A.  Treasury yields staying above 5% My vote is A — but the real signal is not the 5% number itself. It is whether 5% becomes the new floor. When Treasury yields stay elevated, stocks face a tougher hurdle: valuations must compete with a relatively high risk-free return, while corporate refinancing costs also rise. This is especially important for long-duration growth stocks whose value depends heavily on future cash flows. The bigger risk is the chain reaction: oil stays expensive → inflation remains sticky → rate cuts get pushed back → Treasury yields stay high. What makes this cycle interesting is that AI is not completely insulated. Hyperscalers have issued roughly $220 billion of bonds amid massive data-center investment, adding another source of borrowing demand So I’m watching

I’d choose B — but with a bigger interpretation. AMD’s $8.2B World Labs deal isn’t simply about 3D graphics. It is a bet that the next AI frontier will be world models—systems that can understand space, simulate environments and eventually interact with the physical world. That matters because AI workloads could evolve from generating tokens to generating entire environments for robots, autonomous systems and industrial simulation. The strategic value for AMD is even deeper: owning the model layer gives AMD visibility into what future AI workloads will demand from chips, memory, networking and software. So I see this less as “AMD buying a startup” and more as AMD trying to move up the AI stack—from selling compute to helping define what the next generation of compute actually looks like.

BitMine crossing 6 million ETH is impressive, but the bigger story may be what happens after the accumulation. With 5.07 million ETH already staked, BitMine is turning a massive ETH treasury into a potential recurring-yield engine. The company currently cites a 2.62% annualized 7-day staking yield and projects about $358 million in annualized staking revenue. But investors should separate ETH appreciation from staking economics. A 2.62% yield sounds attractive until ETH falls 20%—staking income cannot fully offset a large decline in the underlying asset. That makes BMNR an interesting experiment: it is not simply betting on ETH’s price, but on ETH as a productive treasury asset. The real question is whether staking rewards can compound the treasury faster than dilution, financing costs an

I’d choose B: More AI infrastructure & R&D. Nvidia’s $150B buyback authorization is impressive, but I wouldn’t rush to maximize shareholder returns today. The bigger opportunity is still in extending its AI moat. Nvidia is sitting at the center of a massive infrastructure cycle. Every dollar reinvested into next-generation GPUs, networking, software and systems could potentially strengthen its ecosystem and extend its competitive advantage. Buybacks reduce the share count, but R&D and infrastructure investment can potentially increase the size of the future cash-flow engine itself. That said, this isn’t an argument against buybacks. Nvidia can do both. If free cash flow continues expanding rapidly, management has the flexibility to invest aggressively while repurchasing shares

B — 100K–200K. I expect September payroll growth to land around 120K–150K. August’s 162K gain showed the labor market still has some resilience, but the broader trend is clearly cooling, while private hiring has been relatively subdued. The interesting part is that “good jobs data” may not mean good news for stocks. A strong print could push October hike expectations higher, lifting Treasury yields and pressuring high-duration tech valuations. But a moderate slowdown could be the sweet spot: enough cooling to reduce rate pressure without triggering recession fears. With oil prices already adding inflation risk, I’m watching wages and unemployment more than the headline payroll number. If payrolls come in around 130K with wage growth cooling, markets may interpret it as a soft landing sign

[你懂的]   $Arteris (AIP): The Company Building the “Highways” Inside AI Chips Let’s start with the simplest explanation: Arteris doesn’t manufacture chips. It provides the IP that helps different parts of a chip communicate with each other. Modern AI chips can contain CPUs, GPUs/NPUs, memory controllers, caches, accelerators, security blocks, and I/O interfaces — all of which need to move massive amounts of data. As chips become more complex, especially with the rise of Chiplets and multi-die architectures, moving data efficiently inside the chip becomes a major engineering challenge. That is where Arteris comes in. Its Network-on-Chip (NoC) technology can essentially be viewed as the highway system inside a chip, helping different IP blocks communicate efficiently while balan

