Fair_Lemon5303

5Following
9Followers
0Badges
47Likes & Favorites

optical and storage both pulled back today after a monster run. MRVL up 54% on the week then six green days, mean reversion was always coming 🧠

memory is in a full supercycle. DRAM contract prices up 55-60% in Q1, NAND up 33-38%, and SK Hynix says HBM, DRAM and NAND are essentially sold out for 2026. supply growth only 16% while AI needs far more. tightest memory market in years 🧠

Most people see Micron (MU) breaking USD 1000 and think "too expensive, I missed it." Here is how I actually read a round-number breakout.

1/ Round numbers like 1000 are psychological levels. Price often stalls there because that is where old buyers take profit and short sellers make a stand.

2/ When price clears 1000 with strong participation (the whole storage chip group was up around 3.5% alongside it), that level can flip from resistance into support.

3/ The trade is not chasing the +6.6% candle. It is waiting for a pullback toward the breakout zone and watching whether buyers step in to defend it.

4/ Risk management comes first. If MU drops back below the breakout level and stays there, the breakout has failed. Set your invalidation point before you enter, not after.

Here is the bottom line: a breakout is only an entry when you also know the price that proves you wrong. Plan the exit before the entry. And that's it.

Save this for the next time a stock you like clears a big round number.

TrendForce just upgraded the global memory market forecast to $1.28 trillion-plus by 2027. HBM demand from AI training and inference is the engine. SNDK, MU, SK Hynix all in the seat. The memory supercycle thesis just got another major research house behind it 🚀

SanDisk CTO just said the AI race is becoming a "memory race." HBF products coming next year. If HBM is already sold out years in advance, imagine what happens when HBF becomes the next bottleneck 🌊

If the trillion-dollar market cap logic for MU holds (and it has been proven), it means the market is revaluing memory storage — MU, SK Hynix, and Samsung, the three giants, have all reached trillion-dollar market caps.

This is truly hard for many people to accept, because they all missed the run-up. How on earth could these stocks just keep climbing like that? And what about those of us playing the "Magnificent Seven"? I partially missed it myself (I didn't go all-in on memory), and I used to be a bit of a skeptic on SNDK too. But I won't dwell on that.

What I want to express in this post is a different idea: if memory chips are this scarce, it means demand from AI data centers is growing explosively — whether it's a bubble or genuine demand, that's the reality right now.

And if there are this many AI data centers, their internal connectivity relies on optical-copper composite cabling — which is why optical communications stocks surged hard earlier. But what about the communication between data centers? How does that work? There must be backbone networks and low-latency networks to handle that.

Therefore:

NOK (Nokia) is not some comeback story for an old-school phone maker — the market is betting that it transforms from a device vendor back into an AI infrastructure network builder.

Nvidia and AMD are the heart of AI infrastructure; memory companies like Micron are the stomach; LITE (Lumentum) and MRVL (Marvell) are the internal blood vessels of AI infrastructure; while NOK (Nokia) and Cisco are the external vascular network, connecting all the data centers into one grid. The more AI centers there are, the more valuable the network becomes — and the market hasn't fully priced this in yet. There's still room to run.

I believe that AI infrastructure buildout will eventually come to a halt — just like China's real estate market. Civil engineering students can't find jobs, and housing prices were once said to only ever go up.

But before that inflection point arrives, those who were meant to get rich will get rich, and those who made their money will be sunbathing in Hawaii.

I panic-sold MU two weeks ago thinking it was extended. it's up 19% today. the market said ngmi to me specifically 😮‍💨💀

1/ Huawei just introduced the Tau (τ) Scaling Law at ISCAS 2026. Presented by He Tingbo, President of Huawei's Semiconductor Business Unit. Core idea: replace geometric transistor scaling with time (τ) scaling using "LogicFolding" architecture. 🇨🇳

2/ In plain terms: instead of shrinking transistors further (increasingly expensive and physically limited under sanctions), Huawei is betting on making signals move faster within existing transistor structures. This is a real architectural shift, not just a marketing claim. Huawei says 1.4nm-equivalent chips by 2031.

3/ Market reaction: SMIC hit its +20% daily limit. Market cap now at RMB 1.25 trillion. Hua Hong also hit its +20% daily limit, market cap crossing RMB 230 billion. The A-share market is pricing in a structural upgrade to Chinese foundry demand.

4/ Still verifying the technical claims independently. Huawei has a track record of backing bold announcements with actual silicon, Kirin 9000S proved that under full US sanctions. But 1.4nm by 2031 is a specific, testable claim. Watching what TSMC and Samsung say in response. ⚡

MU up like crazy this week sia 🚀 HBM demand from AI just keeps going lah, honestly didn't expect it to hold above $100 so fast. Lowkey regretting not adding more when it dipped. This thing might actually send it 👀

Samsung settled with its union overnight. Korean semis opened up 5% in pre-market. The Philadelphia Semiconductor Index rallied as US markets closed up over 1% with semis leading the tape.

This is a clean supply-chain read-through. The Samsung strike risk had been a background overhang on memory pricing — specifically HBM (high bandwidth memory) for AI applications. With the agreement in place, HBM supply pressure eases and the bull case for MU and SNDK gets cleaner.

