Gooaye EP704 Notes: When $70B Became $50B, and Why It Didn't Move His Position

Contents

1. What the episode covers
2. 1. Where the $70 billion came from
3. 2. When the direction is right, the numbers are noise
4. 3. Your style came from the people around you
5. 4. Marvell’s $80 billion for 2031, and how to use it
6. 5. Two ways to die betting on a future roadmap
7. 6. A small revenue pool doesn’t mean it isn’t worth taking
8. 7. On impostor syndrome
9. Bad news lands — do I cut or wait?
10. Should I chase a theme that’s still a dream?
11. The One Thing to Take Away

A fish trap exists for the fish; once you have the fish, you forget the trap. A snare exists for the rabbit; once you have the rabbit, you forget the snare. Words exist for the meaning; once you have the meaning, you forget the words.

—— Zhuangzi, “External Things” (Warring States period)

In the 2026-10-10 episode of Gooaye (EP704), host Hsieh Meng-kung spent a segment on OpenAI’s annualized revenue: the $70 billion figure the press had been printing was revised by the Financial Times to $50 billion, markets flash-crashed that day, and the pre-market recovered it the next morning. He said the number changed nothing in his book, because the $70 billion came from taking one month’s revenue times twelve and then applying the 77% growth rate the company had given investors. In the same episode he walked through Marvell’s investor day: 2031 revenue guidance of $70–90 billion against a sell-side consensus he found at $45–50 billion. That kind of indifference has a precondition — you sized the position in advance, and a 50% paper drawdown still lets you sleep.

What the episode covers

The first half is personal: he attended the Manbao Pro annual conference, told the story of how he came to back MY’s work, and admitted he freezes up on stage and knows his ceiling is backstage. The middle turns to markets with two topics — the OpenAI revenue-definition flash crash, and Marvell’s investor day. The back half is listener mail: impostor syndrome, a 25-year-old’s drawdown after a five-bagger, an all-weather portfolio question, and running without headphones.

Compared with recent episodes there is less stock talk and more about how his own temperament has shifted, so what stayed with me is a messier set of things.

1. Where the $70 billion came from

He took the number apart. OpenAI and Anthropic recognize revenue differently; OpenAI signed early deals that make its revenue look worse, so some outlets chose to restate its run rate onto the same basis as its peer — roughly $40 billion. Multiply by the 77% growth the company had told investors, and you get $68 billion, which gets written up as “$70 billion.” The print came in at $50 billion.

His read: $50 billion is not bad. The company went from $2 billion to $50 billion in a span that is rare in human history. He also took a swipe at annualization itself — some people take one day’s return, multiply by 200 trading days, and announce an annualized return. The ARR math does the same thing; it gets accepted because nothing better is available.

2. When the direction is right, the numbers are noise

He described a now-retired sell-side veteran. He asked the man for the latest wafer-start numbers on a company; the answer was that he never looks at those, they are noise — if the direction is right, estimates get revised up and down, a crowd talks it up and then talks it down, and it still arrives at the target.

He says it took him a while to actually believe this, and the honest turn in the story is worth keeping: he admits he used to comfort himself with “the long-term story is fine” when the real reason was that he was underwater. The difference now shows in his SpaceX position, which he added to from the IPO right into the drawdown. Listeners were in his comments begging for the rocket to go up; he wasn’t bothered, because the position is 8% to 10% of his assets, and if it blows up, it blows up.

3. Your style came from the people around you

One line kept me thinking: whoever your friends are when you first reach the market, you will probably end up inside their circle. If they all day-trade, you won’t learn to hold through a swing. If they all run their maintenance margin into the floor, keeping yours at 180 or 200 makes you feel like an idiot, even though neither number is low.

The same mechanism decides what data you watch. If everyone around you tracks monthly revenue, your observation window shrinks from three months to one, and the shorter the window, the more often the tape moves your mood. His current practice is to stop checking intraday and look into what happened after the close.

A letter later in the episode lands on the same point: a 25-year-old made over NT$10 million on a five-bagger in the first half, got hit in July, sits 25% below the high, and now wants modest leverage on the index plus some megacap tech — asking whether he’s gone soft. The host answers with Livermore’s idea: you don’t meet bold old sailors. Either you’re a bold beginner, or you’re a relatively conservative old hand — and only the second kind survives the storms.

4. Marvell’s $80 billion for 2031, and how to use it

He says the revisions surprised even him. 2028 revenue went from roughly $18 billion guided in August to $20 billion. For 2031 the company handed out a $70–90 billion range, midpoint $80 billion, about ten times 2026, against a sell-side consensus he found at $45–50 billion, with long-run gross margin pegged at 59%. Inside that $80 billion, interconnect — optics, NPO, CPO — is close to half, custom compute and its attach content roughly 37%, with switching, storage and other businesses taking the rest.

Some timelines worth noting: 800G is the workhorse today, 1.6T is ramping, 3.2T still runs into physical limits. The company argues NPO and CPO will run in parallel, which cuts against the consensus view of NPO as a stepping stone to CPO. He had assumed 3.2T was a 2029–2030 story; a friend recently told him 2028.

