Gooaye EP701: After a 30% Drawdown, He Rotated the Whole Book

Contents

1. Down 30% — what made him stop holding on?
2. Why did so many of Taiwan’s small IC design names move at once?
3. Does AI-assisted chip design actually help small companies?
4. Can the TSMC-in-Texas rumour be used for anything?
5. Treasury yields are still high. Should I stay out?
6. So how do I use any of this?
7. Worth a look
8. The one thing to take with you

Hard is the road, hard is the road — so many forks; where am I now? A long wind will come to break the waves; I’ll hang my cloud-white sail and cross the sea.

—— Li Bai, “Hard Is the Road” No. 1 (Tang, c. 8th century; my own translation)

Gooaye EP701, released on 30 September 2026, closes out the third quarter. Host Hsieh Meng-kung says July’s selloff took his Taiwan book down about 30% from peak to trough, with leverage running between 1.1x and 1.3x at the time. He expected to spend the rest of the year digging out, yet his Taiwan book made a new high in early September; his US book lagged because one position was sized large and blew up. He credits the recovery to rotating in August and September into whatever was showing strength — optical communications, small IC design houses, satellites — and he adds, in his own words, that some of it was luck. The reasoning comes from his own book and conversations with industry contacts, with no public data to check it line by line.

Down 30% — what made him stop holding on?

His brake wasn’t a percentage. It was structure failing to hold where it should have.

The timeline in the episode is clear: through June he thought it was the easiest market in the world to trade; mid-July brought what he calls a merciless beating; by late July he felt a loss of conviction, because the bigger moving averages weren’t braking at the places they should have braked. That’s when he cut — and the positions he cut were still profitable ones. He had planned to sit and wait, then found he couldn’t carry it.

That part made me pause. Cutting a winner is harder than cutting a loser, because the screen keeps telling you that you were right. His decision ran off a structural signal, not off his P&L. As for why the hole closed so fast, his own answer was that he doesn’t know — maybe he flipped to buying weakness and momentum kept running. He left luck inside the answer.

Why did so many of Taiwan’s small IC design names move at once?

The shape he describes is order spillover. Design capacity and design headcount have become the scarce resource, so the big houses point whatever wafer allocation they can secure at AI-related work, and push the rest downstream.

His example is MediaTek taking large TPU orders, which means some projects simply can’t be taken in-house; those land with other Taiwanese design houses, and keep rolling down to what he calls the smallest tier in the chain. So the extra work in those small companies came from congestion upstream — a different thing from those companies suddenly acquiring new capability.

What I like about this chain is that it can be falsified. If AI wafer demand cools, the spillover disappears first, and the small names lose orders before the big ones do. The thing to watch is whether upstream is still congested, not whether the small-cap charts keep going up.

Does AI-assisted chip design actually help small companies?

This is his second inference, and he flags the uncertainty himself.

The starting point is OpenAI’s in-house chip, publicly described as going from architecture freeze to tape-out in eight or nine months; he puts it closer to a year, and says the point is how fast the whole flow moved rather than physical speed. The official story credits AI collaboration with cutting headcount and time in labour-intensive stages like layout. From there he reasons: if that holds, design effort shrinks, and a small company working on the same two-to-three-year timeline could ship three products where it used to ship one — bigger operating leverage than its size would historically allow.

Then he names the weakness in his own chain: there’s a good chance Broadcom’s design capability is simply that strong, and the official line is just an official line. He describes his own method as shooting the arrow first and painting the target afterwards — join in when something moves, buy first and research after, then go ask industry friends to check the answer.

Checking the answer turned up something else. Design houses won’t let client secrets touch cloud models, so they run open-weight models in their own server rooms; he says there’s a chance frontier closed models get deployed the same way, into customer data centres, letting companies use the leading edge locally. He stamps that one himself: too early to call, treat it as a listen rather than a conclusion. He also notes a precedent — when he said Google would sell its own servers, people online called it nonsense, and a few months later it happened.

Can the TSMC-in-Texas rumour be used for anything?

