The market thinks it sees a card company. I think it is watching a financial operating system emerge. Mastercard has become one of the stranger stories in the market this year. Here is a business generating absurd profitability, growing revenue at double digits, printing cash with the efficiency of a central bank photocopier — and still underperforming the S&P 500 by a painful margin. The stock is down more than 12% year-to-date while the broader market has surged. Normally, that sort of divergence appears when margins are compressing, growth is slowing, or investors realise the story was built on optimistic arithmetic. None of those things are happening here. The market hesitates even while long-term buyers quietly accumulate Instead, I think the market has become oddly conservative a

For most of the AI boom, investors focused obsessively on who makes the chips and who rents the cloud capacity. I think that framing is already becoming outdated. The real constraint inside modern AI infrastructure is no longer raw compute power alone; it is the speed and efficiency with which thousands of GPUs communicate with one another. That shift matters because idle GPUs are financial vandalism. Hyperscalers are spending tens of billions building AI clusters, but if the networking layer cannot move data efficiently between processors, expensive compute hardware simply sits there underutilised. In practical terms, networking has evolved from a supporting technology into one of the central determinants of AI economics. That is why Arista Networks has quietly become one of the most stra

Inference became theatre. Investors arrived before the final act The IPO That Arrived Exactly on Cue Cerebras Systems did not quietly tiptoe onto the public markets. It marched in wearing a brass band, carrying a wafer-sized silicon dinner plate, and demanding Wall Street’s full attention. At one point, investors valued the company at roughly $95 billion following its explosive debut, briefly treating it less like a semiconductor firm and more like the AI equivalent of discovering fire. What fascinates me is not merely the technology. The real story is the timing. Cerebras went public at the precise moment the AI narrative flipped from training models to running them. For the past two years, investors obsessed over who could build the biggest large language model. Now the market cares abou

The AI Toll Collector Nobody Expected Most investors still think $Advanced Micro Devices(AMD)$ is fighting Nvidia for the AI throne. I increasingly think AMD may profit even if it never wins the crown. That is what makes the stock so dangerous at today’s valuation. At nearly $700 billion in market value and more than 140 times trailing earnings, AMD is no longer priced like a challenger. It is priced like a future AI superpower. Yet unlike Nvidia — whose software ecosystem behaves less like a product suite and more like a fully electrified railway network that developers are already locked into — AMD is still racing to prove its moat can widen fast enough to justify the premium investors have assigned it. The paradox is brutal. AMD may not need to

The Cybersecurity Vendor That Quietly Became an Operating System There is a strange irony surrounding CrowdStrike Holdings, Inc.. Despite becoming one of the most dominant software platforms in enterprise security, many investors still discuss it as though it merely sells antivirus software with better marketing and cooler conference booths. That framing now looks wildly outdated. What I increasingly see is a company evolving into the operating layer for modern cybersecurity operations — a position that tends to create frighteningly durable economics. Once enterprises centralise endpoint security, identity protection, cloud workload monitoring, threat intelligence, SIEM, and incident response into a single architecture, the cost of ripping it out becomes about as appealing as replacing a j

Now You Seat Me, Now You Don’t ServiceNow just delivered the sort of quarter that normally sends software stocks sharply higher. Subscription revenue rose 22% year over year to $3.67 billion. Customers spending more than $1 million annually on its AI product, Now Assist, surged over 130%. Management then raised its AI revenue target from $1 billion to $1.5 billion midway through the year. The reward? A near-20% share price collapse in a single trading session. That reaction tells me something important about where markets are in 2026. Investors are no longer questioning whether AI adoption is real. They are questioning whether AI quietly destroys the business model that made enterprise software wildly profitable in the first place. AI may be replacing the very seats SaaS once monetised And

The Case for Novo Nordisk at a Distressed Multiple Novo Nordisk has had the sort of year that makes investors check their portfolios twice just to confirm the numbers are real. The shares have collapsed from a 52-week high of 81.44 to the mid-40s, sentiment has deteriorated sharply, and a company once treated like Europe’s growth crown jewel is now trading at barely 10–11 times earnings. That disconnect is exactly why I think the opportunity has become so compelling. Markets see collapse. The fundamentals suggest strategic transformation The market is behaving as though Novo’s growth engine has permanently stalled. I believe something very different is happening: the company is transitioning from a blockbuster obesity story into a broader cardiometabolic platform, while simultaneously open

The moat nobody notices until it leaks I think investors still underestimate how unusual Berkshire Hathaway’s insurance machine really is. Most conglomerates collect capital and then allocate it. Berkshire’s genius was that it collected capital while often being paid to hold it. That is the magic of float. Warren Buffett effectively turned insurance liabilities into one of the cheapest investment funding sources in financial history. The problem is that float only stays magical if underwriting discipline remains exceptional. Berkshire’s float machine still works — just slightly less flawlessly That is why I think Berkshire Hathaway’s insurance division is entering its first genuine post-Buffett stress test. Not because the business is broken, and certainly not because GEICO suddenly forgot

