Has BTC funds flowed out again? Gold rebounds, SK Hynix opens lower then turns positive! $BTC is currently around $77,000, down about 1% in 24 hours. With escalating US-Iran tensions and ETF net outflows resuming, August profit-taking is starting to loosen. I’m watching for support at 76,000; holding there would mean just consolidation; a quick recovery above 78,000 would signal a halt to the decline, and reclaiming 80,000 would indicate a strengthening trend. $ETH is around $2,390, down about 2% in 24 hours. Spot ETFs have seen net inflows for 12 consecutive trading days, institutions are still buying, but prices remain pressured by high oil prices and high yields. If 2,350–2,370 holds, it can be seen as high-level rotation; only a move back above 2,450 opens new upside. $SKHYNIX is about $161.3 intraday, opening lower and dipping to 157.5 before turning positive, indicating support near 160. A 40 trillion KRW buyback and HBM demand support the long-term logic, but Samsung’s catch-up and rising interest rates will pressure valuations; I continue to hold, watching if 164 can hold in the short term. $XAU is about $4,374, with support at 4,350 and resistance at 4,425; $OKB is about $106.3, must hold 105, only above 110 stops the decline; $QQQ is about $709, pulled back from around 703, but without breaking 715 it’s still consolidation; Friday’s nonfarm payrolls will be the real directional choice. #现货ETF资金回流，BTC与ETH能否接力？

????? Completely crazy, totally crazy Bro, don't be like this I fucking have 20x $NVDA at an average price of 209, damn cnmb I really love your whole family😭

If the US stock market doesn't crash next but instead consolidates sideways for two consecutive months, I actually think it will be harder to trade than a single-day crash. In this kind of market, I won't frequently switch stocks; I will divide companies into three categories. Category one: The more sideways it goes, the more earnings expectations rise. $GOOGL, $META, $AVGO, $MU It doesn't matter if the stock price doesn't rise, as long as EPS, cash flow, and orders keep being revised upward, time is actually on my side. Category two: High elasticity that requires a breakout to confirm. $LITE, $COHR, $ANET, $SNDK I won't rush to add positions just because the price "doesn't fall"; I prefer to wait for a volume breakout at key levels to confirm that capital has truly returned. Category three: Lots of stories, but performance can't keep up. The longer this kind of sideways consolidation lasts, the more cautious I become, because high valuations fear time the most. The market doesn't necessarily need a crash to cut valuations. If the stock price doesn't rise for two years but profits gradually catch up, that itself is a form of valuation cut. So in a choppy market, what I pay most attention to is not who rises 2% today. But who quietly strengthens their fundamentals while the stock price remains unchanged.

Plain as water🥲 Maybe plain and simple is happiness $SNDK $MU $QQQ

I’m not planning to guess whether $AVGO will rise or fall in this earnings report. I’m only preparing three scenarios. Scenario one: AI revenue significantly exceeds expectations + guidance is further raised. In this case, my first reaction might not even be to chase AVGO. I would look at: $MRVL, $ANET, $LITE, $COHR, $MU Because this indicates a very important issue: AI CapEx is continuing to expand from GPUs to ASICs, networking, optical communication, and storage. This could be a confirmation signal for the entire AI infrastructure sector. Scenario two: Good performance, but just meeting the market’s high expectations. This is actually the most interesting. AVGO might have "great earnings but the stock doesn’t rise" because the market has already priced in the good news. At this time, I wouldn’t say the fundamentals have collapsed just because it drops 5% after hours. I would look at: AI revenue growth, orders, customers, and whether the next quarter’s guidance has any substantial changes. If the numbers remain strong but the valuation is just too high, that’s a price issue, not necessarily a company issue. Scenario three: AI growth or guidance is clearly below expectations. This is when you really need to be cautious. Because the market won’t only doubt AVGO. It might further ask: Is ASIC expansion not as fast as expected? Has AI networking been priced in too early? Is hyperscaler CapEx starting to slow down? At that point, $MRVL, $ANET, $LITE, $COHR, and even $MU might all be repriced together. So I think the real importance of this AVGO earnings report is not: Whether it rises or falls the next day. But what it tells us: NVDA has proven AI still needs computing power; can AVGO prove AI infrastructure is still continuing to expand outward? When the market truly crashes, I never first look for: "Who fell the most today?" I look for another kind of stock: Companies with sound fundamentals but indiscriminately sold off by capital. For example, when the market drops the entire tech sector 20% due to oil prices, US bonds, or macro risk-off. At that time, if: $GOOGL’s earnings expectations remain unchanged, $NVDA’s AI orders remain unchanged, $AVGO’s ASIC demand remains unchanged, $MU’s HBM outlook remains unchanged, but their stock prices all fall with the index, that’s the state I’m truly interested in. Conversely, a small-cap stock dropping from 100 to 50 doesn’t mean it’s cheaper than NVDA dropping 20%. Because the market might just finally be admitting: It was never worth 100 dollars in the first place. So in a real crash, what I like to find most are two words: Mispricing. The company’s operations are normal, earnings expectations are normal, industry logic is normal, but because capital needs to reduce risk, good companies and junk companies are sold together. In this kind of market, I’m not in a hurry to guess the bottom. I will slowly pick back up those companies: I originally wanted to hold but thought were too expensive before. The biggest opportunity in a crash is never buying the top decliner. It’s when the market panics, finally willing to sell truly good companies to you at a not-so-ridiculous price.