$纳指100ETF(QQQ)$  [微笑]  QQQ Drops 1%+ — Can Elevated Yields Break the Tech Bull? Monday’s pullback in $QQQ is easy to explain: yields are rising, oil is adding to inflation pressure, and investors are taking profits in high-beta tech. But I think the more interesting question is not whether higher yields hurt technology — they obviously do. The real question is whether AI-driven earnings growth is strong enough to absorb the higher cost of capital. The 10-year Treasury yield climbed to around 5.23%, its highest level since 2007, while the Nasdaq Composite fell 0.9% on Monday. The move came as oil prices and geopolitical uncertainty revived inflation concerns and pushed expectations for further Fed tightening

[思考]  Higher for Longer doesn’t scare me. It changes what I’m willing to pay for. If I had $10,000 to invest today and believed interest rates would remain elevated for longer, I wouldn’t simply move everything into cash—or try to perfectly time the next Fed move. My first question would be: What can still compound earnings and cash flow when the cost of money stays high? That distinction matters. When risk-free yields are attractive, investors no longer have to pay any price for growth. Higher rates can pressure long-duration assets, highly leveraged companies and businesses whose valuations depend heavily on profits far into the future. But that doesn’t mean every growth company becomes unattractive. It means quality, cash flow and pricing power become more valuable. 💰 How woul

My answers: 1-B, 2-B, 3-B, 4-C, 5-B, 6-C, 7-B, 8-B, 9-A, 10-D. The biggest lesson isn’t simply “margin gives you more buying power.” It’s that leverage magnifies both opportunity and risk. An unused margin limit itself doesn’t create interest—the interest comes from the amount actually borrowed. A margin account can also provide buying power before sale proceeds settle, subject to eligibility and available margin. The calculation in Q7 is a good reality check: USD10,000 × 7.99% × 10/360 ≈ USD22.19. But Q8 is the one investors should remember: a USD20,000 position funded with USD10,000 of your own capital loses USD2,000 after a 10% decline—a 20% hit to your own money. @Tiger_AU [正经]

B — CRWV Wall Street’s biggest lesson last week wasn’t about who was bullish or bearish — it was about what investors are actually paying for in AI infrastructure. CoreWeave is the perfect example. Rothschild Redburn focused on leverage, capital intensity and valuation, while JPMorgan argued that rising compute prices and improving margins could outweigh the debt burden. That disagreement matters because AI infrastructure is entering a new phase: demand alone is no longer enough; pricing power, capital efficiency and cash-flow conversion will decide who captures the economics. Microsoft’s upgrade reinforces the same idea from another angle — AI winners need to turn massive infrastructure spending into durable revenue and margins. So I’d watch CRWV less for the $54 vs $125 debate, and more

My pick: Green (5% to 10%) MU’s setup is stronger than a simple “beat the quarter” story. Micron’s own Q4 guide was already $50B revenue and $31 EPS, while the market has pushed expectations higher. The key catalyst is forward visibility. Micron has signed 16 strategic customer agreements, with roughly $22B in cash commitments and many contracts extending through 2030. That changes the traditional memory-cycle equation: if HBM demand remains tight while long-term contracts protect pricing, earnings could stay elevated longer than the market expects. My concern is valuation and expectations—MU now needs not just a beat, but strong FY2027 guidance. **I expect a solid reaction, but probably not a >10% blowout.** @Tiger_Earnings [思考]

C. The bigger opportunity will be AI jobs and talent Singapore’s AI story may ultimately be less about how many AI giants open offices, and more about how deeply AI reshapes the workforce. The early data is encouraging: AI-related job postings have risen sharply, but the opportunity is spreading beyond pure AI engineering. Finance, sales, consulting, cybersecurity and operations increasingly need people who can combine domain expertise with AI tools. That could be Singapore’s real advantage. A small country cannot compete with every market on scale, but it can compete on talent density, enterprise adoption and regional connectivity. The next AI winners may not simply be those who build the models—they could be the people who know how to turn those models into real business value.