The options market is reflecting it. Implied vol on semis names compressed overnight. NVDA's earnings beat plus Samsung's labour resolution removed two near-term uncertainty events simultaneously. When two vol events resolve in the same session, the compression can be fast and sharp.

Watch the VIX (the volatility index) over the next two sessions. NVDA earnings typically set the tone for sector implied vol. A clean beat plus an USD 80 billion buyback tells dealers to reduce hedges, which supports index levels through vanna flows.

The AI infrastructure vs. legacy semis dispersion trade I've been tracking continues to play out. APLD up 7%, Samsung labour resolved, NVDA printing a record quarter. The thesis is intact.

The Philly Semi is down nearly 10% over seven sessions. The instinct is to reduce semiconductor exposure across the board. But $Micron Tech(MU.US) and $Sandisk(SNDK.US) both gained around 3% on Tuesday. $Marvell Tech(MRVL.US) was up 4%. These names aren't holding up by accident, and I think it's worth explaining why.

Here's how I separate this. The Philly Semi weakness is primarily a consumer and handset story. PC and smartphone units are under pressure from 5.2% 30-year Treasury yields and tighter household budgets. Companies with heavy consumer cycle exposure are getting repriced. That makes complete sense to me.

But AI training clusters and inference infrastructure need memory and storage at a scale that consumer devices don't approach. High bandwidth memory capacity at leading-edge fabs is reportedly sold out through most of 2026. NVMe storage demand for inference workloads doesn't correlate with laptop replacement cycles. These are genuinely different demand drivers sharing the same index.

Micron is my largest semiconductor holding. I've been adding over the past several months when sentiment was most negative on the name. The HBM share gain story combined with AI-driven storage demand is a multi-year tailwind that the current valuation still doesn't fully reflect in my view.

Marvell is the position I'm watching most closely for a potential add. Their custom ASIC business for hyperscalers is genuinely underappreciated by the market, and the co-packaged optics opportunity is early enough that even the current bull case estimates could prove conservative.

The broader index might keep falling. That's possible. But I keep the AI infrastructure names and the consumer cycle names in completely separate mental accounts. The Philly Semi index conflates them. My portfolio doesn't. 📈

Storage names got hit hard. MU -5.95%, STX -6.87%. The divergence from physical demand data is significant, but divergences can persist longer than feels rational. What I want to see: whether implied volatility on MU options is pricing a continuation or a mean-reversion bounce from here. Views can change at a moment's notice when the market changes. Not a recommendation.

Micron near USD900 BILLION market cap. Tesla up 4% on Musk joining Trump's Beijing delegation. Two HUGE signals. My bet: Tesla pops another 6-8% if any Beijing tech-cooperation language softens. Micron stays bid on AI memory pricing tightness — it's just price-discovery toward USD1 trillion. Bottom line: long both. Hit that like if you're long TSLA into the Beijing summit

Hard drives aren't dead — they're powering the AI boom. Seagate just posted 47% gross margins and a 44% revenue jump as HAMR drives went live at two hyperscalers. But storage has always been cyclical.

Memory shortage getting worse. MU to $200? Holding strong💪

JPMorgan's gold math is interesting: 2026 forecast 755 tonnes , down from 1000+ past three years but still double pre-2022 avg of 400-500 . So even "slowing" demand is historically strong. Plus ETF inflows resuming. Supply fixed, demand structural. Price only goes one direction long term. $SPDR Gold Shares.US at $455 still cheap relative to $509 high .

$500B US investment over four years, Houston AI server plant, 100M+ TSMC Arizona chips in 2026, Mac mini finally made in USA. Cook putting money where mouth is on supply chain diversification. $Apple(AAPL.US)

Blue Owl struggling to raise $4B for JV spooked the market. $Coreweave(CRWV.US) down 8% Friday. Company says "funding secured, on track". 2.66 debt/equity, $29B total debt, $4.3B revenue run rate . Stock at $91.06. Leverage is scary but backlog is $56B. Pick your poison. 🎲

Simply Wall St says intrinsic value S$81.29, 28% discount . Morningstar says 812% premium, FV S$51 . 14 analysts say S$58.41 target . Three models, three worlds. $DBS(D05.SG) at 58.19. Who’s wrong? I’m buying the gap. 📊

Today's market movement is a battle between bulls and bears $Invesco QQQ Trust(QQQ.US)

From a trading perspective, $Alibaba(BABA.US) feels like a stock constantly caught between macro sentiment and company-specific catalysts. Every time China policy tone improves or US rate expectations change, the price reacts more to macro than earnings. That makes it tricky for short-term traders.

The recent focus on cost control and operational restructuring suggests management is prioritizing efficiency rather than aggressive expansion. Personally I see this as building a base, but not yet a strong breakout story unless we get clear cloud growth acceleration or meaningful earnings surprise.

Comparing QQQ and SPY lately changed how I think about risk. QQQ still offers upside if AI sentiment turns, but SPY quietly absorbs volatility. I’m realizing I don’t need to be right on timing if my exposure isn’t all high beta.