To The Moon

Signal Over Sentiment: Where Hong Kong’s Smart Money Is Actually Moving Today The Hong Kong market still puts on a decent show - flashes of momentum, the occasional surge, enough noise to suggest something exciting is always happening. But I think that’s largely theatre. Behind the curtain, capital is behaving in a far less dramatic, and far more decisive, way. It is no longer rotating across sectors like a well-diversified tourist. It is checking into a few places and refusing to leave. Capital isn’t rotating—it’s clustering, and staying put That shift matters. Because if capital is concentrating rather than rotating, then the real ‘hot spots’ are not the loudest trades - they are the ones quietly absorbing sustained, institutional money. When I look at the market through that lens, three

The Chip That Quietly Runs the Show I find it mildly amusing that while everyone is busy applauding the flashy AI processors, it is memory—decidedly less glamorous—that is holding the entire performance together. In 2026, Micron is no longer just a participant in the semiconductor cycle; it is the constraint. As AI systems shift from generating content to executing reasoning tasks, the workload becomes far more dependent on rapid data access and movement. High-bandwidth memory is no longer a supporting component; it is the pacing mechanism. A cutting-edge processor without sufficient memory bandwidth is, in effect, a race car stuck in traffic—technically impressive, but going nowhere particularly fast. AI runs fast—until memory quietly sets the speed limit Micron’s position here is not jus

From Optional Tool to Institutional Habit I have always found that markets tend to misprice what they cannot neatly categorise, and $Palantir Technologies Inc.(PLTR)$ fits that description almost too well. It is not quite a defence contractor, not quite a SaaS platform, and certainly not your standard AI darling chasing chatbot headlines. That ambiguity, in my experience, is often where the most interesting opportunities—and misjudgements—emerge. What I see here is a company embedding itself into the operational fabric of institutions that do not have the luxury of indecision. Palantir is not selling intelligence; it is embedding judgement. When its systems are used to manage logistics, interpret intelligence, or optimise energy flows, the softwar

The Grid Didn’t Get the Memo I used to think the limiting factor in AI was chips. That was neat, measurable, and—if I’m honest—comforting. You can model semiconductors. You can’t really model whether a regional grid will politely agree to keep up. It turns out the grid didn’t get the memo. If $Oracle(ORCL)$ builds the data centres and $Intel(INTC)$ supplies the chips, $Bloom Energy Corp(BE)$ is the one making sure the lights actually turn on. That line sounds obvious, but the market has been slow to internalise it. We have spent years obsessing over compute while quietly assuming electricity would just… be there. Like oxygen. Or Wi-Fi in a café that claims to have

A Turnaround Measured in Atoms, Not Quarters I have always found that the market struggles to price transitions that are both technical and geopolitical, and Intel sits squarely in that blind spot. This is not merely a semiconductor story; it is an industrial policy experiment awkwardly squeezed into quarterly earnings calls. At its core, the 18A process node is not just another upgrade—it is the dividing line between credibility and irrelevance. If it works, $Intel(INTC)$ does not simply improve margins; it becomes a cornerstone of Western technological sovereignty. If it does not, we are left with a sobering reality: sovereignty is expensive, and markets are not obliged to subsidise it indefinitely. Sovereignty is engineered; profitability remai

The Unlikely Winner of the Compute Crunch I’ll admit, Oracle is not the name most investors instinctively associate with the AI boom. It lacks the glamour, the developer fandom, and the headline dominance. In many ways, it is the unglamorous plumber of the AI gold rush—yet that may be precisely where the value lies. What I see is not a late entrant scrambling for relevance, but a company quietly benefiting from a structural bottleneck. Demand for high-performance computing has surged faster than supply can comfortably accommodate, and $Oracle(ORCL)$ has positioned itself as a willing and increasingly capable provider of that missing capacity. This is not about being the best-known cloud. It is about being available, performant, and—crucially—deepl

I am acutely aware that pitching an aluminium producer at a tech-stock multiple sounds like a violation of several unwritten investing laws. Aluminium is supposed to be dull, cyclical, and cheap—something you hedge, not something you underwrite with conviction. Yet $Century Aluminum(CENX)$ has forced me into an awkward position: defending a 60× P/E smelter as a rational way to play US reshoring, industrial policy, and ESG-driven scarcity. If nothing else, this is proof that the market has entered a strange new phase of adulthood. Aluminium reshaped by innovation, policy, and ESG imperatives This is not a value story. It is not even really a commodity story. It is a policy-enabled growth trade disguised as heavy industry, and the sooner investors a

I have long been sceptical of banks declaring themselves ‘technology companies.’ Most still behave like utilities with better apps and bigger marketing budgets. $Capital One(COF)$, however, is quietly dismantling the universal banking playbook rather than polishing it. By owning the payments network, the software layer, and the balance sheet, Capital One is assembling a full-stack model that collapses boundaries most banks still treat as sacred. The result is a business that looks increasingly mislabelled — benchmarked as a bank, but behaving more like a platform with regulated funding. When payments, software, and banking stop pretending they’re separate The Tollbooth Heist Nobody Is Modelling Most coverage frames the Discover acquisition as a sca

Constellation Brands has chosen the least fashionable strategy in modern markets: deliberate contraction. In an era obsessed with growth-at-any-cost, STZ is doing the opposite — and quietly improving the equity in the process. Shrinking the structure to reveal what actually compounds When Less Starts Pulling Its Weight The consensus looks at Constellation’s mid-single-digit organic sales decline and reaches for familiar diagnoses: fading relevance, demographic decay, or strategic drift. I see something far more intentional. This isn’t erosion; it’s excision. $Constellation(STZ)$ is performing surgery on itself — and the patient is thriving on the operating table. The divestment of lower-margin wine and spirits assets, including SVEDKA and further w

Ciena Corp: Beyond the Silicon Peak — The Quiet Toll-Collector of the AI Efficiency Era AI investing still looks like a race up the silicon mountain. Faster GPUs, denser racks, louder narratives. $Ciena(CIEN)$ sits somewhere less glamorous but far more unavoidable: at the junction where power, distance and data collide. The market largely treats it as a beneficiary of AI growth. I see it as a physical bottleneck with pricing power. The arteries of AI, flowing unseen but indispensable This is not an AI hardware story in the conventional sense. It is an efficiency monopoly operating inside tightening constraints, and that distinction is being mispriced. When Saving Energy Spends More of It The dominant narrative around Ciena is simple: data volumes