The Day I Ran My First 10K, and Rethought What 'Adding to a Position' Means
Notes on Gooaye EP693: what the buyers of MediaTek's $3.5B convertible bond are actually buying, why NVIDIA is fine with customers building their own chips, and whether you should add or trim after a stock has already run. Personal reflections for educational purposes, not investment advice.
Age may wither the body, but never the heart that once reached for the clouds.
—— Wang Bo, “Preface to the Pavilion of Prince Teng” (Tang dynasty, 675 CE; translation mine)
What This Episode Is About
The September 2nd episode of Gooaye doesn’t open with the market. It opens with the host running ten kilometres for the first time in his life. The plan had been to build up to it over two months. Instead, on an idle afternoon the day before training was supposed to start, he just went out and did the whole thing.
His description of the middle stretch is honest in a way I appreciated. Kilometres five through seven felt like falling into a personal low point — his cardio was fine, but his legs started burning and his brain kept manufacturing reasons to stop. Then, somewhere around kilometre eight, something flipped. It got easier. He stopped wanting to change songs or text friends, and just watched the clouds.
The back half returns to markets. MediaTek issued the largest overseas convertible bond in its history — $3.5 billion — and NVIDIA and Google took nearly all of it. He works through what kind of money that actually is, what NVIDIA is buying, and why so much of the recent news looks like everyone trying to take a bite out of Broadcom. Then he answers a listener question that never gets old: when you’re right and the stock keeps running, do you add, or do you trim?
Listening to it, the running and the market talk felt like the same thought seen from two sides: the ceiling you assume is usually one you drew yourself, not one the road drew for you.
The Main Points
1. Five to seven kilometres is the trough; after eight it gets easier. He’s clear that the hard part wasn’t fitness — it was his body sending stop signals while he knew he still had something left. Past that point there was a visible flip: the first half was spent looking for distractions, the second half was blank-minded and calm. He also credits losing close to ten kilos recently. The base was built earlier; it didn’t appear that afternoon.
2. Since August, doing nothing has been more comfortable than chasing momentum. His momentum-trading friends have been complaining all month, but step back and the index keeps grinding higher, closing in on July’s peak. When the market dropped thirty percent from the highs in July, he assumed the recovery would take until year-end. It didn’t. His summary: as long as you’re not constantly trying to trade momentum, this tape is fine.
3. That $3.5 billion carries a 0% coupon, and the buyers still paid a premium in the low teens. So they’re obviously not there for the yield. His read: this is essentially an option — if MediaTek’s XPU business works out and the stock rises, that stake converts. He also recalls Alchip landing an Amazon investment years ago; the market got excited, the stock gapped up and then faded. The headline was bigger than the effect.
4. He doesn’t think this is vendor financing. The test he uses is concrete: MediaTek is already profitable with a healthy balance sheet, unlike some compute-leasing outfits whose business arguably doesn’t function without that investment. The same action means different things depending on whose balance sheet it lands on.
5. NVIDIA is shifting from selling chips to selling the whole thing. His observation is that NVIDIA knows it can’t stop customers from building their own XPUs, so it changed where it stands. An AI factory needs several categories of chip; compute is only one of them. Interconnect, storage, networking racks — there’s still a role in all of it. Build your own accelerator if you like; everything around it can still be ours. He also flags NVHBM: moving the memory controller into the HBM base die on a proprietary spec rather than the industry standard. If that gains traction, it captures value in memory too. He’s explicit that this one needs more time before anyone can call it.
6. Everyone is trying to take a bite out of Broadcom, but being challenged isn’t the same as losing. He points to optical components as the parallel: a new narrative appeared, the related names got hammered, and yet revenue kept printing and many of them later made new highs. The market’s reflex when something new shows up is to sell first and rarely to consider that both things can coexist. Get used to the volatility, but don’t mistake the price reaction for a conclusion about the industry.
7. There’s no universal answer on adding, because it depends on which model you’re running. His analogy is blunt: if you meet two people, one you find attractive and one you don’t, is your behaviour identical? Stocks aren’t equal in your head either. What can be said is structural. If your core is value, rising prices shrink the remaining upside, so you trim — unless you genuinely still think it’s cheap. If your core is theme and momentum, a rising price is itself the reason to add. His own habit also varies by market: he caps positions in US names because earnings calls have blown him up overnight too many times, while in Taiwan he’ll let a single position run larger, as long as liquidity lets him get out.
