# The Feeling Is Awful, the Price Is Fine — Notes on Weak Internals from The Compound and Friends

Source: Realpha Blog (blog.getrealpha.com)
Original article and charts: https://blog.getrealpha.com/en/blog/compound-2026-09-29-bad-feeling-weak-internals-confidence-collapse-nvi/

> Notes on the September 29, 2026 WAYT episode of The Compound and Friends: consumer confidence at a 12-year low, the share of S&P 500 members above their 200-day average down from 73% to 51%, and the index still near its highs. I lay out the bullish case and the strongest counter-case side by side, and name the condition that decides it. Educational personal notes, not investment advice; no stock recommendations or price targets.

Published: 2026-09-30
Locale: en
Tags: market breadth, US equities, interest rates, AI capex, investor psychology


![A person at a high window at night, the far skyline lit up, the streets below mostly dark](/covers/compound-2026-09-29-bad-feeling-weak-internals-confidence-collapse-nvi-cover.png)

> Deprive the eye of a mirror, and it cannot set the brows straight; deprive a man of the Way, and he cannot know his own confusion.
>
> —— Han Fei, *Han Feizi・Guan Xing* (Warring States period; translation mine)

On the September 29, 2026 WAYT episode of *The Compound and Friends*, Josh Brown and Chart Kid Matt spent most of the show on one contradiction. US consumer confidence fell to 81.9, the lowest since 2014. The share of S&P 500 members trading above their 200-day moving average dropped from 73% to 51%. Of the index's 330-point advance since the end of July, Microsoft alone accounted for 181 points. And the index sits a point or two from its high. Their read: the selling is concentrated in rate-sensitive sectors, with only 1% of tech names at 52-week lows and zero financials — rotation, not evacuation. That conclusion rests on a condition: the composition of the new-low list has to stay the way it is, and it changes every day.

![Two bars on the left fall from 73% to 51% while the line on the right still hugs the dashed high, so internals weakened over the same stretch in which price did not.](/figures/internals-fell-price-did-not-en.svg)

## The bullish case: pull out what's actually being sold and look at it

Matt did something I want to copy. He didn't argue about whether breadth had deteriorated. He granted it, then asked who exactly is going down.

The split is clean. Utilities have 3% of members above the 50-day average, staples 27%, real estate 0% — and 45% of utilities are printing 52-week lows. Flip to tech: 66% above the 50-day, 73% above the 200-day, 1% at new lows. Financials: zero. Josh's explanation lands. People bought utilities, staples and real estate for the yield, and the 10-year Treasury now pays 5.25%. Take 3% from a ketchup company, or 5% from the federal government? Unless you think that stock also appreciates, the answer writes itself. What's being sold is the bond substitute. That has nothing to do with the economy.

![Four bars rise from zero to 66%, with real estate and utilities almost flat on the floor and technology highest, showing the selling is concentrated in the dividend-paying sectors.](/figures/which-sectors-are-falling-en.svg)

On why yields keep climbing, Matt counted 18 of the last 24 trading days closing higher on the 10-year. Since 1965 that has happened 11 times, and 80% of those times yields were lower twelve months later, by an average of 90 basis points. He also ruled out the worst explanation: breakeven inflation rates have stayed anchored, so this is a real-yield move rather than an inflation scare. Nominal minus breakeven puts the real 10-year near 2.85%, the highest since 2008.

Then came the table I sat with longest. Matt bucketed history by the 10-year yield and took the median forward P/E for the S&P 500 and the Nasdaq 100 in each bucket. Above 7%, the S&P sits at 13.6x and the Nasdaq at 12.6x — the lowest of any bucket. And the best multiples don't come from 0–3%; they come from the 3–7% middle. Josh's reading: zero rates let capital wander, and an alternative imposes discipline, so investors pay up for companies that have to do the math. The line he wanted to leave people with was that 5% isn't the breaking point. Anyone who's been at this thirty years doesn't need to hear it; anyone who started after 2020 has only seen the 2021–22 hiking panic and treats 5% as the end of the world.

![Three bars form an arch that is high in the middle and low at both ends, with the rightmost above-7% bucket lowest at 13.6 times, showing the best valuations sit in the middle band.](/figures/yield-buckets-valuation-arc-en.svg)

The Nvidia segment runs on the same logic. The stock closed near 228, knocking on the 230–235 level, with rising 50- and 200-day averages and RSI at 58. What Matt looks for is divergence — price making a high while momentum doesn't — and there is none here. Two weeks after Dario Amodei publicly called for slowing down, Jensen Huang announced a $150 billion buyback; added to the $85 billion already authorized, that's $235 billion of repurchases through Q1 2028, roughly 4.3% of shares retired if the price stays flat. Split-adjusted, this stock was about $10 in early 2023. One more chart: over the past year, S&P 500 forward earnings grew 36.7% while the index rose 17.9% — earnings growing at more than double the price — and forward estimates have historically landed within 1% of actuals.

