Energy First, Compute Second I see Arm Holdings as the market’s clearest bet that AI’s next constraint will not be compute, but energy. At roughly $211, with a trailing P/E approaching 280x and a forward multiple still above 100x, the stock is not reflecting what the business is—it is reflecting what the infrastructure will demand. If energy becomes the bottleneck, Arm is essential. If it does not, the valuation begins to look like a very expensive assumption. AI’s real ceiling isn’t compute—it’s electricity From Architect to Toll Collector I find the most misunderstood part of Arm’s story lies in its transition from licensing intellectual property to selling higher-value compute subsystems. Historically, $ARM Holdings(ARM)$ was the architect—desig

The Market’s First Real AI Reckoning I see Microsoft not as a participant in the AI trade, but as its first genuine stress test—where ambition, capital, and monetisation collide in plain sight. After a sharp stumble in early 2026, $Microsoft(MSFT)$ has become the market’s most consequential question mark. It is no longer enough that the company leads in AI. What matters now is whether that leadership produces incremental profit, or simply gets absorbed into an already dominant ecosystem. That distinction is where the entire debate sits. The Hidden Risk: Giving AI Away Too Cheaply AI dominance means little if pricing power quietly evaporates I believe the real battleground is not technological leadership, but pricing architecture. Microsoft is embe

Zuckerberg is no longer being judged on earnings beats—he is being judged on whether $40 billion in AI spending will compound like AWS or combust like the metaverse with better branding Meta is once again doing what it does best: making investors richer, twitchier and occasionally behave as though a 30% profit margin is some sort of corporate distress signal. This time, the market’s anxiety is not about user growth, regulators or TikTok-inspired existential dread. It is about capital allocation—specifically Meta’s decision to drive annual capital expenditure towards roughly $35–$40 billion in AI infrastructure, data centres and computing power. For context, this is not a modest budget increase. It is one of the largest strategic spending surges in modern Big Tech, and a dramatic pivot from

A Valuation That Assumes the Ending I see Tesla in May 2026 as the market’s boldest intellectual gamble—a company priced not on what it earns, but on whether it can industrialise autonomy before reality reasserts itself. At roughly $381 per share and a market capitalisation near $1.4 trillion, Tesla is being valued less as a business and more as a thesis. The numbers themselves are almost mischievous. A trailing P/E above 340 and a forward multiple near 180 would be ambitious even for a pure software firm, let alone a company still generating the majority of its revenue from manufacturing. Yet Tesla sits here comfortably, as if gravity were more of a suggestion than a law. The tension is unmistakable. The financials describe a company still wearing steel-toe boots. The valuation assumes it

Stretch Marks on a Premium Empire For years, $Lululemon Athletica(LULU)$ was treated like the Hermès of yoga pants. The brand did not simply sell leggings; it sold aspiration, discipline, wellness, and the subtle social signal that you probably own a standing desk and drink something involving oat milk before sunrise. Investors adored the formula because it looked nearly bulletproof: premium pricing, cult-like loyalty, industry-leading margins, and a customer base wealthy enough to shrug off inflation with the emotional resilience of a Labrador wearing AirPods. Now the market is asking a far more uncomfortable question: what happens when aspiration becomes ordinary? When aspiration scales, exclusivity quietly begins to fracture That, to me, is the

Scarcity, Not Steel I’ve started to think of Vertiv less as an industrial manufacturer and more as a toll collector on a very crowded motorway—except the motorway is overheating, and everyone is willing to pay extra to keep moving. What it really sells is not equipment, but relief: relief from thermal limits, power instability, and hyperscale timelines that now resemble high-frequency trades rather than infrastructure planning. AI traffic surges; infrastructure quietly charges for every degree This is where 'constraint arbitrage' stops sounding clever and starts sounding accurate. Vertiv sits precisely where digital ambition collides with physical limitation, and right now, limitation is winning just enough to be profitable. As GPUs grow more power-dense and thermally unforgiving, the bott

