the nangoy foundation
A personal record of one small book, run patiently.
october 2026 · book as of 30 october 2026
October was September run backwards. Last month a handful of giants held the index up while the average stock fell hard; this month the giants did the falling and everything else carried. The Nasdaq finished October down 5.3% while the equal-weighted S&P 500, where every company counts the same, rose 5.0%. In September those were the other way round, the Nasdaq up about 2% and equal weight down nearly 5%. What is worth noticing is that the headline S&P 500 was slightly negative in both months, down 0.5% in September and 0.9% in October. If you only looked at the index you would think nothing happened twice, and you would have missed the whole story both times.
What flipped it was the market changing its mind about what it will pay for. For two years it paid for AI growth. In the last week of October it stopped paying for growth on its own and started asking about cash. Alphabet reported on the 28th with revenue up 25% and its cloud business up 103%, and the stock fell 12.6% the next day, the worst session in its history by most accounts, taking something close to half a trillion dollars of value with it. The reason was not the top line. Core earnings came in at $2.67 a share against $3.02 expected, net income fell 25%, and free cash flow was negative $9.1 billion. Markets are no longer paying just for growth.
Microsoft is the one worth sitting with, because it did not get rewarded so much as spared. It rose 3% after hours on what read as spending restraint, having cut its outlook for the year to about $175 billion, and then gave all of it back and closed the next day flat. Part of that lower spending number comes from an accounting change: it extended the assumed useful life of its data centres and buildings from 15 years to 25, which moves some spending off the line investors were watching. Meanwhile its free cash flow covered just 51% of net income, down from 92% a year earlier. So Alphabet was taken apart for burning cash and Microsoft walked away untouched, while its cash conversion went the same direction. The new discipline is real. It is also being applied crudely.
Amazon was the day the market argued with itself. It reported on the 29th with AWS growing 39%, its fastest in nearly five years, alongside $77 billion of capital spending in a single quarter and free cash flow over the trailing year of negative $42 billion, against positive $15 billion a year ago. The stock was down 7.5% before the open, down 6% at its worst, and closed the day off just 1.4%. Nobody seems sure whether to treat the spending as the proof or the problem.
Underneath the headlines the AI complex has already split in two, and the dividing line is debt. The companies that borrowed to build are a long way down from their highs: CoreWeave about 57%, Oracle 56%, Western Digital 54%, Lam Research 47%. The ones funding it out of their own cash flow have barely moved. Nvidia is off about 12%, Taiwan Semi 14%, AMD 16%. One variable explains a 40-point gap in outcomes. I’d rather own the second group, and I do.
The other thing that changed this month was how the bond market behaved. For most of this year the chain ran one way: the Iran war pushed energy prices up, inflation stayed sticky while growth held, rate-hike odds rose, borrowing costs rose, equities got marked down for it. Yields were the cause and equities were the effect. In October it ran the other way. On the days the market fell, money went into Treasuries and yields fell with it. The 10-year went from 5.29% at the end of September to 4.94%, down 35 basis points, with most of that between the 5th and the 22nd, before any of the big earnings. The count is the clearest way to see it: in September the market fell on fifteen days and yields rose on every single one of them. In October it fell on eight days and yields fell on six. And the thing people were running from was not equities in general. It was the debt of the AI buildout, where one cloud provider alone is carrying something like $50 billion of it, and the money came out of corporate bonds as much as out of stocks. Worth adding that oil rose sharply over the month while all this happened. If this were an inflation story, that is not the way round it would read.
As for the Fed, I don’t have a view, and I’ve stopped pretending I need one. It held on the 28th at 3.75–4.00%, though both dissenting votes wanted a hike and the market is now pricing roughly 80 to 90% odds of one in December. Inflation is still 3.7% headline and 2.6% core. The Fed will do what it does regardless of what I think about it. What I can do is position for it and buy good businesses when they go on discount. Same with the midterms on Tuesday: no view, except that an election dropped into a fragile market means more volatility, particularly tied up with the Iran war, where officials have been warning allies to prepare for the possibility of renewed action once the vote is behind us.
So: the AI trade isn’t over, but it is fragile, and I’d be careful here. That is a statement about the market, not about what I am doing, and the two don’t have to match. Four AI-linked names are about a quarter of my book and my view on each of them hasn’t changed. Nearly 29% sits in cash and Treasury bills, and with the Fed still leaning tighter I am content to leave it there. I want prices to fall. I’m a net buyer, so lower prices are the opportunity, and if they don’t come, that isn’t a problem. It never is.
