# Gooaye EP681: The Numbers Are Right, the Price Might Not Follow
> The core tension in the latest Gooaye episode: nearly every AI supply-chain fundamental is on track, some even ahead of plan, yet prices are hard to call. The host's answer is to script multiple scenarios and hide in cheap, defensive names. My notes, plus an honest boundary: a mention is not a buy signal.
Published: 2026-07-24
Locale: en
Tags: gooaye, podcast-notes, valuation, ai-supply-chain, risk-management, taiwan-stocks
> *The whole world is drunk; I alone am sober.*
> *The whole world is muddy; I alone am clear.*
> —— Qu Yuan, "The Fisherman" (Warring States period); translation mine
## What this episode is about
Gooaye EP681 (2026-07-22). One contradiction runs through the whole hour: **the AI supply chain's fundamentals are almost all on track, some ahead of plan, and yet the prices are genuinely hard to call.** Vera Rubin is ramping, AMD is about to launch Venice, CPUs are in short supply, passive-component monthly revenue is posting healthy year-on-year growth. Plenty of names that "should" go up. Many of them have already broken down on the chart.
The host doesn't try to predict his way out of this. He scripts scenarios instead: how to play a range-bound tape, how to chase a breakout, with the target names for each written down in advance, waiting to see which one the market god picks. Woven through it is a stretch I found more valuable than the price talk. When the market pulls back, don't let investing hijack your life, and don't repeat the old lesson of sizing your leverage too big.
**Original episode**: search "股癌 Gooaye" EP681 on any podcast platform (about an hour). Listen to the source. This is only my after-listening summary.
## The main points
- **This sell-off is "index flat, small caps destroyed"**: the Taiwan index is still up near 40,000, but the OTC board and small US tech names are down 30 to 50 percent. The host frames it as a dress rehearsal. The bigger drop is still ahead, and anyone who can't sit through this small stress test should worry more about the next one.
- **The "go full-time" indicator fired again**: every time the market tops, people show up asking about leverage and whether to trade full-time. Someone asked this time too, and blew up right after. It's a classic retail sentiment contra-signal, more reliable than any oscillator.
- **Generational inflation in leverage**: the host used to think his 2.5x made him a daredevil. To today's new traders, "2.5x is kindergarten." Everyone runs huge. The margin calls this round are that stress test coming due.
- **A relief bounce, but hold the celebration**: on the way back up you hit trapped supply at every moving average. Most former leaders are broken. The test is simple. A name that bounces, shakes out, and can push again is still strong. One that can't bounce and rolls back over has fallen out of the pack.
- **Smart money's consensus: hide in cheap, earning, defensive names**: old-AI and ODM forward P/E has fallen to around 10x. The high-multiple stuff gets hit hardest, down 30 to 50 percent, so money ducks into things that "only fall 15 to 20 percent even if they drop again," and parks in TSMC and the index as a shelter.
- **The earnings cards are about to turn over**: Vera Rubin ramping, AMD's Advancing AI event (Helios racks, a big Microsoft order, a possible Anthropic partnership reveal), and the formal Venice launch (the SP7 CPU). That revenue will be good is close to certain. Whether the stocks move is not.
- **CPU is the underrated bottleneck this cycle**: agentic AI is pushing the CPU-to-GPU ratio from 1:8 toward 1:4, 1:2, and some now argue 1:1. Volume is exploding. CPUs are stuck on wafer capacity and memory, so revenue gets recognized slowly. Sockets, a 1:1 bound component, gain pin count each generation, and ASP steps up a notch every time.
- **Google's Frozen v2 chip**: the market is guessing it's Marvell's TPU inference accelerator (TIA). The host's read is clean. The point isn't Google, whose in-house chip is a rounding error on revenue. The point is who builds it. But this is old news in institutional circles, so whether it can catalyze anything is a question mark.
- **Kimi K3 punctures the "low compute" story**: last episode the host said "compute is fine" would get slapped down fast. A day or two later Moonshot announced it was reserving compute for existing customers because there isn't enough. The idea that cheaper models mean fewer servers broke on its own.