## The other side: every sample in that argument grew in the same weather

The show brought its own opposition. Josh read from Barclays: when cumulative NYSE breadth deteriorates by 5,300 over 21 trading days, every instance since 2004 came with an average 9.6% drop in the S&P over the same window. This time the index didn't move. The only precedent for that combination — broken breadth, flat price — is October 2014, and the S&P fell more than 5% the following week. Another chart showed the ratio of new highs to new lows back at the level of the March panic over the Iran war, a state historically associated with a 670 basis point decline and an average peak-to-trough drawdown of 5.2% over the next 21 days. Matt said that if you showed him only that red line, he'd guess the market was in a 20% drawdown.

The strongest objection, though, isn't in those charts. It's in something Matt produced later. He said finance has almost no iron laws, and then gave one that comes close: the starting yield on a 10-year Treasury very nearly determines your annualized return over the following decade, and today's dot sits near 6%. If that holds, the bullish line — every breadth washout of the last decade resolved upward — has a problem, because that entire decade of samples grew in zero rates and easing. Back then there was no 6% risk-free alternative sitting next to you, so missing the rebound cost you everything. Now there is one. Inside a single episode, a relationship called an iron law is quietly removing the floor under a rule of thumb.

![The dots follow a line sloping up to the right, with the past decade's samples clustered inside a grey box in the lower left and today's dot far out in the upper right.](/figures/starting-yield-sets-the-decade-en.svg)

Inigo Fraser Jenkins' *The Everything Trade* from Alliance Bernstein pushes it further. His argument is that AI's appetite for capital has grown large enough to tie equities, credit, the dollar and foreign flows into one trade you can't diversify away from. Of the funding for the 2026–2028 capex build, roughly $1.5 trillion — about 45% — comes from equity capital; the rest comes from private credit, investment grade, ABS, CMBS, high yield and loans. Big tech's free cash flow has gone to zero and the spending is now debt-funded, which floods the bond market, so a plain vanilla bond fund quietly adds AI exposure. Meanwhile corporate buybacks were the largest buyer of US stocks for twenty years, and AI is eating that cash — Nvidia can still do both; most companies can't. The note says explicitly that it isn't bearish and that calling a top is pointless. Its question is whether your portfolio survives if the freight train stalls.

![A single bar shows equity at 45% on the left and various forms of debt at just over 55% on the right, with an arrow running from the debt side to an ordinary-looking bond fund.](/figures/ai-capex-funding-mix-en.svg)

There's one more signal, delivered as a joke. Oura postponed its IPO. Josh said he has nothing against the product but has never seen anyone make money buying a health-gadget IPO: GoPro went public in 2013, revenue peaked at $1.6 billion in 2015 and fell to $651 million last year, and in September it announced a merger with an optical photonics company, shareholders receiving $1.14 in cash plus a sliver of equity. He remembers the day Fitbit listed — a giant treadmill erected on Broad Street outside the exchange with fifty people running on it wearing Fitbits. That stock is down 95%. Risk appetite at the fringe is already contracting, and it's contracting near the index highs.

As for the bullish evidence that credit default swaps on the hyperscalers haven't blown out: that reads to me as a coincident measure. It breaks when things break. It won't tell you early.

## The one thing to take away

One idea survived the episode for me: **when something feels wrong, find out who is going down before you decide whether to trust the feeling.**

What the show got right was refusing to stop at the aggregate. They took "breadth is bad" and split it into eleven sectors, one cell at a time, and the same number came out meaning something else entirely. Forty-five percent of utilities at new lows and 45% of financials at new lows are two different worlds. My own most expensive mistakes over the past few years weren't about direction. They happened at the aggregate: my account turned red, I felt uneasy, and I started collecting reasons to justify the unease instead of sitting down to count — what do the red ones have in common? Is this one cause, or eleven?

The habit travels well beyond a brokerage account. Something I've tried is moving it onto feelings. Next time you feel generally bad about a person, a job, or a situation, take a sheet of paper and write down every specific thing from the past month that bothered you, one per line. Next to each, write where it came from: this person, your own exhaustion, some third party, or plain bad sleep. Get to ten, and you'll often find the "general bad feeling" is held up by two or three events from the same source — and that one of those can be dealt with today. The other eight usually have nothing to do with whoever you thought this was about.

![On the left one solid grey block of overall bad feeling; on the right it is split into four source buckets holding ten small squares, one of which is marked as something you can handle today.](/figures/sort-the-bad-feeling-en.svg)

While you're writing, don't summarize and don't judge. Two columns only: the event and its source. The judgment can wait until you've finished counting.