He treats it as rumour commentary, but gives it one use: watch whether the related names move first.

His reasoning for the location is infrastructure — water, power, and a large chemical cluster tied to oil production, which is why materials suppliers are also looking there, not only TSMC. Reports mention six front-end fabs, while what he had heard pointed to back-end first with front-end under evaluation; the whole thing sits at a very early stage. His observation rule: if the direction is real, the related names run before the news confirms, and the harder they run, the more it signals that the decision is already made underneath. He also gives the other side — the fab-services names are famously cheap, he believes the future is good, and they still don’t move; he traded a few swings in Q1 and Q2 and found it hard work.

There’s a piece of honesty here I enjoyed. He points out that it isn’t only the US government asking for local supply — customers ask too, with Musk’s non-China, non-Taiwan stance the most visible case — and he had just put down a deposit on a Roadster 2. His line: I’m not dating the man, I just need to know what game everyone is playing. The deposit is about fifty thousand dollars, he guesses the car might land near NT$10 million and take years, and he books it the way he books a new phone — a cost of participation that might hand him an idea.

Treasury yields are still high. Should I stay out?

He thinks it’s worth worrying about, and he still puts macro second, for a specific reason: rates have sat high for a while, macro commentary will always tell you this is dangerous, and if that’s all you watch, you end up unable to act, because there’s always something new being thrown at you to scare you.

His alternative is to stay on individual companies and individual themes: where the money is, and which names are about to post revenue and profit. His image for it is a stock walking from 100 to 200 — macro headlines put bumps in the road, and the job is to hold that stretch of road. His comment on the fear-and-greed index sits on the same line: the reading flips around all day, so trading off it leaves you unable to hold anything.

He also gives FOMO a definition you can observe: more and more things you can’t buy today and will only cost more tomorrow, several days running, until everyone believes it and piles in. He feels that mood drifting back, wants a few more sessions before calling it, and is deliberately slowing down himself after three good quarters.

So how do I use any of this?

Most readers are sitting in one of two spots: burned in July and still out, or watching September strength and afraid of being the last one in. The episode touches both.

On the first, his line is blunt: someone who got scared may never touch it again, and when they finally do come back, be careful — the person who spent months saying stay out suddenly flipping can mark the next place people get hurt. The use of that for me isn’t predicting other people. It’s the mirror. My own fear and greed tend to arrive in sync with the crowd’s, and the moment they’re in sync is the moment risk is highest.

For the second, I split it into two separate questions. Is the reason this theme works still intact? And is the reason I bought it still intact? Those can have different answers. Upstream congestion continuing keeps the spillover story alive — but if my reason for buying was that it was going up, a healthy theme won’t save me, because I never had a test to check against.

On his shoot-first method, I’ve tried a tamer version: put in an amount small enough that it can’t move my mood, book it as a research expense, and commit to writing a note that checks the answer by a set date — no note, position closed. His version trades size for speed; mine trades a small amount of money for the motivation to do the work. It may not be right, but it does tie “I bought first” to “did I ever actually understand it”.

Worth a look

• Gooaye EP701 (30 September 2026), the source for these notes
• Earnings decks and filings from companies such as MediaTek and TSMC, to check claims about orders and capacity
• US Commerce Department releases and company press releases, to follow up rumours like the Texas site
• The US Treasury yield curve published daily, so you can read the rate level yourself rather than through someone’s summary

The one thing to take with you

Write down why you’re still holding, in a sentence that can be proven wrong.

His decision to cut holds up because he had a switch made of something other people could see too: the big moving averages failing to brake where they should have. “I think that’s about enough” is a different animal — no switch, so it never fires.

Here’s the version I’ve tried, and it works away from money too. Pick something you’re currently persisting with — a relationship, a skill you’ve been at for six months with nothing to show, a job you’re carrying — and write one sentence: “If X happens, I stop.” X has to be a fact someone else could verify, like “three months running I practise less than twice a week” or “they break the same promise twice”. When I can’t fill in X, it usually means my reason for persisting was never clear in the first place, and that’s the thing worth sorting out.

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.