The most useful part is what he does with the guidance. Marvell has no monopoly on data center connectivity, so if it can raise numbers like this, peers will likely raise too — which sends him looking at the things where expectations are still low, instead of chasing whatever story just finished being told.

5. Two ways to die betting on a future roadmap

Talking about the TFLN material in 1.6T, he said something I felt in my stomach: this still looks like a future theme to him, and people are already trading 3.2T. His warning has two layers.

First, roadmaps usually have more than one branch, and picking the wrong one blows you up outright. Second, and harder: even when you pick right, you can still eat a halving on the way. His example is the Taiwanese investing crowd and AOI — the direction turned out correct, but the ride took 50-60% out of them, and holding that was brutal. Dream themes like this need a sustained bull market under them; when they correct, the drawdowns stack.

6. A small revenue pool doesn’t mean it isn’t worth taking

This is my favorite segment, because he overturns himself in it. He once traded the BMC story — server management chips — through Nuvoton and lost money. His post-mortem blamed the size of the market: hand the incumbent’s entire revenue to Nuvoton and it still wouldn’t matter to that company.

Marvell’s investor day put a data-center management chip on the table, Aspeed’s stock fell that day, and his reading went the other way. Marvell wants compute, storage, switching and power integrated into one total solution, because that is how performance gets pushed; and that requires the chip that manages the whole data center, however small the market looks today. His line: revenue that is small now doesn’t mean it stays small — before 2020, when people discussed Nvidia, AI was a dream too.

He doesn’t say his original call was wrong. He says his post-mortem missed one possibility: someone takes the position early because the pool itself is going to grow, and everyone ends up with more.

7. On impostor syndrome

A listener, recently graduated into a startup, wrote that his boss gives him too much authority, he often has no direction, and with a weak job market the impostor feeling keeps growing. The answer came in two parts.

First: if you got into the room, the interviewer had already read your file and probably combed your social accounts, so you did something right. He argues that being genuinely unqualified for a position is something experience tells you directly rather than something you reason your way into — if you truly can’t hold it, you won’t be able to hold it.

Second, and more practical: you assume your boss has a direction, and often he doesn’t. He cites Jensen Huang describing moments when the company nearly went under while the team didn’t know, because the only people he could be fully honest with were the customers holding his fate. The fog you feel is what proximity to decisions feels like.

What I actually took from this segment was something else. All episode long he demonstrates one habit: remember what people did for you, and make sure they know. He says that strictly speaking he isn’t a good person — he has bad thoughts, he just chooses not to act on them; but when he sees someone living and working seriously, he makes sure that person knows they did something great.

Bad news lands — do I cut or wait?

I’ve gotten this wrong plenty of times: the headline hits, the group chat starts buzzing, and my finger is already on sell. This episode offers something more usable than “stay calm.”

The line is whether the news moved the price of the thing. He puts it bluntly: if you find that the quoted price genuinely can’t go up anymore, something has actually broken and you run. If it’s a question of methodology, sampling month, or which outlet restated which basis, the long story is intact. The OpenAI crash was the second kind, and the evidence is the pre-market recovery the next morning — the people who cut and bought back paid for both trips.

Push the test one step earlier and it becomes a buying checklist: what does this story rest on, does that thing have a price, and where do I look that price up? Without that step every headline looks equally dangerous, and all you have left to pick sides with is your mood.

Should I chase a theme that’s still a dream?

The other familiar spot: you understand a roadmap, you think the market hasn’t priced it, and you want in. Two questions come first.

One, does the road fork? 1.6T and 3.2T use different materials, and 3.2T’s physical limits aren’t solved, so nobody knows which approach wins. A bet concentrated on one branch looks generously priced because it genuinely has a decent chance of being worth zero.

Two, and this one is about you: if you pick right but the position halves first, can you hold it? AOI is exactly that story — right direction, people shaken out. The answer has nothing to do with how well you read the technology and everything to do with size. His own SpaceX position is the control case: same asset, 8% lets him ignore the comments, 30% has him refreshing headlines nightly.

So my habit now is to write down how much I’m willing to see go to zero on that branch before I look at whether it’s worth buying. Every time I did it in the other order, I was the one who got shaken out.

The One Thing to Take Away

The quality of your judgment depends on how much you care about the numbers, and how much you care is set by how big your position is.

The same headline is something you read and put down at 8%, and something you argue about at midnight and buy back the next morning at 30%. That retired veteran could call the numbers noise because his direction and his allocation were already settled, which earned him the right not to care. Indifference is the output of sizing.

Here’s something I’ve tried that works outside investing too. Pick one thing you’ve been checking over and over — whether someone read your message, whether an application came back, what a particular person thinks of you — and write one sentence: “if the answer is the worst one, here’s how I spend the following week.” If you can describe that week concretely, you can stop refreshing. If you can’t, the problem isn’t your composure; you’ve staked more on this than you can carry, and the weight is what needs adjusting.

One more, worth a try: he’s started running with no phone and no headphones, just a key, and calls it a kind of practice. This week, take one thing you normally do while scrolling, and do it once with the phone left at home.

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.