71 a year ago. Such a drastic change is not a normal fluctuation but an invisible tax covering the entire supply chain. What's more troublesome is that the Federal Reserve cannot produce diesel by raising interest rates. It can only suppress demand, causing businesses and consumers to spend less. This creates a frustrating situation: living costs are pushed up by energy prices, and financing costs are increased by monetary policy, hitting ordinary people twice. Diesel breaking $6 is not just gas station news; it is a preview of the next round of price pressures. #美国柴油价格首次突破6美元

Oracle's AI cloud revenue surged 121%, but a software company that once consistently generated cash is now starting to show negative free cash flow. This is the most exciting and also the most dangerous change in my view. This quarter, Oracle's OCI revenue reached $7.4 billion, with unfulfilled contracts totaling $664 billion. Orders are very strong, but to convert these orders into revenue, the company must build data centers in advance, purchase GPUs, secure power, and sign long-term leases. The income statement still retains the high-growth imagination of a software company, but the cash flow increasingly resembles that of a heavy-asset infrastructure business. The issue is not whether AI demand is real or not, but whether the realization speed can outpace capital consumption. As long as customers use the capacity on schedule, these investments will turn into astonishing growth leverage; if model efficiency improves, customers delay deployment, or demand concentrates among a few large clients, the data centers and leases will not disappear accordingly. The capital market liked Oracle in the past because of stable databases, high profits, and predictable cash flow. Now it is using that stability to exchange for a second growth phase in the AI era. This transformation is very brave, but there is no turning back. #财报观察员：甲骨文AI云收入增121%

"BTC现货ETF三日流出近4.5亿美元" is scary, but if you extend the time window a bit, the story changes immediately. From September 8 to 10, the ETF outflows were about 46.6 million, 120.2 million, and 282.7 million USD respectively, totaling nearly 450 million USD over three days. However, as of September 10, the entire month still saw a net inflow of about 320 million USD. The same set of data, when taken over three days, can be described as "institutional retreat," but when counted from the beginning of the month, it’s "funds are still flowing in." This is where market sentiment is most easily manipulated: the numbers don’t lie, but those who choose which segment of numbers to show might be lying. What I’m more concerned about now is who is absorbing after the sell-off. If the ETF outflow is close to 450 million USD, but the price hasn’t experienced a corresponding sharp drop, it means off-exchange funds, long-term holders, or buyers from other regions are absorbing the chips. This kind of support is more important than the net inflow on any single day. ETF funds are not always the right money; they also chase rallies, cut losses, and rebalance portfolios. Don’t treat inflows as faith, and don’t treat outflows as a verdict. The real strength or weakness is hidden in the price reaction after selling appears. #BTC现货ETF三日流出近4.5亿美元

Institutions raising rate hike expectations does not mean the Federal Reserve has already decided to raise rates. After the release of PPI and CPI, Wall Street quickly revised its forecasts. This move looks professional but actually reveals the most genuine side of institutional forecasts: often they are not predicting the future, but rather reducing the risk of being contrary to market consensus. After the data heats up, if they still insist on no rate hike, once the Fed really acts, fund managers will find it hard to explain to clients; by collectively raising expectations, even if the final judgment is wrong, it can be attributed to sudden changes. This kind of "career-safe forecasting" causes market probabilities to suddenly crowd to one side, but it does not mean the outcome is already determined. What is truly worth observing is how the Fed judges supply shocks. If inflation mainly comes from energy and transportation, further rate hikes will only first hit demand, employment, and financing, and may not bring down commodity prices. The worst scenario is not a single rate hike, but the central bank neither being able to suppress inflation nor daring to stop tightening. The market is currently trading on policy anxiety, not policy answers. The closer to the meeting, the more cautious you should be when everyone stands on the same side. #PPI、CPI公布后，多家机构上调9月加息预期