Break the Cycle
Bad memories haunt investors. The Telecom bust. The Great Financial Crisis. The Dutch tulip mania back in 1634. Investors remember these disasters and try to hold on to some thread of caution. Bad memories keep investors' expectations in check. Market cycles happen anyway.
The computer memory market cycle is the same. But more.
Bad Memories
High fixed costs paid for by borrowing and parts interchangeable by design help memory act like a commodity market. Wild. Always too much or too little. Never the right amount. A price chart for memory could be one for copper, or coffee, or corn. The fact that you can chart price means the thing is pretty commoditized! Memory companies over-built when times were good. Prices crashed. Memory companies lost money. So did investors a lot of the time.
Investors don't have to think too far back to recall memory cycles. Three companies—Micron, SK Hynix and Samsung—make all of the most advanced memory. Two of them lost money on the memory chips they sold as recently as 2023, and Samsung only made money by selling other stuff. That’s for the high-end stuff. The market for not-quite state-of-the-art memory is still struggling.
A bit longer ago—just over a decade—memory stopped advancing technologically and economically at Moore's Law rates. It stopped being a technology—at least one that attracted technology investors. "Technology investment" over the past 20 years meant software investment. But, then, in the fall of 2022, OpenAI released ChatGPT. NVIDIA hardware powered it. NVIDIA sales rose 800%. NVDA stock has been electric. Up 1000%.
NVIDIA makes processors, not memory. But its solders memory to those processors. And that memory is less like the commodity stuff with each new model and each new release.
New Memories
Just like AI needs much more processing power compared to traditional tasks, it needs much more and better memory. Just like GPUs are better suited to these tasks than traditional logic processors, High-Bandwidth Memory—HBM—is better suited to AI than traditional memory solutions.
SK Hynix pioneered HBM back in 2013. It’s still the leader today. Samsung and Micron have, so far, struggled to keep up. SK Hynix just blew the top off earnings. It tripled profit compared to last year. Sales grew 75%. All while its traditional memory market struggles.
AI has changed the way companies buy memory too. SK Hynix sold its HBM capacity for 2025 already! It knows, more-or-less, what it'll make on that. It has contracts and agreements for future orders. Customers aren't buying this stuff from a warehouse any more. It hopes to pin down its 2026 HBM capacity in the next few months. That would lock-in a good chunk of its profit for next year. HBM has graduated from the commodity cycle.
Those profits? Look like they'll be substantial. SK Hynix earned as much in the past six months as it did in the previous 2 1/2 years. Customers can't get enough HBM right now. The outfit expects HBM sales to double this year. In the past year, HBM made up 40% of its sales. HBM pushed SK Hynix into an era of greater stability and improved profitability.
Don’t Forget
AI hasn't rinsed cycles from the memory market. It changed what's important. HBM demand and AI proliferation have pushed the traditional memory cycle to the side.
Investors clutch the old paradigm of boom and bust. Shares of SK Hynix don't reflect the shift. They expect a swift return to the wild, profitless old days of 24 months ago.
The business is growing into quality. It’s working with customers under forward purchase agreements. Its designing custom solutions—with purchase agreements. Its developing systems solutions that require much more coordination and cooperation in the computing ecosystem.
Shares don’t expect any of this to result in much. So, we can invest in a bustling business, the market leader in an industry of three shifting from a highly cyclical one to a more stable one featuring long-term customer commitments for a bargain price.
Disclaimer: None of this is investment advice. It's meant to illustrate ways LCM thinks about investing. Things that LCM decides are good investments for LCM and its clients are based on many criteria, not all of which are covered here. Some or all of LCM's ideas may not be suitable for other investors. LCM does not recommend investing either long or short any position mentioned. LCM may own positions in some of the companies mentioned. Some of its ideas will lose money — investing entails risk. See full disclaimer here.