Opening the Ledger

Purpose — Why I Am Writing This

This is my first quarterly report, where I will be sharing my current portfolio allocations, the changes I have made, and the reasoning behind them. The discipline of writing it down is most of the point. A thesis committed to paper before the outcome is known is a thesis that can be checked later — against what actually happened rather than against what I remember believing.

My goal is to add around half of my pay cheque into the portfolio each month. That has a consequence worth stating plainly up front, because it will otherwise look like erratic behaviour: the portfolio is still relatively early and small in comparison to my monthly contributions. Changes in allocation will therefore look quite drastic over the short term.

A position can double in weight without my having any stronger view on it, simply because two months of contributions went in that direction. Conversely, a holding I still like can drift down the table purely through arithmetic. For the next several quarters, read the allocation table as a record of where new capital went at least as much as a statement of relative conviction. That distinction will fade as the base grows and monthly contributions become a smaller share of the whole.

What I am looking for

The portfolio is consistent with my philosophy: I invest in high-quality companies trading at a discount to fair value, where the discount exists because fears of disruption have impacted market sentiment rather than because the business has deteriorated.

The pattern is specific and it recurs. A dominant business is confronted by a technology or a regulatory threat that the market believes is existential. The narrative is simple and easy to headline. The defence requires understanding the incumbent's actual moat — usually distribution, data, switching costs, or network density rather than the technology itself. The market discounts the whole business on the narrative, while the fundamentals keep compounding.

Every holding in this letter fits that shape. Google was bought when ChatGPT was going to replace Search. Mastercard when stablecoins were going to disintermediate the card networks. Microsoft when capex fears and AI competition in Office were the story. S&P Global when AI was going to commoditise analytics. Uber into the robotaxi panic. SoFi into a rate cycle that has swung from expected cuts to expected hikes.

Allocation — Current Positions

Seven holdings. The concentration in SoFi is deliberate but larger than I would like as a steady state, and I address that directly below.

Weights are of invested capital at the date of this letter and will move materially with monthly contributions. Cost basis is my average entry price per share across all purchases.

S&P Global SPGI · 18.4%

One of the strongest, widest-moat companies in the world, trading at a historic discount on two fears that I believe will thin out: AI disruption of its analytics business, and higher interest rate expectations.

On the moat: the credit ratings business is effectively half of a global duopoly, protected by regulation, decades of accumulated methodology and the simple fact that debt issuers need a rating the market recognises. The indices business — the S&P 500 itself among them — earns fees on assets tracking benchmarks that have become industry standards, which is close to a toll on the growth of passive investing. Neither is a business a competitor can simply build.

On AI: I expect it to be a tailwind rather than a threat, for the same reason it has been at Google and Microsoft. The scarce asset is not the ability to analyse data — it is the proprietary data itself, the regulatory licence, and the trust of the market in the output. AI makes S&P Global's data more valuable and its analytics products cheaper to deliver. A model with no access to the underlying dataset cannot substitute for the dataset.

On rates: higher rates suppress debt issuance, which is the cyclical part of the ratings business, and this is the mechanism behind the current de-rating. But the rate pressure is being applied to combat what are essentially one-time hits from higher oil prices and tariffs rather than an entrenched inflation problem. Those are supply-side shocks that work through the numbers and out again. As they do, issuance recovers.

My expectation is that both threats thin out over time, the company continues to grow healthily, and the multiple returns to something closer to its historical range. This is the most patient position in the portfolio — nothing needs to happen quickly for it to work.

SoFi Technologies SOFI · 17.6%

My second-largest position and, I would readily accept, my highest-risk name. I have been a shareholder for around two years and the performance of the management team over that period has been outstanding. Anthony Noto made a hard, specific, falsifiable promise during the 2022 depths — GAAP profitability by Q4 2023, made while the company was losing money heavily and widely doubted — and delivered it on schedule. He delivered the bank charter. He said he would drive stock-based compensation to single-digit margins and it has fallen from roughly 15.5% of revenue in Q4 2022 to about 6.3% today. That is a track record of doing what he said he would.

The company continues to show extreme growth despite lesser macroeconomic conditions. Q2 2026 delivered adjusted net revenue up 40% to $1.2bn, net income up 61%, members up 35% to 15.8 million, and a nineteenth consecutive quarter clearing the Rule of 40 — with a score of 70. Cross-buy accelerated to 51% of new products opened by existing members, up from 35% a year ago. Credit improved rather than deteriorated, with personal loan charge-offs falling to 2.62%.

