② Earnings are still strong, but the best growth rate is behind us Samsung’s 783% profit surge proves how powerful the AI memory cycle has become. HBM demand remains strong, while server DRAM and enterprise SSDs continue benefiting from hyperscaler spending. But investors should distinguish record earnings from accelerating earnings. Memory is cyclical. Extraordinary margins eventually attract new capacity from Samsung, SK hynix, Micron and Chinese suppliers. If supply catches up faster than demand, pricing power can weaken before AI demand actually slows. That is why Samsung’s muted stock reaction matters. The market is looking beyond Q3 profits and asking whether these margins can survive into 2027. I choose ②. I’m not calling the memory cycle over, but the easiest part of the trade may

The SaaSpocalypse thesis was too simplistic. AI can replace features, but replacing enterprise software means replacing data infrastructure, permissions, compliance, workflows and years of integration. That is a far higher hurdle. The bigger risk is not disappearance—it is pricing power. If agents dramatically increase productivity, customers may demand fewer seats and lower subscription costs. The winners will therefore be SaaS companies that turn AI into measurable incremental revenue, not merely cheaper software. CRM, ITSM and cybersecurity incumbents have an important advantage: they already sit where enterprise data and workflows live. The real test is simple: Does AI expand the economic value of the platform faster than it destroys the old pricing model? If yes, SaaS may be entering

The biggest mistake investors can make in the 2026 midterms is betting on red or blue instead of betting on policy. History is encouraging: the S&P 500 has averaged roughly 6.6% in Q4 of midterm years, and has posted positive returns in the 12 months after every midterm since 1950. But 2026 is different. With the 10-year Treasury near 5.3%, elevated valuations and massive AI capex, rates and earnings may matter more than election headlines. My focus would be on AI infrastructure, power and defense. A divided Congress could actually be constructive by limiting major policy shocks, while a Republican sweep could favor deregulation, energy and AI infrastructure. The real trade isn't Republicans vs Democrats. It's policy uncertainty vs. policy clarity. For me: A — AI & Technology, but

[你懂的]  Semtech (SMTC): The AI Infrastructure Stock Most Investors Still Overlook When investors talk about AI infrastructure, the usual names are NVIDIA, AMD, Broadcom, Marvell, Micron and SanDisk. But there is another part of the AI infrastructure equation that is becoming increasingly important: How do all those GPUs actually communicate with each other? That is where Semtech (SMTC) gets interesting. Semtech is not a GPU company and it is not simply another “AI chip stock.” It provides critical semiconductor technologies for high-speed data transmission, signal integrity and connectivity inside modern data centers. In simple terms: NVIDIA provides the computing power. Networks move the data. Semtech helps make that data move faster and more reliably. And as AI clusters move fro

My answers: 1A, 2B, 3C, 4B, 5A, 6B, 7B, 8B, 9C, 10B. ​ETFs are fantastic, versatile tools for building long-term portfolio core holdings, but products like TQQQ or SOXL require extra caution. Because of daily resets and volatility decay, leveraged ETFs are better suited for short-term tactical trades rather than buy-and-hold investing. ​My key takeaway is that successful ETF investing goes far beyond choosing the right underlying index. Expense ratios, concentration risk, interest rate dynamics, and structural leverage mechanics significantly influence long-term performance. Keeping fees low and staying disciplined with risk management are essential strategies to avoid emotional trading and maintain consistent portfolio growth. @TigerEv

The AI Rally Is Getting Broader — and That Matters The most interesting signal from these new all-time highs isn’t simply that NVDA, TSM and LITE are rallying. It’s that the AI investment cycle is spreading beyond GPUs into optical networking, cybersecurity, HVAC and electronic testing. The Nasdaq just closed at another record high, even with the 10-year Treasury yield around 5.3%. That makes LITE and KEYS especially interesting to watch: they are less obvious AI plays, but benefit as data centers become larger and more complex. Meanwhile, Nvidia’s $150 billion additional buyback—bringing total authorization to $235 billion—shows just how much cash the AI leader is generating. My takeaway: **don’t only chase the headline AI names. The bigger opportunity may be in the “picks and shovels” un

[你懂的]   Penguin Solutions (PENG): The Small AI Infrastructure Company Few Investors Are Watching When investors talk about AI infrastructure, the conversation usually starts with NVIDIA, AMD, Broadcom, Micron or the major hyperscalers. But there is another part of the AI buildout that is becoming increasingly important: Who actually helps companies build and operate the AI factory? That is where Penguin Solutions ($PENG) becomes interesting. Penguin Solutions is not a GPU manufacturer. It provides AI and high-performance computing infrastructure, integrated memory solutions, software and services that help customers design, deploy and manage complex computing environments. Think of it this way: NVIDIA provides the engine. PENG helps build the machine around it. And that dist

