Gigachad Korea. 7+ years of investing. Financial mesmerizer, unknown chess player Follow for some Korean exposure
Joined August 2019
- Tweets4.3K
- Following304
- Followers203
- Likes3.9K
Pinned Tweet
The rally after the July selloff happened right after Amazon and Microsoft reported earnings, although there was also Leopold's liquidation. Anyway, it's time to rethink this setup, because their earnings are now just two weeks away.
I think there's a very high probability that we'll see another rally this time as well, but the outcome will play out differently. Here's why.
First, as I mentioned before, going forward, AI trades won't all be supported at the same time. The narrative will shift to winners and losers, because that's simply the natural outcome.
Also, unlike before, now that concerns about neoclouds have emerged, there will have to be some kind of adjustment to earnings estimates across the entire AI basket.
And eventually, with winners and losers splitting apart and liquidity inevitably flowing out of the broader trade, this rally will result in lower highs and lower lows. In other words, I expect any rally to be sold, with the divergence depending on which companies turn out to be winners and which turn out to be losers.
Even this much is already expected.
What we need to worry about is what happens after Anthropic's IPO and after the midterms. As I mentioned, you've probably noticed that all three event windows are overlapping right now.
That's why I'm not recommending bottom-fishing in this AI trade or trading it broadly in general. It's just too reckless.
But if you really want to catch the bottom, waiting for Amazon and Microsoft's earnings is probably your best bet.
What we need to see is the exact opposite move... We need a major outflow.
Leveraged Long ETFs just saw an inflow of $9.5 Billion last week, the largest in history, most of which was semiconductors 🚨 Risk On 🤑
I don't know if it's just me, but over the past few weeks, it feels like the market has gone up every single time the futures open on Monday.
All the escalating crises over the weekend just disappear on Monday. So today, it's going up again, by more than 1% no less.
But as expected, the structure is still a mess, and it looks like it'll be sold off soon.
As I've consistently emphasized, they love the classic RSI 30.
$QQQ
And surprisingly, the Nasdaq is plunging toward RSI 30 again. I told you something this stupid could happen... These guys love the classic so much. Anyway, it's going back to RSI 30.
$QQQ
KOSPI closed up 2% today, and the foreign selling streak has finally paused for now. They're now net buyers for two consecutive days.
But in terms of dollar amounts, they net sold $18 billion over the past two weeks, while their purchases over the last two trading days amounted to only $800 million, so it seems best to approach this cautiously for now.
It's just that, as you know, the selloff was so sharp and deep that this feels like an area where a technical rebound could happen. And I expect any rebound to be sold.
I think there's a visible gap 5,900 and 6,300 right now, so I can see the possibility of a rebound reaching that area. But I expect it to be sold off easily.
KOSPI is bouncing 3% today, but you're probably not going to want to hear this, because the actual trading flows are absolutely terrible. Yeah, I'll tell you: foreign investors are selling again today. So, how bad is it? They've been net sellers on 13 of the past 14 trading days. Are these guys trying to send KOSPI to zero or something? lol.
But the bigger problem is that other participants are selling too. Retail investors are selling, and institutions are selling as well. And this is something we've seen several times over the past two months during KOSPI's ghost-town period.
The reason every participant can be selling at the same time is that SK Hynix's buyback is still ongoing. Literally, the buyback is the only thing participating on the buy side.
After that 3% rally in overnight futures, nobody is buying anymore, and they're selling into every rebound.
And this is the phenomenon we've been watching and talking about continuously for the past two months.
Many people are still confusing two separate things: declining demand and the possibility that future data centers won't get built as planned.
Problems with future data centers mean that future demand projections will be affected. It doesn't mean there's no demand today or that demand is going to decline. The demand projections were based on the data centers that were supposed to be built in the future in the first place, and that's what's now being called into question.
Nobody is claiming that current demand doesn't exist or that demand is about to decline. The issue right now is that there's a growing consensus that the data centers planned for the future are highly unlikely to be completed at 100%.
And that long-tail impact on future demand is affecting valuations right now.
Ultimately, the problem is that data centers are highly sophisticated, massive structures, yet the plans and projections built on the assumption that you can simply bring the funding and land and make data-maxing happen are now falling apart.
This isn't a game. Even if you have the funding and land to build 1,000 data centers, obviously you can't just build them all according to plan. And those stupid execution plans are what's affecting the present.
Let me explain this again. The truth behind the demand everyone was talking about is simply something like the utilization rates of data centers that will be built in the future.
When people see that it looks like a lot of data centers will be built in the future, and everyone is pulling in money through debt to fund them, they call that insane demand and say orders are sold out for years.
Now, when you consider that not all those data centers will actually get built, all that changes is the future equipment demand and those situations where people claimed orders were sold out for years.
Put simply, taking into account that there are other GPU providers as well, if you assume that only 800 out of the 1,000 data centers planned for the future actually get built, you could end up with calculations like that.
Morgan Stanley has revised AMD's GPU demand forecast for 2026-2030 from 14.1 million units to 8.0 million units, or -43.3%, while keeping Nvidia's the same.
$AMD $NVDA
Readers added context they thought people might want to know
There is no record of Morgan Stanley issuing this report on Oct 14, 2026, and no major financial media has reported these figures. The attached chart is not in Morgan Stanley's official format and contains a basic math error (14.1M minus 8.0M is a 6.1M decrease, not 5.9M).
marketbeat.com/stocks/NASDAQ/…
And as I mentioned, it's now extremely difficult for individual to predict the actual extent of delays, cancellations, and reductions among the data centers that will eventually be built in the future. That's why I said this interconnected AI bottleneck trade has entered a very difficult phase, and why it's hard to recommend it.
It's simply because those future orders that were supposedly sold out can effectively drop to zero depending on the utilization rates of future data centers. And again, predicting those numbers is extremely difficult.
I've talked a lot about Solidigm for a while now, how it's going to IPO and is weighing on SK Hynix, with an IPO valuation of $150 billion. But I think when this IPO news comes out, SK Hynix will initially decline. However, I'm starting to get the feeling that this could be the bottom of the current decline. It's just intuition.
In other words, there won't be a bottom in SK Hynix before that news comes out. And that's fine. The IPO announcement will probably come within the next two months.
And once again, everything is lined up across this timeline: Anthropic's IPO, the midterms, further escalation with Iran, and the CoreWeave situation. I don't think that's a coincidence. If a bottom forms, I think that'll be it.
SK Hynix is now 17 weeks into its decline from the peak, down 60%, and it made a new low this week.
By the way, it's now been just over a month since the US 10-year Treasury yield crossed 5%, and the longer this period drags on, the more everything will start to deteriorate.
And that time has come again lol. According to the data, Koreans naturally bought $SOXL, the 3x leveraged semiconductor ETF, this week! A huge amount. They net bought $2.6 billion this week. That means Koreans have now been net buyers for two consecutive weeks after five straight weeks of net selling.
As I mentioned earlier, the signal flipped two weeks ago, and the market has been falling ever since. Must be a huge coincidence lmao.
Oh shit.. the signal has finally changed. Koreans are buying $SOXL again, the 3x leveraged semiconductor ETF. Koreans net bought $350 million worth of SOXL this week, marking the first shift to net buying after five straight weeks of selling.
As you know, we started properly tracking this on a weekly basis eight weeks ago, and I combined everything into a single quote, so you can verify it all by following that quote. Simply put, two weeks ago, you could already see that Koreans had been selling $SOXL for five consecutive weeks. And you could also see that the signal had flipped last week.