What I am underwriting rests on three legs. First, the transition: that SoFi converts today's spread-driven profits into a fee-and-scale business before the spread compresses. Fee-based revenue is already 39% of the total. Second, the cohort: SoFi has positioned itself as the all-in-one banking app for a younger, high-earning demographic, and those members age into higher incomes, larger balances, and more products — revenue per member should rise without proportionate acquisition cost. Third, the asymmetry: the macro conditions currently suppressing the stock are already discounted, and improvement from here is largely uncompensated upside.

That last point is why I have built the position up heavily as sentiment reached what I consider undeservedly low levels. Rate expectations swung from cuts to one or two hikes on an energy-driven inflation shock, and SoFi — with a beta above two and genuine rate sensitivity — was marked down for a macro environment it did not cause. Those conditions will likely improve over time and act as a new tailwind, at which point better loan-sale execution, higher origination volume, improving credit and multiple expansion all arrive together.

What gives me most confidence is how management has handled the macro turn. Despite rate expectations moving from cuts to hikes over the course of the year — the single biggest headwind to SoFi's lending and capital-markets economics — management has maintained its guidance. At Q2 it raised full-year adjusted net revenue guidance to $4.75–4.85bn, while holding adjusted EBITDA at around $1.6bn, adjusted net income at around $825m, adjusted EPS at around $0.60, and member growth of at least 30%. A business that holds its targets through a rate environment moving against it is telling you something about the durability of the model.

Microsoft MSFT · 15.6%

I started this position under $400 per share, when the company was mispriced on two related anxieties: the scale of AI capital expenditure, and the threat of competing models — Claude in particular — taking share in the productivity software space that Office has owned for decades.

Microsoft's response has been the most instructive part of the story. Rather than defend a proprietary model at all costs, it has implemented Claude into its own applications alongside its other AI offerings, and continues to improve the product with whatever works best. That is a company treating AI as an input to its distribution advantage rather than a threat to it — and it is the same pattern as Google. The supposed disruptor becomes a tailwind because the incumbent owns the surface where the technology gets used.

On the capex question, the distinction that matters is between spending on a hope and spending against a revenue line. Microsoft is directly monetising its capital expenditure through Azure — a business that is both massively profitable and still growing fast. The capex is building capacity that is already being sold. That is a materially different proposition from infrastructure spend in search of a business model.

Underneath the AI narrative sits the thing that actually makes Microsoft difficult to dislodge: enterprise switching costs. Identity, security, productivity and cloud are sold as an integrated stack into organisations that would face enormous friction unpicking any single layer.

Mastercard MA · 15.3%

An extremely wide-moat compounder, mispriced on fears that stablecoins will disintermediate the card networks. I think this is close to the clearest example in my portfolio of a narrative threat meeting a business that is almost impossible to disrupt.

The network is a genuine two-sided moat built over decades: billions of cardholders on one side, tens of millions of accepting merchants on the other, and neither will move without the other. A new payment rail does not merely need better technology — it needs to solve acceptance, fraud liability, chargebacks, cross-border settlement, regulatory compliance, and consumer protections simultaneously, and then persuade both sides to switch at once. That is why challengers have consistently ended up running on the networks rather than replacing them.

Mastercard's own response reinforces the point: it has acquired a stablecoin company rather than resisting the technology. If stablecoins become a meaningful settlement layer, Mastercard intends to be part of the plumbing, not the roadkill. Again — the disruptor absorbed as a capability.

The strongest evidence for the moat is what the actual bear case has to be. Mastercard's biggest genuine threat is regulation — interchange caps, antitrust intervention — which is a testament to the strength of the business. When the only credible way to hurt a company is for a government to legislate against its profitability, competition is not the problem.

Meanwhile it continues to grow quickly with expanding margins. And it carries a property I value highly in the current environment: Mastercard is one of the best holdings to own if inflation is a worry, because it takes a percentage of the value of purchases. Higher prices mechanically increase its revenue per transaction. It is one of the few businesses where an inflation shock flows straight into earnings rather than against them.

At today's price of around $600 per share, against my cost basis of $523.51, Mastercard trades closer to fair value than when I bought it. I am happy to hold at this level, but I am looking to add on weakness rather than chase the price higher.