Going Further
”The news was huge — why was it flat again the next day?”
Two giants stepping in, the largest convertible in the company’s history, and the result was one limit-up day followed by a return to flat.
The way I think about it: the question is which year’s cash flow this news actually changed. The buyers paid a premium for a zero-coupon instrument. What they bought is the right to convert if, several years out, this path works. Translated: even the people writing the cheque think the answer arrives years from now. There’s no reason for the price to book that today.
Three questions I now ask of any piece of good news: does this change orders, capability, or merely a relationship? Orders show up in the numbers fastest. Capability waits for a product cycle. Relationships often just move some future possibility from zero to not-zero. This convertible looks like the latter two. And the Alchip-Amazon episode is worth remembering precisely because it had the same shape — an enormous name attached, and no quarter’s revenue created.
I used to lose money right here, seeing the words “strategic partner” and deciding the story was settled. Strategic partnerships are usually the opening of a story, not the ending. The useful things to track are the ones the episode itself lists as open questions: who the customer is, when it goes into production, and how big the volume gets.
”Someone says this company is about to be displaced — should I get out?”
The optical-components parallel landed for me, because I fumbled my way through that exact episode.
The logic sounded airtight at the time: new approach arrives, old approach gets challenged, the related stocks should fall. And the market played that script for a while. What actually happened is that revenue kept printing quarter after quarter and the prices later made new highs. What got sold off wasn’t the fundamentals — it was the market’s imagination about the fundamentals.
So I gave myself a crude but workable rule: when you hear “X is about to be displaced,” first separate whether the challenge is to demand or to share. If the total market is expanding quickly, losing share and earning less are two different outcomes — that’s exactly the lens used here for whether rising XPU adoption hurts GPU vendors. Share gets taken, and the pie grows at the same time. If demand itself has stalled, then the fight over share really is zero-sum, and that’s when leaving makes sense.
Then there’s verification. Imagination can argue for a year, but revenue speaks for itself. So I write down, in advance, which quarter and which line item this bad news should show up in if it’s real, and I go check on that date. If I can write it, I can hold. If I can’t, I don’t actually know what I’m betting on — and that’s the reason to trim. Not because the price fell, but because I couldn’t state a test.
”It keeps going up. Do I add or do I trim?”
He doesn’t give an answer, and I think the absence is the answer.
Adding requires knowing which system is making you money in the first place. Value logic says price converges to worth, so higher prices mean less room and smaller size. Momentum logic says trends persist, so the rise is itself the reason to add. Both work. Mixing them doesn’t. Buy on a value thesis, add on a momentum impulse, then switch back to value to justify holding through a drawdown, and by the third switch no rule set is protecting that position at all. I’ve done this, more than once.
The second thing worth stealing: the reason for a position cap should come from market structure, not from mood. He caps US positions because overnight earnings gaps genuinely happen there. He’ll size up in Taiwan on the condition that liquidity lets him exit. Both constraints are checkable in advance. Compare that to “this one’s up a lot and I’m nervous” — that cap expires the moment your mood improves.
So I make myself write the cap as a sentence with a reason attached: this name maxes out at X percent, because it gaps on earnings, or because the book is too thin to exit in one go. A reason you can write down is discipline. One you can’t is usually just how you feel today.
Worth a Look
- Gooaye EP693 (2026-09-02), the starting point for this piece
- MediaTek’s public filings on the convertible and the follow-up earnings commentary, to check terms and stated use of proceeds
- Broadcom’s next earnings call — the episode itself says this will be the fullest read on where everyone actually stands
- NVIDIA’s published AI factory architecture material, for which layers it defines as its own territory
One Thing to Take With You
The single idea I want to keep: the ceiling you assume is usually your body talking, not the road.
What stayed with me wasn’t that he finished. It was his description of kilometres five to seven — cardio fine, body insisting on stopping. That signal wasn’t a fact. It was a default. Markets do this too: down thirty percent, my default said “this takes until year-end,” and that was also a default, not a calculation.
Here’s something I’ve tried, if you want it: pick one thing you’ve already given up on this week, and rewrite “I can’t do this” into a sentence you can check. Not “I can’t run 10K” but “last time I stopped at kilometre X, and it was my legs, not my lungs.” Not “I’m bad at English” but “last time I froze, it was in this specific kind of conversation.”
When I did this, about half of my reasons collapsed on the spot — not because I couldn’t do the thing, but because I had never actually measured it.
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.