C. 🧬 Moderna — Citi says valuation has gone too far Wall Street’s Moderna downgrade is the most interesting call because it highlights a lesson investors often overlook: great fundamentals do not automatically mean a great stock at any price. Moderna’s business may improve, its pipeline may deliver, and earnings expectations may recover — yet if the share price has already priced in too much future success, the risk/reward can deteriorate quickly. That is why analyst upgrades and downgrades should be read as valuation signals, not simple buy-or-sell instructions. Netflix and Target show how improving expectations can create upside, while Moderna shows the opposite: sometimes the biggest risk isn't a bad company — it's paying too much for a good story. For me, this is the key takeaway from

The Broadcom–Anthropic deal is less about a $42 billion headline and more about who is financing AI demand. Broadcom may fund roughly one-third of Anthropic’s $125.2 billion TPU commitment, while potentially becoming its largest chip customer. This is not proof that AI demand is artificial. But it changes what investors should measure. Revenue growth alone is no longer enough; we need to examine cash flow, infrastructure commitments and who ultimately funds the expansion. If financing accelerates genuine AI adoption, it is powerful operating leverage. If suppliers increasingly finance customers who then buy more infrastructure from those same suppliers, the ecosystem becomes more interconnected—and more vulnerable if monetization disappoints. The real AI question is no longer simply “Who

[Thinking]  Tesla just gave investors something they desperately wanted: evidence that demand may be stabilizing. In Q3 2026, Tesla delivered 486,532 vehicles, compared with production of 464,391. That means deliveries exceeded production by 22,141 vehicles. Tesla also delivered roughly 24,600 more vehicles than the company-compiled Wall Street consensus of 461,974. The stock responded immediately, rising 4.65% to $370.59 on Friday. But here's where I think investors need to be careful: A delivery beat is bullish. A sustainable earnings recovery is a completely different question. My score for this rebound: 7/10. Not 9/10. Not yet. --- 1. The 22K inventory drawdown is actually meaningful The headline number is 486,532 deliveries. The more interesting number is: 486,532 deliveries

[你懂的]  The market’s biggest story today isn’t simply whether AI stocks can keep rising. The more important question is: Where does the next wave of AI capital spending go? The latest U.S. jobs data showed a sharp slowdown in employment growth, reducing expectations for another near-term Fed rate hike. That is supportive for growth stocks. But with long-term Treasury yields still elevated, investors have less room to pay any price for future growth. That makes earnings power and real AI demand increasingly important. And this is where I think the next opportunity may be hiding. 1️⃣ NVIDIA — Still the AI Core NVIDIA remains the undisputed center of AI computing. Its massive buyback authorization also highlights the extraordinary cash generation of the business. But there is a probl

④ Too early — wait for Toshiba’s capacity to actually come online. A 10% plunge in STX and WDC looks dramatic, but the market may be pricing in a supply problem well before it actually arrives. Toshiba’s expansion is important because it confirms that AI-driven storage demand is strong enough to justify major investment. Yet new HDD capacity cannot instantly become excess supply; equipment, components, yields, customer qualification and hyperscaler contracts all take time. The real risk is not that AI suddenly stops needing HDDs. It is that by 2027–28, supply growth catches up with demand and erodes today’s pricing power and margins. For now, I would watch Toshiba’s actual ramp, STX/WDC contract pricing and nearline exabyte growth. Until those indicators deteriorate, this looks more like a

Gold Below $4,200: Is This a Correction — or a Break in the Old Gold Narrative? [暗中观察]  Gold has finally cracked below the $4,200 level. COMEX gold settled around $4,133.70/oz on October 2, extending its decline for a second consecutive week. After reaching a record high above $5,300 earlier this year, gold has now pulled back by more than 20%. At first glance, this looks like a classic profit-taking correction. But the bigger story is more interesting. Gold is now being tested by several forces at the same time: higher Treasury yields, a stronger U.S. dollar, renewed inflation concerns, oil prices and changing Federal Reserve expectations. And that creates an important question: Has gold simply become too expensive — or is the market finally challenging the assumptions behind it

C. Compute Services The AI infrastructure race is moving beyond “who has the most GPUs?” toward a more important question: who can turn those GPUs into productive compute? AI agents and inference-heavy workloads will require not only accelerators, but also networking, memory, power, cooling and reliable data centers. A GPU sitting idle creates little value; a fully utilized GPU becomes revenue-generating infrastructure. That makes compute services increasingly important. They can monetize enormous infrastructure investments while giving AI companies flexible access to capacity without owning every layer themselves. For investors, I’d watch capacity growth, utilization rates, pricing power and recurring revenue. The next phase of AI may reward the companies that successfully convert scarce