Uber Technologies UBER · 14.6%

Most recent addition — full research note →

My most recent addition and the position that follows the Google template most directly: a high-growth compounder trading at a historically low valuation because of a disruption I do not believe will land materially.

The business is compounding at a scale where it should not be able to. Gross bookings grew 22% and trips 18% in Q2 2026, the fourth consecutive quarter above 20%. Non-GAAP operating income grew 40% — profit growing nearly twice as fast as volume — with margin on gross bookings expanding from 3.3% to 3.7%. Trailing free cash flow crossed $10bn for the first time and the share count is shrinking. Growth is coming from genuinely durable places: fewer than 10% of eligible consumers in sparse markets have used Uber in the past year, against over 50% in dense markets, and new low-cost formats finally make thin-density geographies viable.

On autonomy, scaling robotaxis is a manufacturing and capital problem, not a software problem — AV trips are currently under 0.5% of Uber's roughly 300 million weekly trips, and until an operator has metro-wide density its ETAs are long. Consumers choose the shortest wait, not the brand. The early evidence supports this: in Phoenix, Los Angeles and San Francisco, where robotaxis are most available outside the Uber app, Uber's trip growth accelerated. Autonomy appears to be expanding the market rather than taking share.

Longer term, the case rests on commoditisation. Nvidia's DRIVE platform is turning autonomy into a reference architecture any manufacturer can buy, which makes Waymo's decade-long head start purchasable. When supply commoditises, the winner is the aggregator holding the demand, the data and the network effects. Uber is uniquely placed because it already has the demand — the moment its fleets go live they monetise instantly, which means it can afford to be behind. Its roughly thirty AV partnerships exist to ensure no single supplier gains leverage over it.

There is also a structural argument for the hybrid network. Demand is enormously variable — a Friday night, a rainstorm, a stadium emptying. A pure-AV operator must either build fleet for peak and leave capital idle most of the week, or fail customers at the moments they care most. Human drivers flex to absorb peaks with the capital cost on someone else's balance sheet. And a portion of riders will simply prefer a human driver. Both point to hybrid as a durable architecture rather than a transitional one.

Finally, even the bear case is survivable. Mobility is the exposed segment and it is becoming a smaller share of the business — Delivery is nearly the same size, growing faster, with faster margin expansion, and is untouched by robotaxis. Advertising is compounding at roughly a $2.5bn run rate. A scenario where AVs take real Mobility share is damaging but not catastrophic from a multiple that already discounts it.

Amazon AMZN · 11.1%

A high-quality business with low disruption risk that continues to grow healthily. Amazon is less a mispriced-fear position than the others here and more a straightforward compounder I want to own more of at the right price.

The interesting part of the current story is the rapid expansion in margins driven by AWS. Amazon spent two decades being valued as a low-margin retailer that happened to own a cloud business; the mix has now shifted far enough that the profit engine is unambiguous. Retail provides scale, logistics density and a captive customer relationship; AWS and advertising provide the margin.

The part I find most compelling on a long view concerns operating leverage. Because of the sheer size of the company and the number of people it employs, Amazon is among the best-positioned businesses in the world to benefit as AI and robotics mature. Most companies applying AI are looking for incremental efficiency. Amazon has warehouses, delivery networks, customer service operations and a labour base measured in the hundreds of thousands — an automation opportunity of a scale almost nobody else possesses. If robotics reach a higher level, the operating leverage available here is enormous.

This is a position I am looking to add to. I am being patient for more appealing prices rather than paying up, which is the correct discipline for a business whose quality is widely recognised and therefore rarely on sale.

Alphabet GOOG · 7.4%

The template for everything else in this portfolio, and the position that most shaped how I look for opportunities.

I started this position at a cost basis of $175.16, when the dominant concern was that ChatGPT would replace Google Search — an existential narrative, universally reported, and reflected in a multiple of roughly 16 times earnings for one of the highest-quality businesses in the world.

The early findings showed the opposite of the thesis the market was pricing. Search was benefiting from AI, particularly through AI Overviews, which increased engagement rather than cannibalising it. Gemini was consistently among the highest-performing models available, so the technology gap the bear case assumed did not exist. And the other segments — YouTube and Google Cloud — were continuing to execute at a high level and were being valued at close to nothing inside the conglomerate discount.

