Heroic Lifesaver

$Micron Tech(MU.US)

Post 2 of Episode 24

Context: What has strengthened my Micron thesis is that AI is increasing the value of memory across the entire data-centre stack, not just HBM. Micron is already shipping HBM4 in high volume for its lead customer, while its latest 512GB DDR5 server module targets increasingly memory-intensive AI and analytics workloads. At the same time, HBM consumes disproportionately more wafer capacity, which can tighten supply across conventional DRAM as well.

My trade: I therefore see MU as more than an HBM beneficiary. It gives me exposure to a broader structural shift where AI servers require substantially more high-performance memory at multiple levels. That is already showing financially: fiscal Q3 revenue reached a record $41.46 billion, while Micron guided Q4 revenue to roughly $50 billion with an extraordinary ~86% gross margin.

Takeaway: For years, memory was treated as the cyclical component sitting beside the processor. AI may be changing that relationship. When compute gets faster, memory increasingly determines how much of that compute can actually be used — turning memory from supporting hardware into part of the bottleneck itself.

* I have reduced my Micron position ahead of earnings to reduce risk. Might add back if price action and earnings are good.

@Captain's Treasure

Markets are rather poised at the moment. They have done remarkably well in the face of rising yields, oil prices etc. Inflation numbers if lower than expected might provide a boost.

$NVIDIA(NVDA.US)

Post 1 of Episode 24

Context: What stands out to me in NVIDIA’s latest numbers is how the business is evolving from a GPU supplier into the architect of the entire AI factory. Q2 FY2027 revenue reached $96.2 billion, up 106% YoY, while Data Center revenue reached $89.0 billion, up 117%. More importantly, NVIDIA says its revenue opportunity per gigawatt has expanded from roughly $18 billion with Hopper, to $25 billion with Blackwell, and potentially $40 billion with Vera Rubin as CPUs, GPUs, networking, storage and software become increasingly integrated.

My trade: That changes how I think about NVIDIA. I’m no longer viewing it simply as ‘selling more accelerators’. With Vera Rubin now entering production shipments, NVIDIA is trying to capture an increasing share of the entire AI-infrastructure build — from compute and NVLink to Spectrum-X networking and BlueField.

Takeaway: The next leg of NVIDIA’s story may not be about selling more chips alone. It could be about monetising more of every AI factory that gets built. When each new generation increases both performance and NVIDIA’s content per deployment, the size of the opportunity can grow even without relying purely on GPU unit growth.

*At the moment, Nvidia is a very small percentage of my portfolio and serves as a placeholder as I believe there are other names that will give better next dollar rewards.

@Captain's Treasure

I believe the markets are more resilient than the headlines would make them to be. They have done remarkably well in the face of rising yields, oil prices, etc., and the names that are being rotated into indicate economic growth ahead.

Haha….Looks like the TACO script is in play again. And markets still fall for it! Expecting talks to progress and the president’s stance to soften again….

$GE Vernova(GEV.US)

Post 5 of Episode 23

Context: My original interest in GE Vernova was simple: AI needs enormous amounts of electricity, and power is becoming one of the biggest constraints on data-centre growth. What has strengthened the thesis is that this is now showing up in the numbers. GEV ended Q2 with a $176 billion backlog, while gas-power equipment backlog and slot reservations reached 116 GW. Even more striking, data-centre orders within Electrification exceeded $5 billion in the first half of 2026 — already more than double the total for all of 2025.

My trade: I increasingly see GEV not simply as an AI-adjacent power play, but as a company monetising scarcity across both generation and the grid. Gas turbines address the need for dependable new power, while its electrification businesses help move and transform that electricity. The Prolec GE acquisition adds further exposure to the transformer and grid-equipment side of that bottleneck.

Takeaway: A good thematic investment becomes more interesting when the story starts turning into orders. For GEV, AI’s power problem is no longer just a future forecast — increasingly, it is becoming backlog.

@Captain's Treasure

Lol. They called for an AI slowdown and signed another deal. Talk about actions speaking louder than words! And looking at how sentiments are in the basement yet markets are resilient seems to indicate the trend is up. We might get a dip into mid-October or November but it looks like a rally is in the making!

$SPDR Gold Shares(GLD.US)

Post 4 of Episode 23

Update on GLD.

Context: What I find interesting about gold today is that the macro picture is pulling it in opposite directions. Higher rates and bond yields are clear headwinds, yet inflation remains elevated, geopolitical uncertainty persists and central banks continue to view gold as an increasingly important reserve asset.

My trade: That is why I don’t view GLD simply as a proxy to falling interest rates. I hold it as a macro counterweight — something that can respond to risks that may not be captured by my growth and technology positions. I’m comfortable letting it play that role even when the immediate monetary-policy environment is challenging.

Takeaway: Gold doesn’t need every macro variable to line up perfectly. Sometimes its value is greatest when the economic signals themselves are contradictory. When inflation, rates, geopolitics and asset prices are pulling in different directions, diversification becomes the thesis.