## A few pearls in the Q&A
- **How to actually use forward P/E**: the host says valuation multiples are "worth a glance, but not that important." Two things matter more. First, do you buy into the story at all. Second, is there something you know that no report has written down yet. That last one is the real alpha, the gap that pays you when everyone else catches up and revises late. As for the sell-side's "sky-high" price targets, he treats them as headline marketing.
- **How to read a serial capital-reducer like Yageo**: his answer is grown-up. The nice thing about public markets is you can choose not to play. Don't buy what you don't understand. If you do understand that a company likes to manage its stock and stir things up, some people board on purpose. Better to know what you're buying than to curse "retail slaughter" afterward.
- **Discipline around averaging down and stops**: he admits he's grown less fond of hard stops in recent years, because he trusts his read. But he's emphatic that **beginners need a hard fail-safe**, like a floppy disk that won't go in backwards, to stop themselves from being dragged around by the tape. Once you have experience, you switch to position-size limits instead.
- **800V racks and CDU as a theme**: asked about Navitas, Wolfspeed, STM. His suggestion is to bundle the group and watch. Wait for the day they move together (5, 8, 10 percent on the day), then think about stepping in. There are too many themes right now and money is pulling back, so pure sit-and-wait could last forever.
## Extending the thought
**One: "good earnings" and "goes up" are two separate ledgers.** The most important line in this episode is really the valuation anchor in plain speech. Fundamentals set the downside, the level where buyers show up once it gets cheap. Money and chart shape set the timing of the upside, who lifts it and when. Blur the two and you burn time stuck in the "the numbers are great, why won't it move" mood. Good earnings don't arrive late, but they can skip a whole quarter.
**Two: the alpha in forward P/E hides in the line nobody else wrote.** What the host describes is inference ahead of consensus. When every report gives 40x, watch out for being the last mouse in. But when you hold one piece of the puzzle that every report missed, that's the money you can actually take. The multiple isn't the edge. The information gap is.
**Three: the real decision in a pullback is opportunity cost, not "run or stay."** Hiding in cheap defensive names is a trade: you swap upside meat for downside safety. Whether that trade is worth it depends on whether you think the market has one more stab lower in it. There's no clean answer, but you should at least know you're making that swap, instead of chasing highs and dumping lows on autopilot.
## How we apply it
Three points here plug straight into what we're building. Writing them down as working notes.
1. **The valuation anchor only works on the cheap end, and this episode adds a footnote**: old AI became smart money's shelter only after it fell to 10x, while the expensive stuff got sold no matter how good the earnings. Our anchor rule already restricts entries to the cheap end. This is a live market check on it.
2. **Multi-scenario planning beats a single forecast**: the host doesn't bet direction. He preps both a range-bound script and a breakout script with target names for each. That's the same logic we use with scenario probabilities instead of one price target. Accept that you can't call the turn, then build a position that isn't a disaster under any script.
3. **Position limits, not hard stops**: he shifted from hard stops to size-based risk control, on the condition that his read is strong enough. That's a reminder for us. Your risk tools should match your circle of competence. Beginners want the hard fail-safe, veterans want position management, and using the wrong one hurts.
## An honest boundary
To be clear: **a podcast naming a stock is not a buy signal.** We backtested the "mentioned by a well-known podcast" factor ourselves. The average 20-day excess return after a mention was about +1%, with a hit rate under 50 percent. No tradeable alpha. Gooaye's value is narrative intelligence. It's a high-quality proxy for what mainstream Taiwan retail is feeling, so you know what the crowd is thinking, not what to buy. Every stock in this episode is context from the show, not a recommendation.
## Worth a look
- Original episode: *Gooaye* EP681, on any podcast platform (Spotify / Apple Podcasts / SoundOn)
- Related post: Gooaye EP679: This Wasn't Fundamentals, It Was Leverage Getting Called (the first half of the same pullback)
- To understand "good earnings that don't rally," search two terms: valuation anchor and information asymmetry
---
*This post is after-listening notes and educational discussion, not investment advice. Any stocks mentioned are context relayed from the show, not buy or sell recommendations. Do your own research before holding anything. Quoted content is the property of the original show; please listen to the source.*