Google's subsequent appreciation is the clearest evidence I have that this pattern is real rather than a story I tell myself. Search was not disrupted; it was strengthened, and growth accelerated. I lean on this case explicitly when assessing new opportunities — it is the reference point for what an overpriced disruption fear looks like on the way in, and how quickly sentiment can re-rate once the evidence arrives.

Why it is now my smallest position

I trimmed substantially around $350 per share — roughly double my cost basis — because I was low on cash and wanted to allocate to other bets where I saw more asymmetry. That is the honest reason: not a change of view on Google, but a funding decision. The capital went into positions where the disruption fear was still loud and the discount still available, which is where my method says the returns are.

I continue to hold the remainder, and intend to. Google trades at what I consider a fair valuation rather than a discounted one, and it remains an incredibly strong business — Search, YouTube, Cloud, and a genuinely leading position in frontier models. Fair value in a business of this quality is a perfectly good place to leave capital compounding. It is simply no longer the best use of a marginal pound while other holdings are mispriced.

Changes This Quarter — What Moved, and Why

Initiated: Uber

The newest position, opened into the post-earnings weakness. Shares fell 5.3% on a Q3 guidance midpoint that missed consensus by roughly 0.14% — about $80m on a $59bn number — after four consecutive quarterly beats had elevated expectations. That is a sentiment reset rather than a thesis break, and part of the shortfall is a mechanical acquisition-lapping effect that will wash out. Buying into a de-rate caused by positioning rather than fundamentals is precisely the set-up I look for.

The quarter itself was also really encouraging on the AV plans specifically. Uber is live with autonomous vehicles in seven cities and targeting fifteen by year end, has committed around $10bn to AV partners and fleet purchases, and is working with Nvidia towards 100,000 vehicles from 2027. Most importantly, in Phoenix, Los Angeles and San Francisco — the markets where robotaxis are most widely available outside the Uber app — Uber's trip growth accelerated rather than slowed. That is the first real evidence on the question the market is most worried about, and so far it points my way.

Added significantly: SoFi

The largest capital deployment of the quarter. Sentiment reached levels I consider undeserved given that the operating results were, on almost every measure, the best in the company's history. The de-rating was driven by a swing in rate expectations from cuts to hikes on an energy-driven inflation shock — an external macro variable, not a deterioration in the business.

Trimmed: Alphabet

Reduced substantially around $350 per share, roughly double my cost basis of $175.16, taking Google from a core position to my smallest at 7.4%. The reason was funding rather than conviction: I was low on cash and wanted to allocate to other bets where the discount was wider. Google now trades at fair value rather than a discount, which made it the logical source of capital — but I want to be clear that this was a cash-management decision, and those are the ones most likely to look wrong later.

On position sizing

The resulting book is more evenly weighted than I expected when I set out to write this — 18.4% at the top, 7.4% at the bottom, and five positions clustered between 14% and 19%. The top three account for 51.6%.

That is a reasonable shape. What it understates is risk-weighted concentration: SoFi and Uber together are 32.2% of the book and carry substantially more volatility than Mastercard or S&P Global at similar weights. My intention is to correct that by adding elsewhere rather than by trimming. Given that monthly contributions are large relative to the existing base, dilution is a faster and less costly mechanism than selling.

Amazon at 11.1% is the position I most want to increase, and I am waiting for a more appealing price rather than paying up. Mastercard would take further capital on weakness.

Risk — Where This Portfolio Is Exposed

A letter that only sets out the case for each holding is marketing. These are the three structural risks I see in the portfolio as it is currently constructed.

Sources & Disclosure

Basis of preparation. Company figures are drawn from the most recent reported quarterly results of each holding and from contemporaneous market reporting, current to September 2026. Allocation weights are of invested capital at the date of this letter. Cost basis figures are average entry prices per share. Where a view is an inference or estimate rather than a reported figure, it is identified as such in the relevant research note.

Disclaimer. This is a personal record of my own capital allocation and reasoning. It is not investment advice, a recommendation, or an offer to buy or sell any security. It has not been prepared in accordance with regulatory requirements designed to promote the independence of investment research. I hold positions in every security discussed and my views are therefore not disinterested. Forward-looking assessments are inherently uncertain and may prove incorrect. Anyone acting on this material does so entirely at their own risk and should consult a suitably qualified and regulated adviser.