# Gooaye EP708: A Car Is a Depreciating Asset, and So Are You Source: Realpha Blog (blog.getrealpha.com) Original article and charts: https://blog.getrealpha.com/en/blog/gooaye-2026-10-24-ep708/ > Notes on Gooaye EP708 (24 October 2026): the Ken Griffin interview the host relays, which parks GPUs next to the airline business; a friend's question about whether a gross margin at a record high can go any higher; why positioning only sets the depth of a fall; and the depreciation talk from a hotel room during a typhoon holiday. Educational listening notes only — no investment advice, no price targets, and the judgements here may be wrong. Published: 2026-10-25 Locale: en Tags: Gooaye, podcast-notes, GPU economics, shortage pricing, investing discipline ![A night flight cabin, a long row of empty seats running deep toward the tail, a single reading lamp lit by the window](/covers/gooaye-2026-10-24-ep708-cover.png) > I would buy osmanthus and carry wine along — yet it would never be that journey of my younger years. > > —— Liu Guo, "Tang Duo Ling: Reed Leaves Fill the Sandbar" (Southern Song, c. 1204; my own translation) In Gooaye EP708, released 24 October 2026, host Hsieh Meng-kung spends half the episode on a Ken Griffin interview he relays. The Citadel founder's line is that for two years you could buy the whole supply chain behind the data-centre capex story and make money, and that the market has now started doing research company by company. The sharpest stretch parks GPUs next to an airline: once a flight clears breakeven, selling one more seat at twenty dollars still earns, and once capacity runs long, fares go into free fall. Hsieh adds the other half himself — the script runs eventually, even if not this month — and the compute rental prices he was watching at the time were still rising and holding high, so that price is what the argument has to be tested against. Later in the same episode he says something else: a car is a depreciating asset, and so are you. ## He says the tape has started to smell wrong The episode was recorded in a hotel room during a typhoon holiday. Four days off arrived with no warning, he knew that keeping two boys at home would end in damage, so he booked a hotel at ten that night and drove onto a jammed motorway. At the front desk he asked whether everyone had had the same idea and fled; the clerk said cancellations outnumbered bookings. He heard later from a friend that the three-day holiday had become four, so people already heading south left early — that was the traffic. Same phenomenon, and the reason you can see is usually not the reason. Then he talks about the tape. The week kept building that uneasy feeling, and when the feeling stacks to its ceiling a bounce is usually close by — he guesses it is coming and says he has no plan to trade it. He sets July beside it: in July he trimmed nothing and wore the drawdown, on the logic that May and June had earned enough and a continued leg would earn more, and most of what fell in July never walked back up. This time he cut early, and he will be careful buying back, because a price on the way down anchors you to where it has been, and that level sometimes already borrowed three years of valuation, which takes a long time to repair. ![The price line falls away from its high while a dashed horizontal line stays at the old peak, marking a level that already borrowed three years of future, yet the eye stays glued to that line.](/figures/gooaye-ep708-anchored-to-the-old-peak-en.svg) The backdrop is that money got expensive. Issuance costs have risen, trading capital will not go into anything that pays back slowly, and the industrial side and the investing side are pulling in together. He quotes Buffett's line that interest works as a kind of gravity, dragging things back down; with no large rate cut nearby, some of those highs carry no guarantee of returning. ## The question his friend asked: can you open it any wider? A friend who does good research was talking to him about one group of stocks — he does not name it — and asked why people still buy when the gross margin is already printing like that. The question sounds odd at first; good things get bought. What the friend added was this: with your margin opened that wide, can you open it any wider? If it cannot go up, the remaining paths are flat or down, and there is nothing left to trade. ![The gross margin curve climbs to a high level and flattens, then splits into three roads ahead: a dashed one rising with a question mark, a grey dashed one staying flat, and a grey dashed one heading down.](/figures/gooaye-ep708-margin-ceiling-three-roads-en.svg) Hsieh says what it broke was an assumption he had been sitting inside: it is good, so I can keep buying. The rally already explained why it earns that margin today, and to pull in fresh money to push it higher, it has to get better. He holds the same blade against the shortage-pricing story. The scarcest things in the market are memory and substrates, and if those two cannot even hold up at these levels — no need to rally, just hold — then anything less scarce has even less room above it. His split: a name selling nothing but shortage pricing is dangerous from here, a name with volume growth and a step up in value is still selectable. ![Two panels side by side: on the left gross margin spikes up and falls back to its original level, while on the right gross margin steps up and stays there.](/figures/gooaye-ep708-shortage-spike-vs-value-step-en.svg) ## An interview that parks GPUs next to an airline The Griffin interview comes to us through him, and he says he is unsure where the later part came from — possibly somebody else's write-up — so the details need discounting. Griffin's starting point is that the market has begun doing research. For two years all you had to do was follow the data-centre story — where the halls go, what