@Captain's Treasure

It’s typical of markets to worry over uncertainty. Ooh rates are so high. Have a look at rates over the past 50 years and you realise it is not that high after all. Then the markets will push higher and narratives change to something else!

The rotation within AI tech was at it again last night as memory names got bid up while optical names moved sideways. Tech needs to move up in tandem with the general market and make new highs with improving breadth for me to be convinced the true blue bull is in town.

Apparently all the market needed was… a Muse. 😂

$Meta Platforms(META.US) launches its new personal AI agent, it shoots to the top of the App Store, Meta jumps about 11% in a day… and suddenly the whole market remembers:

“Oh yes. We like AI again.” 😂

Muse has already pulled in around 2.8 million downloads in its first 12 days, and the excitement spilled beyond Meta into the broader AI trade, including names like AMD and Intel.

What I find interesting isn’t just Meta’s move.

It’s what Muse potentially represents.

We’ve spent the last few years talking about AI models getting smarter.

Now we’re moving toward AI that actually does things — books, buys, searches, organises, communicates and works in the background for you. Meta itself describes Muse as an agent designed to take actions across apps rather than simply answer questions.

And if agentic AI really takes off, that means a lot more inference, orchestration, memory, networking and compute happening continuously in the background.

Which probably explains why Mr Market saw one popular AI app and immediately went:

“MORE CHIPS!” 😂

Of course, a few million downloads don’t suddenly justify every AI valuation on earth.

But Muse has done something important.

It has reminded the market that the next phase of AI may not just be about building bigger models.

It may be about getting AI into the hands of ordinary people — and actually giving them a reason to use it every day.

And judging by Monday’s market reaction…

Wall Street appears to have found its Muse too.

Aren’t we all a’muse’d? Sorry I can’t help the pun! The topic was calling for a (pun) chline! 🤓😂

@Captain Leo

$Sandisk(SNDK.US)

Post 3 of Episode 23

An update on Sandisk

Context: What makes SanDisk interesting to me is that NAND is still a cyclical business — but this cycle may be developing differently. Years of disciplined capacity investment, rising data-centre demand and tighter supply can create an environment where storage manufacturers regain pricing power instead of simply competing on volume.

My trade: I see SNDK as exposure to that potential shift in industry economics. If demand continues growing while supply remains constrained, the upside may come not only from selling more NAND, but from selling it at better prices and under increasingly favourable long-term arrangements.

However, I have trimmed down and might trim further my Sndk holdings for technical reasons one being the time cycle we are entering hence time is of more importance than price here and how price behaves into this time cycle is more important. Hence, I am taking gains and practicing caution here with a view to a re-entry.

Takeaway: Sometimes the investment opportunity is not about discovering a new technology. It is about recognising when the economics of an old one are changing. When something essential becomes scarce, pricing power can become the real catalyst. However, excellent fundamentals do not equate to value. And IMHO, technical analysis provides a good window alongside fundamental valuations for this purpose.

@Captain's Treasure

It was quite the rally by AI tech and the beneficiaries on another proof that AI actually makes money and investments in them do pay off if managed wisely.

The Clarity Act is a good example of how politics play a part in how a country moves forward. When you have a clear divide in votes aligned with the political parties for a bill that is helpful for regulation clarity, it shows them up. Funny thing is crypto regulations are still moving ahead!

$AEM SGD(AWX.SG)

Post 2 of Episode 23

This is the stock I was talking about in the previous post.

Context: What interests me about AEM is that as semiconductors become more complex and expensive, the cost of allowing a defective chip to progress further into packaging and system assembly rises dramatically. That makes advanced testing, handling and reliability increasingly important — especially in high-performance computing and AI.

My trade: I see AEM as exposure to the test and reliability layer of the semiconductor ecosystem rather than to any particular accelerator, memory or networking architecture. The thesis is simple: regardless of which chip ultimately wins, increasingly sophisticated devices still need to be tested efficiently and accurately before deployment.

Takeaway: The more valuable the chip, the more valuable it becomes to know that it works. In an AI world obsessed with faster compute, the less glamorous business of testing and reliability may become increasingly important — and that is where AEM earns its place in my portfolio.

@Captain's Treasure

A quick reminder to myself!

The market rallies for a few days… and suddenly we’re all geniuses again 😂

Funny how quickly investing psychology changes.

A few red days and I’m questioning every position in my portfolio.

A few green days and suddenly I’m thinking:

“Yes. Exactly as I predicted.” 😂

Of course, nothing has magically changed about my investing ability.

That’s probably one of the things about a strong market.

When everything is going up, it becomes very easy to confuse a rising tide with stock-picking skill.

So whenever my portfolio starts looking particularly clever, I try to remind myself:

The thesis still matters.

Valuation still matters.

Position sizing still matters.

And perhaps most importantly…

The market has a wonderful way of humbling anyone who starts feeling a little too intelligent. 😂

Enjoy the green days.

Just don’t let them go to your head.

I am watching $VanEck Semiconductor ETF(SMH.US)and $iShares Semiconductor ETF(SOXX.US)closely as a proxy to my portfolio. 👀

@Captain Leo