Heroic Lifesaver

$UMS(558.SG)

Context: What attracted me to UMS is that it sits further upstream in the semiconductor value chain. It is not an accelerator, memory or networking company; instead, it provides the precision manufacturing and engineering capabilities needed by semiconductor-equipment makers as fabs expand capacity and adopt more advanced processes.

My trade: I added UMS as a Singapore-listed way to participate in the broader semiconductor capital-spending cycle. Rather than relying solely on which AI chip architecture wins, the thesis is that increasing chip complexity and continued fab investment should support demand for the equipment ecosystem that enables production.

Takeaway: In a technology boom, there can be value in owning the companies that help build the factories and tools behind the winners. Sometimes the less visible “picks-and-shovels” businesses offer a different way to participate in the same structural trend.

@Captain's Treasure

Fed day. And I’m not sure the rate decision is even the most important part.

The market has already formed a pretty strong expectation about what the Fed will do tonight, so I’m actually more interested in what comes next.

Is this just a one-off move in response to inflation pressures?

Or does the Fed think rates need to stay higher for longer?

That distinction matters much more to me as an investor.

So tonight I’ll mainly be watching three things:

1. Bond yields. 2. The US dollar. 3. And how tech and crypto react after the announcement.

Sometimes the market reaction tells you much more than the headline itself.

If the Fed sounds hawkish but Treasury yields struggle to move higher, or the dollar fails to strengthen, that would definitely get my attention.

On the other hand, if yields break higher and the dollar follows, liquidity-sensitive assets like tech and crypto could have a tougher time.

So for me, tonight isn’t just about:

“What does the Fed do?”

It’s more about:

“How much has the market already priced in — and what does it do once the news is actually out?”

That’s what I’ll be watching. 👀

@Captain Leo

Is the “AI slowdown” narrative really bearish?

I’ve been thinking about the recent comments from Anthropic’s CEO about slowing down AI development.

And I have to say… I’m not completely buying the way the market is interpreting it.

The safety concerns may well be genuine. But I also can’t ignore the incentives.

If you’re already one of the companies at the frontier of AI, slowing things down — or making the regulatory hurdles higher — doesn’t necessarily hurt you. In fact, it could make life much harder for everyone trying to catch up.

That’s also why I found it interesting that Elon Musk $Tesla(TSLA.US)agreed.

Maybe I’m being cynical 😂, but when competitors suddenly agree that everyone should slow down, one of the first questions I ask is:

Who benefits?

For me, the bigger question as an investor is whether anything has actually changed in the AI infrastructure story.

Are hyperscalers cutting capex?

Are accelerator orders being cancelled?

Is demand for networking, memory, storage or power starting to weaken?

If the answer is no, then I’m not sure a few comments about slowing AI development automatically mean the AI investment cycle is slowing too.

For now, I see this more as a narrative shock than a thesis break.

And perhaps there’s another possibility:

Slower model development could actually mean more spending on inference, safety, testing and infrastructure before the next generation gets released.

That’s the part I’m watching.

What do you think? Genuine concern, strategic positioning… or a bit of both? 😂

@Captain Leo

$Sandisk(SNDK.US)

Post 5 of Episode 22

Context: What changed my view on SanDisk was recognising that AI inference creates a very different storage problem. As context windows grow and workloads become more data-intensive, fast NAND can increasingly be used to offload and support KV-cache workloads instead of relying entirely on much more expensive memory.

My trade: I see SNDK as a way to gain exposure to this less obvious second-order effect of AI. The attraction is not just higher NAND demand, but the possibility that tighter supply, improving pricing and greater data-centre usage can lift the economics of the business at the same time.

Takeaway: The next phase of AI may reward more than just compute and memory. When workloads scale, the boundary between “storage” and “performance infrastructure” starts to blur — and that is where SanDisk becomes much more interesting.

@Captain's Treasure

$GE Vernova(GEV.US)

Post 4 of Episode 22

Context: The deeper I looked into the AI infrastructure buildout, the clearer it became that computing power is only useful if there is enough electrical power to support it. Data centres, grid expansion and electrification are creating demand that cannot be solved simply by adding more GPUs.

My trade: I added GEV as a way to diversify my AI exposure beyond semiconductors. Its position across gas power, grid equipment and energy infrastructure gives me exposure to the physical systems needed to bring new generation online and move that electricity to where demand is growing.

Takeaway: Sometimes the most interesting way to invest in a technology revolution is to own the bottleneck it creates. AI may be the headline, but reliable power could be one of the resources that determines how quickly the story can actually scale.

@Captain's Treasure

$Broadcom(AVGO.US)

Post 3 of Episode 22

Context: What attracted me to Broadcom is that the AI buildout is not only about GPUs. As clusters scale, demand also rises for custom silicon, high-speed networking and connectivity — areas where Broadcom has become increasingly important.

My trade: I viewed Broadcom as a way to participate in AI infrastructure from a different angle, with exposure to both semiconductor growth and a highly cash-generative software business. Rather than treating it as a short-term AI trade, I approached it as a strategic holding tied to the broader expansion of data-centre architecture.

Broadcom is not a big position. It is just a placeholder in my portfolio.

Takeaway: Sometimes the best opportunities are not in the most obvious part of a theme, but in the companies supplying the critical infrastructure that allows the entire system to scale. A good company however is usually not the cheapest. So patience is key in getting good value.

@Captain's Treasure

$NVIDIA(NVDA.US)

Post 2 of Episode 22

Context: NVIDIA has been one of the clearest beneficiaries of the AI infrastructure buildout, but a great company does not automatically mean an unlimited position size. As the share price appreciated and my portfolio gained exposure to other parts of the AI stack, I became more conscious of concentration and valuation risk.

My trade: I gradually trimmed my NVIDIA position rather than exiting completely. I wanted to lock in gains while retaining some exposure to a company that remains central to AI compute. At the same time, I redeployed capital toward less obvious bottlenecks across networking, memory, optics and power.

Takeaway: You don’t have to stop believing in a company to reduce a position. Sometimes trimming a winner is simply good portfolio management—especially when it creates room for opportunities with a better risk/reward setup.

@Captain's Treasure