they cost, which layer is the bottleneck, which layer carries the fat margin — and buying the lot made money; Hsieh says his own hands were full of supply-chain names. Griffin says DRAM and MLCC still have supply bottlenecks and rising margins, and that what the market is now chewing on is whether this holds two or three years out. Then comes the analogy. GPUs and airlines line up on three points. High fixed cost, low marginal cost: the plane is bought and the flight departs regardless, so one more passenger costs close to nothing, and a finished data centre running an extra hour or two is an electricity question. Inventory expires: the seats empty on this flight have no time machine to go back and sell into, and the hours a GPU sits idle are gone the same way. Commoditised product: the same GPU rented from different operators differs in service, while the compute itself is near-standard, like a seat on a flight. ![The unit price curve is gentle while it sits above the breakeven line, then falls straight down past the vertical dashed line marking overcapacity until it nearly touches the marginal cost line at the bottom.](/figures/gooaye-ep708-breakeven-then-free-fall-en.svg) So it cuts both ways. Past breakeven, one more seat at twenty dollars earns; below breakeven the flight still departs, everyone wants to lose a little less, and the price falls freely. Griffin's second example is the telecom build of 2000: nobody laid that fibre expecting it to sit unused, more went in than demand needed, and some of it was never lit. The host asks him straight whether GPUs are walking the same road, and his answer is a question mark — once installed capacity runs past demand, marginal pricing goes into free fall. ![Two pairs of bars stand side by side, each with capacity built rising well above demand; the excess in the left pair is labelled never used, while the demand bar in the right pair is dashed with a question mark.](/figures/gooaye-ep708-fiber-built-more-than-used-en.svg) That cut lands on a few assumptions in circulation. A few episodes back Hsieh covered Musk saying on a call that they sign short contracts rather than long ones because prices are still rising; that story faces a challenge. There is also useful life: cards from a generation or two back still rent at decent prices, but can the capitalisation rate and revenue model in years seven, eight and nine carry the last twelve months' assumptions forward? ## He builds a rebuttal on Jensen Huang's behalf Hsieh calls himself a committed bull, so he goes looking for bearish arguments to test himself with. After this stretch he drafted a rebuttal for Jensen Huang: the Jevons paradox. In plain terms, when something gets cheaper the volume jumps, and the extra volume can hold total spend up. That is where GPUs part ways with airlines — an airline cannot manufacture more passengers, so falling fares just mean losses; when compute gets cheaper, uses like agentic workflows eat the capacity. He says Huang would argue exactly that, and the rebuttal holds water. ![A tall narrow block of price times volume forks after the price falls: the upper path becomes shorter and wider with the same area, while the lower path only becomes shorter and loses area.](/figures/gooaye-ep708-jevons-fork-area-en.svg) He keeps both sides on the table: a rebuttal does not make the Citadel founder wrong, and the fixed-cost and commoditisation points stand. He leaves himself one indicator to track — compute rental prices — and when he looked, those prices were still rising and holding high. Then he adds the layer that puts a chill in it: share prices lead revenue by at least six months, sometimes a year or two. Why are prices that smart? His explanation skips prophecy — on the way up, good news stops moving the thing, so it plateaus, and what is left after that is bad news. So the big decline does not wait for quoted prices to turn down; it goes first. ![The share price line rises first, runs along a plateau and then falls, while the revenue line is still climbing, and a bracket marks the lag between the two turning points.](/figures/gooaye-ep708-price-leads-revenue-en.svg) ## He crosses out the question "who is selling?" One more stretch comes from the same relay. Plenty of people explain this decline as hedge funds deleveraging, and Griffin says there is fact in that — unwinding positions does move markets — but you should not say it fell because hedge funds unwound. Something that was going to fall falls without them; the unwind only speeds that leg up. Hsieh says this sent him back to his own habits. Margin debt works the same way: it was going to fall anyway, and the leverage stacked on top makes it fall deeper, deep enough to trigger forced selling. He admits that when he cannot find a reason he reaches for positioning too, cursing whichever fund blew up. Yet think back to the start of a bull run, when some big holder was selling and some fund was cutting — everybody just laughed at the guy who sold before the bus left. When selling does take it down for good, the problem sits somewhere else. He closes the stretch on a judgement of his own: the question ahead is whether you can hold on to it, and how hard you earned in the past barely counts as a question. A supply-chain mania tends to end back where it started or worse, so stops and profit-taking both have to work. ## The line he misquoted in a hotel room In the listener questions that fill the second half, one of them connects the market discussion to somewhere I did not expect. A listener describes a kind of accounting he is stuck in: two hundred dollars at Starbucks feels expensive, while a 5% drop in his portfolio in a day registers as nothing; he wants a Tesla but hates losing thirty percent the moment it leaves the lot; he wants to go to Europe but figures sleeping at home saves a few hundred thousand that becomes real money