MacroMicro EP209: The Same Capex Was Vision Last Quarter and a Crime This Quarter

He ate a peach, found it sweet, and gave the ruler the half he had not finished.
The ruler said: “How he loves me — he forgot his own appetite to feed me.”
Later, when his beauty faded and the ruler’s affection cooled and he fell out of favour,
the ruler said: “This man once … fed me a peach he had already bitten.”
Mizi’s conduct had not changed from the beginning. What made him admirable before and criminal after was the turning of love into loathing.
—— Han Feizi, “The Difficulties of Persuasion” (Warring States period; my translation)

These are my personal notes on MacroMicro’s After Meeting EP209 (released 2026-08-02, “Read Earnings Through Cash, Read the Fed Through Warsh”). Not a transcript, not official show content. Please support the original programme for the full discussion.

What the episode is about

The show runs on two tracks: the Federal Reserve, and the big-tech earnings released so far.

Put the two together and the same thing surfaces on both — the market has changed which line item it prices, and it changed far faster than the fundamentals did.

The passage from Han Feizi is about Mizi Xia. He hands his ruler a peach he has already bitten into, and in the moment it is read as devotion. Years later, out of favour, the identical act is produced as evidence against him. The behaviour never changed. The affection did.

The platform companies got the same treatment this quarter.

The main points

1. The earnings were good — almost counter-intuitively so. Microsoft’s cloud revenue grew 43% year over year, with Microsoft 365 group subscriptions holding double-digit growth at 14%. Meta’s advertising revenue grew 27%, with ad impressions and pricing up 14% and 12% respectively. Google’s advertising grew 15% on the back of Search and YouTube, while Google Cloud grew 82%. Microsoft and Google each added another 50 billion dollars of backlog versus the prior quarter. AI monetisation has not stalled; it is speeding up.

2. The share price reaction went entirely the other way. For the previous two quarters the market rewarded how much you dared to spend. This quarter it rewards how much cash you still hold. The sharpest illustration is Apple, marked down in earlier quarters for moving slowly on AI, now favoured precisely because it did not commit heavily — and back at the top by market capitalisation. Among the rest, Microsoft, which retained more cash, traded better after reporting than Meta or Google.

3. The pressure point is free cash flow. Operating cash flow remains strong, but capital spending has risen enough that Google and Tesla both turned free cash flow negative in the second quarter, with Meta close to zero. Google raised close to 50 billion dollars in equity financing in June. What the market has started to worry about is not an absence of AI demand — it is the debt being layered on to fund the capex, and the domino risk if demand ever slips.

4. Microsoft’s “cut” was accounting, not a decision. This is the detail I most wanted to write down. Microsoft revised this year’s capex from 190 billion dollars down to 175 billion — which sounds like braking, until you learn the reason: it extended the depreciation life of data centres and office buildings by ten years, pushing the cost out to later periods. The actual investment plan is unchanged, and next year’s spend is still going up. For scale, the latest sell-side estimates put big-tech capex near 800 billion dollars this year, with a shot at a trillion next year.

5. The Fed held, but the margin for error is narrowing. Rates unchanged, asset purchases unchanged, statement nearly unchanged — placid on the surface. But three voters thought July already warranted a hike. The real story is what happens now that Warsh has discarded forward guidance: the market prices itself. The ten-year yield went from under 4.4% in late June to above 4.6%, and that self-administered tightening did part of the Fed’s work for it while buying more time to watch the data.

6. The number to track is the monthly core inflation print. All three components of June’s core CPI pointed to inflation staying contained: core goods still declining month over month as tariff effects fade, shelter decelerating, and services excluding shelter flipping outright negative — ending the first half’s run of consecutive increases. The show cites Williams’ reference threshold: if core monthly prints hold near 0.2% in the second half (roughly 2.4% annualised), the Fed does not need to hike; sustained readings above 0.2% raise the odds materially. The next meeting is September, and the last report available before it is August’s.

7. The membership of the five reform working groups tells you the direction. All three leads on the communications group — former Bank of England governor King, former New York Fed SOMA manager Fisher, and former Brazilian central bank governor Fraga — are critical or sceptical of forward guidance, yet all three value transparency. So the destination is not a return to deliberate obscurity; it is communicating the logic of decisions without communicating the predicted levels. The balance sheet group is more genuinely split, with no consensus pushing for balance sheet runoff, which makes aggressive tightening there less likely than previously assumed.

8. Cheaper open models are not the thing that pops the bubble. The show points to an Artificial Analysis chart plotting API cost against capability tier: since ChatGPT arrived in late 2022, the cost of a given level of intelligence has fallen in steps, across both open and closed models. That is the ordinary path of a diffusing technology — costs fall, usage rises, which is simply Jevons’ paradox. Global AI penetration sits somewhere around 15% to 20%, so there is a long way left to run.

