SK Hynix: The final leg of the memory cycle
In my previous post, I highlighted how Micron could reach 800$ (unfortunately for me I was too stupid and sold at 440 after the Q2 earnings, it was actually the right move [stock dropped like 30% in the next 12 days] but I didnt buy back). Recently I have bought into SK Hynix after the pullback from 3 million KRW, I started buying at about 1.8 million KRW, unfortunately once again due to my low IQ I sized too big and didn't manage to keep buying all the way to the bottom (though I did pick up a lot at about 1.4 million). Position is roughly at breakeven at these prices. In my last post I mentioned that the longer you hold these the more risky they become, after studying the 1990's DRAM cycle (shout out Gavin Baker). I have come to a realization that this cycle is in fact more similar to the 1990 cycle than the other consumer led cycles. LTA's will also change the game, the only reason I don't think we have topped yet in the memory cycle is because no one believes in the LTA's therefore the prices are still trading as though its cyclical and not structural and that's the opportunity. I do want to bear in mind that their gross margins are already at record highs. But there is 3 distinct differences in this cycle. Those that don't understand this cycle is definitely different might clown on me for saying this time is different, but the facts HAVE ALREADY proven this time is different (look at gross margins and LTA's if it wasn't different would these things be present??)
- Sales are mostly to businesses
- Long term agreements (LTA's)
- Demand still outpacing supply as far as we can project
I could tell you all about AI and how it needs more compute but I think the numbers and the stock prices speak for themselves. There is clearly a lot of demand and I hope if you are reading this your understand the memory story already this post is mostly to explain my price targets, write out my thinking and also break down why I think certain things will happen. I have a picture that addresses #3 above. As for #1 I think we all know about the hyperscaler Capex numbers so no need to harp on it (it is constantly getting revised upwards though so might we see 1.2 Trillion$ in Capex next year??[I have questions as to how they will fund it but some of the cloud figures and OpenAI & Anthropic numbers seem to suggest they are actually generating revenues]) I will focus on #2 for this post as this is where I feel the market is not accurately pricing it in. The market and commentators seem to suggest that LTA's are not actual contracts and can be thrown in the garbage bin by hyperscalers at anytime, if that was the case why would they put up deposits?? There are ways in which market dynamics can play out that lead them to break LTA's (spot prices drop more than the deposits that they have placed is one of them), but I think for now since we can see HBM demand still outstripping demand and most memory companies saying they can only produce half of demanded supply the skew is to the upside. But I do hope that the next phase of the cycle will be from capacity expansion and not higher average selling prices (ASP's) as higher ASP's might cause demand destruction. My base case is that memory companies are more than happy at these prices and will ramp up production to meet demand, so even if ASP's dont continue to go up it will be highly profitable as long as ASP's dont DROP.
So wheres the opportunity??
I will first break down what the market is currently pricing SK Hynix/ memory players although I will focus on SK Hynix since I believe they are the cheapest and have the biggest moat. Will exclude Sandisk & other HDD players since I'm not super familiar/ haven't taken the time to study them properly [I have tried but I came to the conclusion their businesses are only booming because of lack of capacity from the big 3, therefore if big 3 ramp up capacity they will be the first to go] also excluding CXMT cause they are producing mostly DRAM for consumer devices and the stock seems to have gone nuts]
Positioning: Most of Korean retail has been washed out at this point, and positioning in SK Hynix is incredibly light
SK Hynix: Forward P/E 4-5x
Micron: Forward P/E 10-12x
Samsung: Forward P/E 5-7x
EOY 2027 Numbers:
Lets start off with the bear case, SK Hynix is expected to do 200,000 KRW at the low end in earnings this year, which means its 8 times trailing P/E at the low end and it doesn't grow, if this was actually what I believed, I would be avoiding the stock 8x P/E for a stable stock would be a nice return if it was durable, but not worth the risk if its cyclical
Base case: 250,000 KRW growing to 320,000 KRW next year (this is how we get 5x Forward P/E). Not gonna elaborate too much cause this is what the market believes right now and if we are right but the market already agrees with us there will be no outsized performance
Bull Case: KB Securities estimates that SK Hynix will do 389 trillion KRW (285 Billion USD), lets use a band of 250 billion to 300 billion USD (will use USD cause its easier for me to calculate). In this case if the markets believe in LTA's it would stand to reason it would easily assign 10x P/E on it even if they didnt expect any growth. That equals to 2.5 trillion to 3 trillion in marketcap. (As I'm writing this SK Hynix trades at 825 Billion marketcap)
So I guess my bullish estimate is 2.5 trillion to 3 trillion $ marketcap this would be 3-3.6x gains? If there are no share buybacks this equals to 5 million to 6 million KRW sounds a bit ridiculous even to myself as I'm typing this but I think stranger things have happened and I wont be carrying the full position all the way till the top will probably start trimming above 4 trillion KRW (KRW might also go down vs USD so this might be artificially high)
What needs to happen for the market to give SK Hynix 10x P/E??
- The Korean companies must show they actually care about their shareholders, if that happens SK Hynix is a double just off multiple expansion (Micron premium is due to better perceived shareholder friendliness in the US)
- They have to hit earnings the bullish earnings expectations
- The market must believe in LTA's (for this to happen either DRAM, HBM and NAND prices stabilize or the companies honor the LTA's for the next 1-2 years)
Anyone that see's this post please feel free to remind me to sell if it ever gets to 10x peak cycle P/E (might be a giga bull and say it will go to 20x P/E), I still think SK might perform market returns after that but if I wanted to get market returns I can just buy an index fund, be diversified and take much less risk.
Thank you all to anyone that read the full post.