Cash & T-bills combines the cash balance with the SGOV position — the true dry powder, now a little under a third of the book. US equity includes the IVV index holding. Emerging Asia is Tencent Music and Taiwan Semi; International developed is the VEA holding.
SGOV 0–3 Month Treasury · 12.9%Treasury bills, where cash sits when I’m not ready to spend it. It pays me to wait, and with the Fed still leaning toward one more hike, waiting is being reasonably well compensated.
VEA Vanguard Developed Markets · 12.1%The un-clever part of the book, on purpose: cheap, broad exposure to developed markets outside the US, and my bet that the whole world isn’t priced like the S&P. October was the month that bet stopped looking eccentric.
MCO Moody’s · 11.0%A toll booth on global debt issuance, and one of the good businesses I like. It fell 10% in September and took most of that back in October. I’m happy with it as a business and don’t much care where the price goes in the short term. I’d buy more on a bigger drop.
AXP American Express · 8.8%Reported on the 23rd. Profit up 31%, which isn’t real growth: nearly all of it is the gain on selling its business-travel stake, less restructuring. Strip the one-offs and it grew about 13%. The parts I own it for got better — annual card fees growing 17% and accelerating, and credit the best it has been, with 1.8% of balances written off and 1.2% late. Small-business spending stayed soft and the promised turnaround quietly lost its date, but that was never what I was betting on; with a business this size and this mature I look at the whole thing, not one segment. There is a $350m fine over money-laundering controls and a 5% headcount cut to watch. A quiet quarter. I expect Amex to still be running in ten years, and I’m a net buyer here — if it falls, I’ll be happy to own more.
NVDA Nvidia · 7.8%A monster business at a price that demands near-flawless execution. I own it, I’m not adding, and I count it as part of one AI position rather than a separate bet. It’s about 12% off its high, which in a month when the borrowers fell by half tells you which side of the line it sits on.
IVV iShares Core S&P 500 · 6.3%The S&P 500, in place of August’s QQQ. It puts some cash to work while nothing on my list is at my price.
AMZN Amazon · 6.2%Really an AWS bet. AWS grew 39% this quarter, its fastest in nearly five years, and that is the good news. The capital going in is the open question, and it got bigger rather than smaller: $77bn of capital spending in the quarter, and free cash flow over the past year of negative $42bn against positive $15bn a year ago. I still can’t tell you whether AWS earns a decent return on what it is pouring in — and until I can, this is a position in a business I haven’t finished underwriting. That one is on me.
TME Tencent Music · 5.8%Priced for the business to shrink, which it isn’t. The risk that matters isn’t the numbers, it’s the China ownership structure behind it.
GOOGL Alphabet · 5.3%I like the business, and I’ve stopped pretending my July valuation still speaks for it. That work is out of date and I haven’t redone it. What I’m watching is narrower. Google is eating its own search: AI Mode earns less per query than the old blue links, and the queries being eaten are the valuable commercial ones, shopping and travel and local. I’d still rather it cannibalise itself than hand those users to somebody else. The question I’m tracking is what Google has to do to make AI Mode earn more per search. There is more work owed here and I’ll do it when I have the time.
TSM Taiwan Semiconductor · 4.7%The company the whole AI complex quietly runs on, and an elite compounder on returns. I own it, I’m not chasing it up here, and I’d want more on a broad selloff.
PGR Progressive · 3.1%A best-in-class insurer in a tougher patch, with the share of premiums paid out in claims and costs still well inside the level that would worry me.
Since inception on 3 July 2025, the book has returned +17.79% on a time-weighted basis — the clean measure of how the portfolio itself performed — against +20.71% for the S&P 500 over the same stretch. The gap narrowed this month, from five points behind at the last letter to 2.9 points. It closed not because I did anything clever but because the index fell faster than I did.
Time-weighted return since inception (3 Jul 2025 – 30 Oct 2026).
October itself cost 0.8%, and almost all of it arrived at the very end. Going into the 29th I was up about 0.7% for the month. Then I lost 0.46% on the 29th and 1.06% on the 30th, and a positive month became a negative one in two sessions. Those are the two days the hyperscalers reported. My book is about a quarter AI-linked, so the thing I spent this letter describing happened to me directly, on the way out the door.
october — breadth, the other way round
Calendar October (30 Sep – 30 Oct 2026). Bars run from a zero line in the middle: losses to the left, gains to the right, on a scale of plus or minus 6%. Equal weight measured by the RSP fund’s price.
