THIS ISN'T ONE CHAIN

Follow the instruments, not the logos

There is a story people like to tell about the computer industry, and it goes IBM, then Microsoft, then Apple — three companies taking turns at the front. Told that way it is a race, and the interesting question is who won.

Read the instruments instead and it stops being a race. IBM licensed its operating system in 1980 without taking exclusivity, and the standard it set became the asset Microsoft owned. Microsoft, seventeen years later, put $150 million of its own accumulated cash into a competitor that was running out of road. Apple survived, and went on to hold a securities portfolio larger than most sovereign wealth funds. The companies compete. What moves between them does not.

Edition III asked whether the closed-loop revenue structure was one company or all of them. This edition asks the same question of a different structure: when an operating company starts allocating capital instead of selling a product, is that a story about three logos, or is it a condition of the balance sheet? Same method as Edition III. Blunt regexes, public filings, replicable command.

The specimen, with its terms

6 August 1997, Macworld Boston. Gates on a screen above the stage, the audience booing. What actually changed hands that day is a matter of public record, and it is not what the anecdote says it is. Microsoft did not lend Apple $150 million.

That is a recapitalization. A lender wants the money back; an equity holder wants the company to exist. And the $150 million is the least important line in the table. It was not large against Apple's losses that year. The five-year Office commitment is what told the market Apple had a future, and the browser default is what Microsoft took for it.

Note what is not in the table. No votes. No board seat. A software company performed a financier's function — deciding that another company would continue to exist, and on what terms — and acquired no formal control in exchange. This is the point that makes the "Microsoft is secretly a bank" version both wrong and unnecessary. Nothing here requires a charter. It requires a balance sheet.

So we grepped for the instruments

Edition III's scanner already had a cohort and a pattern list. We added Apple — which, embarrassingly, had never been in it, despite holding the largest pile of securities in the group — and six new patterns aimed at the vocabulary of allocation rather than the vocabulary of selling: non-voting-stock, convertible-preferred, supplier-prepayment, marketable-securities, extend-credit, loans-receivable.

The bars below count only that financier block, across the four most recent 10-K / 10-Q filings for each of the nine companies. Live, from /data/edgar-prevalence.json.

Salesforce and Alphabet score high for the same boring reason Apple does: they hold a lot of securities and they say so repeatedly. Microsoft scores 4, which is not evidence that Microsoft does less of this. It is evidence that Microsoft's disclosure uses equity method investments — a phrase that lives in Edition III's pattern block, not this one. The regexes are blunt and they are supposed to be.

Apple is the end of the chain, and it incorporated one

In 1997 Apple is the company being kept alive. Follow it forward and it becomes the thing that does the keeping — and it is the one member of this cohort that stopped doing the financier function by implication and simply set up a company to do it.

This is the least metaphorical point in the whole sequence. A manufacturer did not become bank-like by analogy or by anyone's clever reading. It incorporated a securities manager and staffed it, in a state with no capital-gains tax, four hours from head office. Eight years after taking $150 million from a rival to stay solvent.

Now look back at the chart. Apple scores 62 — seventh out of nine. Alphabet scores 161. That is not Apple doing less of this; it is Apple's filings stating the position once and moving on, while Alphabet's restate it across every note. Counting words measures how much a company says, not how much it holds. Apple sits near the bottom of a chart measuring the exact behaviour it institutionalised. Keep that in view whenever you read a bar chart built out of regexes, including this one.

Three of the six patterns returned zero

marketable-securities fires in nine companies out of nine, 628 times. Not one of these firms is only an operating company; every one of them runs a portfolio. That is the finding, and it is unglamorous: at this scale the financier function is not an exception, it is a standard feature of the balance sheet.

Then the zeros. supplier-prepayment, extend-credit and loans-receivable return zero across the entire corpus. Not one hit in 36 filings. And we know that at least one of these things happens: Apple's Advanced Manufacturing Fund put $450 million into Corning and $390 million, then a further $410 million, into II-VI. That is capital advanced to a supplier so the supplier can build capacity. It is in the public record. It is not in these filings under any of these words.

Edition III learned this lesson from Amazon and we will state it the same way. When a regex returns zero on something you have prior reason to believe is happening, the regex is the answer, not the absence. But there is a second reading available here, and it is the more interesting one: the function is disclosed as a portfolio and never as lending. The money does the same work. The vocabulary that would make it recognizable as finance has gone missing.

Where the chain breaks

A recursion that collects only its successes is a story, not a finding. So: on 8 July 2015 Microsoft took a $7.6 billion impairment on the Nokia devices business it had bought for $7.2 billion fifteen months earlier, and cut 7,800 jobs. It was larger than the $6.2 billion already written off on aQuantive in 2012. Apple launched a lending product in March 2023 and withdrew from it in June 2024, thirteen months later, handing the function back to the firms that already had it.

Allocation power is not foresight. Holding the capital lets you decide who gets a next year; it does not tell you who should. Both of those write-downs were taken by companies that were, at the moment they took them, the most capable allocators in the industry.

Which is also the correction to the tidy version of this thesis. "Capital is continuous" is a useful structural reading and a bad literal one. Ownership changes. Governance rights change. The investors change, the decision-makers change, and the asset classes change. What persists across the handoffs is not a thing but a capacity — and the reason to map it in filings rather than narrate it in logos is that filings record instruments, and instruments are the only part of this that can be checked.

How to rebuild this

Same instrument as Edition III, one round further on, and the same deal: nothing to install, nothing to ask for. The SEC's submissions API needs no account and no key — it needs a User-Agent header that identifies you with an email. The sixteen regexes are published with their sources in the data file this page reads, and the financier block is the last six of them.

If you think extend-credit should have caught something and it did not, send me the pattern that catches it — a zero that turns into a number is the most useful correction this page can receive, and Apple's Advanced Manufacturing Fund says one is out there. Open for replication. Open for correction. Corrections to [email protected].