For me, a cash-secured put is not simply a strategy to collect premium—it is a commitment to buy a stock at a price I have already decided is attractive. I prefer OTM strikes with enough downside buffer, typically giving myself time for theta to work without taking unnecessary assignment risk. But the biggest lesson is that a high premium often comes with a reason: elevated IV usually means the market expects bigger moves. I also prefer limit orders, especially when spreads are wide. A few cents of execution difference may look insignificant, but repeated across multiple contracts, it adds up. Most importantly, I treat assignment as part of the original plan, not a failure. Before entering, I ask one question: If this stock falls another 30%, would I still be comfortable owning 100 shares

Bitcoin’s bull case is becoming less about hype and more about how the market reacts to bad news. The Fed just hiked rates, the CLARITY Act stalled, and BTC still recovered toward $87K. More importantly, U.S. spot Bitcoin ETFs recorded five straight inflow sessions, including nearly $999M on September 21. That tells us something important: buyers are increasingly willing to absorb macro and regulatory shocks. Tiger Research’s $250K target by 2029 is therefore interesting not because $250K sounds exciting, but because its framework is based on Bitcoin’s expanding monetary role, investor cost bases and its valuation relative to gold. But the key risk remains liquidity. If yields keep rising and ETF flows reverse, the bullish structure could be tested again. For me, the next question isn’t “

Full Moon, Bright Future ​As the Mid-Autumn moon illuminates the iconic skyline of Marina Bay, it brings a spirit of warmth, gratitude, and togetherness. ​Happy Mid-Autumn Festival to all fellow Tigers! May this season of reunion bring joy, harmony, and peace to you and your loved ones. ​A special congratulations to Tiger Brokers! Wishing you continued success, steady growth, and global momentum. Here's to soaring to new heights together! ​May your portfolio be as full as tonight’s moon, and your investments yield golden rewards! @TigerEvents

[你懂的]  The “Boring” IT Distributor Quietly Riding the AI Boom $TD SYNNEX (SNX) At first glance, SNX looks incredibly boring. It is a huge IT distributor with more than $60 billion in annual revenue. It sells hardware, software, networking equipment and technology solutions to businesses and resellers. But there is something hiding underneath that traditional business: Hyve Solutions. And this is where the AI story gets interesting. So, how does SNX actually make money? The traditional SNX business is basically a giant technology supply chain. A manufacturer produces the equipment → SNX buys and distributes it → resellers, system integrators and enterprise customers buy it. SNX makes money through distribution margins and value-added services. The catch? Margins are thin. That mea

I’d choose ③ — DRAM can stay strong, but NAND may peak first. AI is changing memory demand, but not every segment benefits equally. HBM and server DRAM remain closely tied to AI infrastructure, with rising memory content per server helping support pricing. NAND is different. Enterprise SSD demand is strong, but NAND still has greater exposure to consumer electronics. If new capacity ramps faster than demand, NAND pricing could weaken earlier. That’s why I wouldn’t ask whether the entire memory cycle has peaked. The more important question is which segment turns first. Burry’s warning still matters: high margins eventually attract supply. But timing is everything. For MU, SNDK and SKHY, I’d watch pricing, inventories and 2027 capacity growth closely. The memory trade may not be simply bulli

The memory rally is real—but the next phase is about proving earnings can catch up with expectations. AI is absorbing enormous amounts of DRAM, HBM and NAND, while new capacity takes years to build. That gives $MU and $SKHY unusual pricing power. But I wouldn’t confuse “sold out” with “risk-free.” CXMT is already expanding advanced DRAM production, while memory is still a cyclical industry. For me, the real signal is simple: watch whether strong pricing translates into sustained margins and cash flow. If MU’s September 30 results confirm that, the thesis gets stronger. If demand or pricing disappoints, today’s high expectations could amplify the downside. Memory isn’t just a capacity story anymore—it’s a test of whether AI demand can permanently reshape the cycle.

I’d pick ③ Hybrid cloud + local becomes the standard. The AI industry probably won’t move entirely from the cloud back to PCs. Instead, workloads will be split based on economics and performance. Frontier models, large-scale training and complex reasoning will remain in data centers, where NVIDIA’s ecosystem has a major advantage. But repetitive agent tasks, private enterprise data and latency-sensitive inference could increasingly run locally. The key change is that AI compute may become workload-dependent rather than cloud-dependent. If local hardware becomes powerful enough, companies can avoid paying inference fees for every single task. Over thousands or millions of daily operations, that difference could become significant. So the next AI infrastructure battle may not be cloud vs. lo