Against the S&P I was marginally ahead, and against the equal-weighted index I was nowhere. The more interesting read is what that gap says about the market rather than about me. Breadth improved while tech cooled: money spread out instead of piling into the same handful of names. After a long stretch of the index being carried by a few giants, that looks like conditions getting more normal, not less. It changes nothing about how I invest. It is just worth noticing.
And while I report a monthly number, I am not working to a month. There is no clock on this and no pressure to produce a return by a deadline, which is exactly the freedom that lets me hold 29% in cash without flinching. I trust the approach, I am still refining it, and I would rather keep working on the book than manufacture a move to fill a reporting period.
One comparison I’d rather use going forward: since inception the equal-weighted index has returned +17.89% against my +17.79%. Within a tenth of a point. The S&P’s extra three points over the same stretch came almost entirely from a handful of giants getting bigger. That isn’t an excuse, it’s a description of what I’ve been competing against.
for context — other markets, since inception
Zero line as above: losses left, gains right, on a scale running from minus 4% to plus 20%.
This month’s stock is one I don’t own, and that is the point of including it.
I did the work on Mastercard in June. It was trading around $463. I built an owner’s-earnings model on it, came out with a number, and didn’t buy. It closed October around $638. Up roughly 38% in four months, while I watched.
So the obvious question is whether I bottled it. I’ve gone back through what I wrote at the time, and the answer is no. My estimate of what the business was worth sat below where it was trading, and the price I’d actually have bought at sat well below that again. The stock was never close to my number. Not buying wasn’t nerve failing. It was the process doing precisely what it is built to do.
Worth saying what the business is, because it is the opposite of last month’s. Amex owns its whole loop: it issues the card, runs the network, and faces the merchant. Mastercard owns none of that. It doesn’t issue cards, doesn’t lend, and carries no credit risk. It runs the rails, and takes a small cut of everything that travels over them. Banks issue, merchants accept, Mastercard sits in the middle and clips. Two sides that each need the other, a network that gets more valuable the more people are on it, and almost no capital required to run it. It is one of the most wonderful business models ever constructed, and everybody knows it, which is the whole problem.
I use the same discount rate on every business I look at, and I don’t move it for quality. Moving it for quality is just picking the answer first and working back to whatever inputs get you there. Quality is meant to show up somewhere else, in how big a discount to my own estimate I insist on before buying. I still think that is the right way round.
But run that over a fortress business and it will rarely let you buy one. A hurdle that demanding, applied to a company the market has already worked out is excellent, lands below the market price nearly every time. So my rules point me at wonderful businesses and then forbid me from owning them at any price actually on offer.
That is the old argument, Graham against Munger. Demand a deep discount and sit on your hands, or pay a fair price for a wonderful business and let it compound. If you asked me which camp I am in, I’d say Munger without hesitating. If you looked at how I actually behave, you’d say Graham. Mastercard at $638 is the invoice for that gap.
I’m not going to pretend I have resolved it, because I haven’t. What I’d push back on is the idea that the last four months settle anything. A stock going up after I declined to buy it is not evidence that declining was wrong. If I only ever judge a decision by what happened next, I have stopped having a process and started having a scoreboard. The question worth asking is whether my hurdle plus a deep discount is the right bar for a business like this one, or whether I have set a test that nothing excellent can pass.
I don’t have the answer yet. I’m aware that is an unsatisfying place to end. It is where I actually am.
One move in the book this month.
On the 29th I bought one more share of Alphabet, taking it from 4.0% of the book to 5.4%. That was the day it fell 12.6%, its worst session since it listed in 2004, after a quarter where revenue grew 25% and free cash flow came in at negative $9.1 billion.
The part that stood out to me was search being disrupted, and that is something I had been expecting rather than something that blindsided me. A selloff you saw coming is the one you are supposed to be ready for.
Two things made this easy rather than brave. Alphabet was one of the smallest positions in the book, and after the add it still is: a shade over 5%, with only two holdings smaller than it. This isn’t a conviction bet and it isn’t sized like one. If I’m wrong about Alphabet it costs me something; it doesn’t do me real damage. And I’m sitting on a well-accumulated pile of cash, which is precisely what it is there for.
Whether this turns out to be a real opportunity, I don’t know. I like the business, the price is a lot lower than it was, and the position stays modest while I do the work.
What I owe it is work. I’ll be watching this one closely.