The Stablecoin, the Chain, and the Protocol Most investors still analyse Coinbase as though it lives and dies by whether Bitcoin is having a good hair day. That framing increasingly looks outdated. In 2026, Coinbase resembles less of a crypto casino and more of a digital toll road operator quietly building the financial plumbing for autonomous commerce. That distinction matters enormously. The market remains obsessed with Coinbase's trading revenue volatility, its lofty valuation multiples, and its tendency to behave like a caffeinated semiconductor stock whenever crypto sentiment swings. Yet beneath the noise, something far more interesting is happening: Coinbase is assembling an integrated stack around stablecoins, settlement infrastructure, and programmable payments that no other listed

For years, investors treated Amazon as a retailer with a cloud side hustle. Then they treated it as a cloud titan carrying an unruly retail operation on its back like an exhausted pack mule. I increasingly think both views miss what Amazon has become: an advertising machine disguised as civilisation’s most convenient delivery network. That distinction matters because markets often misprice companies during identity changes. Amazon’s business model has evolved faster than the valuation framework investors use to analyse it. Wall Street still debates retail margins and AWS growth rates while a higher-margin engine hums beneath the floorboards, subsidising nearly the entire empire. Advertising is no longer an accessory business for Amazon. It is becoming the economic glue holding the ecosyste

The Most Important Company in AI Isn’t the One Designing the Chips Wall Street spent the past three years obsessing over who could design the fastest AI accelerator. $NVIDIA(NVDA)$ became the poster child, $Advanced Micro Devices(AMD)$ became the challenger, and $Intel(INTC)$ continued its long-running hobby of promising turnarounds with the confidence of a man trying to restart a lawnmower in the rain. But by mid-2026, the market has quietly shifted to a more uncomfortable question: who can actually manufacture these chips at scale? That answer is Taiwan Semiconductor Manufacturing Company. The AI economy runs atop foundations most investors never see I think inv

Broadcom’s Quiet Empire The first wave of artificial intelligence created a simple investment story: buy the companies building the fastest chips and enjoy the ride. The second wave is becoming far more complicated — and, in my view, far more interesting. AI is now colliding with economic reality. Hyperscalers are discovering that training and deploying large language models at global scale requires far more than raw compute power. It demands lower energy costs, faster connectivity, optimised architectures, and infrastructure capable of moving staggering amounts of data without collapsing into latency chaos. That shift is precisely why I believe Broadcom has quietly become one of the most strategically important companies in the entire AI ecosystem. Broadcom is no longer simply benefiting

Why the Most Important AI Infrastructure Battle May Be Happening Far Below Big Tech’s Pay Grade For the past two years, the AI investment boom has revolved around giants. $NVIDIA(NVDA)$ sold the shovels, $Microsoft(MSFT)$ rented the mine, and $Amazon.com(AMZN)$ charged everybody extra for bringing their own wheelbarrow. Yet while Wall Street obsessed over trillion-dollar firms, a quieter shift began unfolding underneath them. DigitalOcean was never supposed to become one of the defining AI stocks of 2026. It lacked the scale, balance sheet and political gravity of hyperscalers. For years, it occupied a fairly unglamorous corner of the cloud market serving startup

Reinvention in Plain Sight Alphabet is no longer the market’s favourite ‘obvious’ company, which is precisely why I find it compelling. At just under $4.7 trillion in market capitalisation and trading near its highs, it looks fully appreciated on the surface. A trailing P/E of roughly 29 and a forward multiple above 30 do not scream bargain. Yet those numbers obscure a deeper transition: Alphabet is shifting from a predominantly advertising-driven enterprise into a vertically integrated AI infrastructure company. This is not a cosmetic pivot. It is a structural rewrite of how revenue is generated, how costs are controlled, and how competitive advantage is sustained. The market, in my view, is still pricing Alphabet as a highly efficient incumbent rather than a company attempting to control

A Utility Hiding in Plain Sight I think the mistake most investors are making with Cellebrite is not analytical—it’s categorical. The market is still trying to decide whether this is SaaS, AI, or cybersecurity, when in reality it is drifting into something closer to sovereign digital infrastructure. That framing matters upfront because it reframes everything else: revenue durability, pricing power, and competitive risk. Governments do not treat digital forensics as discretionary spend. They treat it like evidentiary plumbing—if it fails, the consequences are legal, not just operational. The broader 'DiSaaSter' narrative—that AI compresses software value by commoditising features—has merit in horizontal SaaS. But it is being applied far too bluntly. Cellebrite is not selling convenience; it

The AI Boom’s Most Uncomfortable Question Super Micro Computer has become the stock market equivalent of a Formula One car being rebuilt while still racing at 300 kilometres per hour. One side of the market sees an AI infrastructure champion powering the next phase of Nvidia’s Blackwell rollout. The other sees a company drowning in legal risk, collapsing margins, and geopolitical scrutiny. Personally, I think both camps are right — and that is precisely what makes $SUPER MICRO COMPUTER INC(SMCI)$ one of the most fascinating stocks in the market today. Most AI commentary still treats Nvidia as the sole protagonist of the artificial intelligence boom. Yet the uncomfortable truth is that GPUs are useless without the server infrastructure surrounding

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