Extensions

Separate “the fundamentals changed” from “the priced line item changed”

This is the most useful measuring stick the episode gave me.

When a company’s cloud revenue is growing 82% and backlog is up another 50 billion, and the stock falls anyway, the market has usually not decided the business is worse. It has switched which line item it prices — from how large you might become, to how much cash you still have.

Those two situations demand opposite responses. If the fundamentals changed, you re-examine your reason for holding. If the priced line item changed, you check whether your holding period can survive the re-rating. Conflating them tends to produce the worst decision at the worst moment.

Telling them apart is not hard: look at the numbers that are not downstream of the share price. Revenue growth, backlog, paying users. When those roll over, that is a change in fundamentals.

The market is currently rewarding the signal, not the fact

The Microsoft 190-to-175 revision is the most interesting thing I took from this episode.

What actually happened: depreciation life extended ten years, this year’s reported figure shrinks, the investment plan loses nothing. And the market’s response was relatively positive.

In other words, what is being rewarded right now is the appearance of restraint, not the fact of spending less.

The use of this observation is not to sneer at the market. It is a reminder that when a crowd starts rewarding signals over facts, it has entered an anxious state — and anxious pricing is distorted in both directions. So rather than guessing when it corrects, put attention back on the numbers that cannot be changed by an accounting treatment.

There is a practical reading habit in here too: when a capex figure moves, ask first whether the investment decision changed or the amortisation method did.

Negative free cash flow comes in two kinds

“Google and Tesla turned FCF negative” sounds alarming, but it can arise from two very different situations:

• The core business stopped earning — operating cash flow deteriorated
• The core business earns fine, and the company chose to spend — capex rose

The episode is explicit that this is the second kind, with operating cash flow still at strong levels.

The distinction matters because it determines what you track. In the first case you track demand. In the second you track financing conditions — rates, credit spreads, whether money can still be raised at a tolerable cost. Which loops straight back to the Fed track: if we actually reach a hike, the damage to a sector funding capex through leverage is amplified.

The two threads meet there. That is the part of this episode I thought was well built.

What retail investors should change now that forward guidance is gone

“The market prices itself” sounds abstract, but it has a concrete consequence: from here, what a Fed official says carries less information and what the data shows carries more.

The dot plot and forward guidance used to save you a lot of homework. That shortcut is gone. What replaces it is a threshold like 0.2% that you can track yourself, monthly.

For a retail investor this is mostly good news. A threshold is public, checkable, and requires no interpretation of anyone’s tone. The bad news is that volatility rises, and structurally so — not as a transitional phase.

My own adjustment is to move attention from “what will this meeting do” to “have the last few monthly prints cleared the bar consecutively.” The first has to be guessed afresh every time. The second accumulates.

A self-check about judging the same thing two ways

Back to Mizi Xia.

The sharp part of that story is not that the ruler was fickle. It is that when affection changes, people go back and re-narrate facts that were already settled. The peach is the same peach. It has simply been described again.

Applied to investing, the check I set myself is this: when I notice I have changed my language about the same set of numbers, I need to establish whether what changed was my model or my mood. If a company’s capital spending is “investing in the future” while the stock rises and “burning cash” while it falls, and I acquired no new information in between, then I did not change my judgement. I changed my feelings.

Simple check. Not easy to run honestly.

One honest boundary

The above is my synthesis and extension, not advice to anyone. The figures cited (growth rates, capex amounts, yield levels, cash flow positions) are as presented by the show’s research team; I quote them to illustrate a line of reasoning and have not independently verified each one. Companies are named to illustrate the phenomenon, not as recommendations or assessments.

To be clear about my position: I hold assets affected by this move, so I have a natural preference for the conclusion that this is a re-rating rather than a deterioration in fundamentals. I am writing it down so a later version of me knows where I was standing.

References

• MacroMicro After Meeting EP209 (released 2026-08-02, “Read Earnings Through Cash, Read the Fed Through Warsh”) — the source of inspiration for this piece. Please listen to the original and support the creators.
• Han Feizi, “The Difficulties of Persuasion” — origin of the half-eaten peach and “the turning of love into loathing”

---

Disclaimer: This article contains personal listening notes and study reflections for educational purposes. It does not constitute investment advice, an offer, or a solicitation. No specific securities are recommended and no price targets are given. Investing involves risk; past performance does not indicate future results. Please make independent decisions based on your own financial circumstances and risk tolerance, and consult a qualified professional where appropriate. The author may hold positions in the types of assets discussed.

Raw Markdown

This article is an educational discussion of investment method. It is not advice to buy or sell any individual security, offers no target prices, and does not analyze any current holding. Investing carries risk; make your own decisions or consult a qualified professional.