The Biological Gold Rush Nobody Modelled Properly I have watched countless companies try to rebrand themselves as ‘platforms’, usually with the subtlety of a re-labelled spreadsheet. $Eli Lilly(LLY)$ feels different. What I am seeing is not a pharmaceutical firm stretching into adjacent markets, but infrastructure being laid beneath a global metabolic economy. While much of Wall Street remains hypnotised by silicon and servers, Lilly is quietly monetising something far more persistent: human biology. The demand curve here is not cyclical, nor particularly sensitive to gadget refresh cycles. It is, quite literally, embedded in the global population. Biology, not silicon, may become the century’s most profitable infrastructure The recent share price

From Metronome to Maverick The transition from Tim Cook to John Ternus feels, to me, like swapping a world-class conductor for a lead guitarist. Both can produce brilliance, but one is measured in precision and tempo, while the other occasionally leans into controlled chaos and sees what happens when the amp goes a little too far. For years, Apple Inc. has behaved like that conductor-led orchestra: perfectly timed, obsessively disciplined, and almost unnervingly consistent. Under Cook, $Apple(AAPL)$ became the market’s most reliable profit engine—less a growth story in the traditional sense, more a compounding machine disguised as a consumer tech company. That consistency is exactly what justified a premium valuation. When investors pay over 30 ti

The Market’s Favourite Noise Machine There are moments when a stock stops behaving like a spreadsheet and starts behaving like a headline. Boeing is firmly in that camp. At roughly $231, pressing toward the upper end of its 52-week range, it is trading less like a recovering industrial and more like a narrative in motion. I find the fascination understandable. $Boeing(BA)$ sits at the crossroads of geopolitics, energy shocks, and industrial execution. Yet what interests me is not the noise itself, but whether that noise is masking a deeper repricing of what Boeing actually is: a constrained asset with unusually visible demand. The market appears to be pricing Boeing as though its problems are cyclical. I am not convinced they are. When noise outpace

The Bottleneck Nobody Can Ship I think the market is still treating AI as a race for better chips. That made perfect sense not long ago. It feels slightly outdated now. The constraint has shifted—from silicon to electricity. Training and running large-scale AI systems now requires sustained energy loads measured in tens, sometimes hundreds, of megawatts. That is not a procurement problem; it is an infrastructure one. You can queue for GPUs. You cannot queue for grid capacity in quite the same way. There is a slightly absurd reality here: we are building machines capable of simulating intelligence at scale, yet their deployment hinges on whether the local grid can cope. It is as if Formula 1 has been reduced to arguing over petrol stations. That mismatch is where I think

The Network That Doesn’t Need the Ground I do not see AST SpaceMobile as another entrant in the satellite race. I see it as a direct challenge to the economic architecture of global telecoms. Traditional operators spend decades and billions layering towers, fibre backhaul, and spectrum licences to chase incremental coverage. $AST SpaceMobile, Inc.(ASTS)$ is attempting to bypass that entire stack—an approach that, on paper, makes building thousands of towers look almost quaint. If this works at scale, coverage stops being a geographic constraint and becomes a capacity question. That is a subtle but important shift. Instead of competing for subscribers in saturated urban markets, the company is effectively unlocking a new tier of demand—users who we

A King Meets His Reflection I’ve always found the market becomes most revealing when a company stops being judged on what it does and starts being judged on what it represents. $NVIDIA(NVDA)$ has reached that point. At nearly $5 trillion in market capitalisation, it now functions as a proxy for artificial intelligence and, by extension, the market’s confidence in future growth itself. That shift sounds subtle, but it changes how the stock behaves. Nvidia is no longer just analysed—it is interpreted. Nvidia has shifted from asset to narrative anchor And that’s where things become interesting. Because when a company turns symbolic, expectations tend to drift away from operational reality. Nvidia’s challenge in this second act is not proving it is ex

When Silicon Meets the Socket AI scaled fast—power infrastructure didn’t get the memo For years, I watched investors obsess over who would build the fastest chips and the smartest models. It was all very glamorous—like watching a Formula 1 race where everyone kept upgrading the engine. In 2026, the race has hit a rather less glamorous obstacle: the fuel tank. The constraint has shifted—quietly but decisively—from compute to electricity. Data centres are no longer just temples of silicon; they are industrial-scale energy sinks, and the grid is starting to wheeze under the pressure. Turns out, intelligence at scale requires an almost unfashionable amount of power. That’s where I see $GE Vernova Inc.(GEV)$ stepping in—not as a participant in the AI tr

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