in ten years of stocks. He asks whether Hsieh has lived in that head, how to get out, and whether it just comes down to having enough capital. Hsieh says this grows on anyone who stays in the capital markets long enough. The tech friends drive the good car and book the good restaurant on a date, then cry poor to their friends all week, because every dollar they have only wants to buy more stock. He says having enough capital never dissolves it — at some level you still think this way — though there are moments when the grip loosens. His loosened after his children were born: his life had other people in it, and some memories are worth buying. His example is the trip to the US west coast with his elder son before the second boy arrived, including a beach hotel that shows up in GTA, not cheap, and he is still glad about it. Then comes the line. People say a car is a depreciating asset; he says, brother, you are a depreciating asset too. Dropping into a sports car at thirty-something and climbing into one at sixty-something are different atmospheres. He reaches for a line of classical verse and gets it slightly wrong from memory at the microphone — I would buy osmanthus and carry wine along, yet it would never be that journey of my younger years. His point: buying the flowers and taking the wine out on the water comes with a different heart than the swagger of being young. ![A curve slides downward with age, and the same amount of money buys a tall bar in your thirties but only a short stub in your sixties.](/figures/gooaye-ep708-you-depreciate-too-en.svg) He closes on money he burned himself: he came back to an old game as a whale, spent a fair amount, does not play any more and cannot be bothered to resell the gear, and he is glad he went back to flatten the bosses he once lost to with the strongest kit available. The money can be earned again, the experience stays for life, and experience accumulated earlier gives you longer to think about it — compounding inside your head, as he puts it. ## My stock's margin is at a record high, so why isn't it moving? I have walked into this one: the report is clean, the gross margin prints a record, the analysts are complimentary, and the price sits there and then starts drifting down. The tool in his friend's question is swapping "is it good" for "can it get better". Good decides what it is worth; better decides whether new money has a reason to arrive. With the margin already at a historical high, somebody has to be able to name the next notch up — a new product, a new customer, a structural change from a new spec — and when nobody can, neither flat nor down pays you. The second tool splits price increases into two kinds. Shortage pricing: demand lifts the price, margins jump with it, and that high rides the cycle back down when the cycle turns. A step up in value: the product moved a generation, unit price settled on a new shelf, the customer cannot swap it out — that margin builds a step, and a turn in the cycle does not necessarily sweep it away. Which of the two you own decides how to feel about a drawdown. ## Was this drop caused by institutions cutting? Being stuck in a loss makes you want a name to blame, and the positioning story conveniently supplies one. Griffin's point transfers directly: positioning sets the depth, fundamentals set the direction. Deleveraging, forced liquidations and algos trading into each other make the fall deeper than it should have been and leave that long lower wick at the bottom; something genuinely still in an uptrend does not lose the path because one holder cut. So on a drawdown my first question is this: if nobody were a forced seller today, would this price stand? ![Two paths descend from the same point along the same fundamentals slope, and one of them pokes out a long lower wick along the way, yet both end at the same point.](/figures/gooaye-ep708-chips-set-depth-not-direction-en.svg) Asking it that way pushes attention back into the fundamentals box — can the margin still open wider, is the quoted price still high, is the scarcest layer holding? If it stands, the flush is an opportunity; if it does not, the flush only makes the fall slower. What Hsieh repeats in this episode is to do more research while the tape is bad. What I took from it is that research has a specific job here: replacing "who is selling" with "why should it be worth this". ## Worth a look - Gooaye EP708 (24 October 2026), the source for these notes - Ken Griffin's public interviews and Citadel's public comments; the episode relays them, the host is unsure it was one sitting, so check the original audio or video - The Jevons paradox in its original context, from William Stanley Jevons, The Coal Question (1865) - Published rates from the major clouds and GPU rental platforms, to track whether compute rental prices have turned - Earnings calls and monthly revenue from memory and substrate makers, to check how far the shortage-pricing story has run - Liu Guo's "Tang Duo Ling: Reed Leaves Fill the Sandbar", the source of the opening quote ## The one thing to take with you **You are a depreciating asset too, and some experiences are only for sale at a particular age.** An asset's depreciation gets written into the accounts; nobody keeps the ledger on yours. The same money spent on the same thing returns something different at thirty than at sixty, because the feeling needs the body and the frame of mind you had at the time. That is what stings in his line — time runs one way, and past a certain age some purchases can no longer buy back the thing they once bought. Here's a version I've tried that works away from money too. Take something you keep saying you'll do "once there's money", write it down, then find out what the smallest version that keeps the core would cost — small enough that this month's budget covers it. Once you have the price, book it today.