The craziest IPO of the AI era might be coming. SoftBank's SB Energy has officially applied for a U.S. listing, ticker $SBE, with rumored target valuation exceeding: $50 billion. But the truly outrageous part is: This company currently does not have a single fully operational data center. Revenue in the first half of 2026 is only $139 million, and this income basically still comes from its original energy business; Net loss during the same period: $3.21 billion. Then look at the story it tells the market: Data center backlog is close to: $439 billion. No mistake. Revenue $139 million, backlog $439 billion. A difference of over 3,000 times. Why dare to play like this? Because it is backed by three entities: SoftBank, OpenAI, NVIDIA. NVIDIA plans to invest $1.5 billion in the IPO; OpenAI holds SB Energy warrants worth about $5.5 billion; SoftBank and OpenAI are also its most important customers. Currently, the company has signed or is constructing data center capacity reaching 8.8 GW. But this is also where I think caution is most warranted. This 8.8 GW currently all comes from: OpenAI and SoftBank. And the $439 billion backlog is very long-term. Only about 10% is expected to be realized in the next six years. So this company might become the most interesting IPO to test the AI bubble: Is the market really willing to pay a $50 billion valuation in advance for a company that hasn't truly started generating data center revenue? But from another perspective: If Wall Street is really willing to pay, the signal it sends is equally terrifying. It means the market is no longer pricing AI infrastructure based on: How much money you make today. But rather: How much electricity, land, GPUs, and data center capacity you have locked in for the future. The internet bubble fought for users. The AI era might be fighting for: GW.

$AAOI has secured over 1.1 million square feet of new industrial space in Houston through two 10-year lease agreements. Is the light coming back again?

Today I saw a piece of news, and my first reaction was: GPUs might really become the "oil" of the AI era. CME is preparing to launch two very special futures contracts: NVIDIA H100 rental price futures NVIDIA B200 rental price futures In the future, it won't just be $NVDA stock being traded. Even how much a GPU can be rented for in the next month can be traded and hedged. Why create this product? Because the current pricing for computing power is really chaotic. The same H100 GPU rents for about $7.23 per hour on hyperscalers like Microsoft and Amazon; while on some smaller cloud providers, it’s only about $2.65. The price for the same GPU can differ by nearly 3 times. This is very much like the early oil market: Everyone needs it, but there was no unified price, making it hard to lock in future costs. Then came WTI and Brent. Now computing power might be starting down the same path. What’s even more noteworthy is that BCG estimates the global AI Compute market could reach as high as $2.4 trillion by 2030. So I think the truly impressive thing about this isn’t just the addition of a financial product. It’s that Wall Street is implicitly acknowledging one thing: Computing power has become so important that it needs its own "commodity market." And currently, CME is choosing as the benchmark not AMD, nor some proprietary ASIC. But: NVIDIA H100 and B200. If in the future everyone quotes AI compute power prices like they do a barrel of WTI crude today, then $NVDA’s deepest moat might not even be GPU performance. But rather that it is slowly becoming: The pricing unit for the entire AI compute industry.

Last night the market was not just a normal pullback; oil prices, US Treasuries, and rate hike expectations all pressured risk assets together. $BTC is currently around $77,700, failing to reclaim $79,000 for two consecutive days. Brent crude rose to $95, the 10-year US Treasury yield hit 4.81%, and liquidity expectations have been tightened again. $77,700 is the first support; holding it could lead to another attempt at $78,500–$79,000; if it breaks, look first at $75,500, but don't rush to buy. $ETH is about $2,395, down nearly 3% in 24 hours, underperforming BTC. $2,400 has been lost; the downside target is $2,350–$2,370; only a rebound above $2,430 counts as a stop to the decline, and above $2,480 there is room for a catch-up rally. $SKHY closed yesterday at $160.78, down 2.38%, dragged down more by the drop in the Korean market than company news. $158.5–$160 is the support zone; reclaiming $164 would mean recovery; HBM's long-term logic remains intact, but under high interest rates, valuation digestion is expected first. $SOL is about $100, which is the key line between strength and weakness; $OKB is about $110.1, with support at $108.5 and resistance at $112.5; $QQQ closed yesterday at $707.64, $705 must hold, and only above $715 does it turn strong. The core focus today is whether the 10-year US Treasury yield can fall back from 4.81%. #非农前数据分化，9月加息预期升温 #BTC高位回落，黄金联动受考验