The memory crisis could continue until 2030. Are SK hynix, Micron and Samsung a once-in-a-lifetime opportunity, or is it a TRAP?

The memory market has historically been one of the most cyclical segments of the entire semiconductor industry. Periods of high demand and rising prices were regularly followed by overproduction, falling memory prices and a dramatic drop in manufacturers' profitability.

This time, however, the question arises whether the structure of the entire market is beginning to change.

SK hynix CEO Kwak Noh-jung recently stated that he sees no clear signs of an impending downturn and that the current tight market situation could continue until the end of 2030. He even previously said that customer demand could exceed the company's production capacity even after 2030.

This is primarily driven by AI.

The development of data centres is dramatically increasing the consumption of traditional DRAM memory, but especially high-performance HBM memory. These are an essential component of modern AI accelerators and their production also consumes significantly more manufacturing capacity than traditional DRAM.

The result is a situation where manufacturers are shifting ever more capacity towards HBM, while the supply of traditional memory remains limited. In recent months we have seen significant price increases across the memory market, and it is not yet very clear where rapid new supply could come from.

But the business model itself is also changing

This, in my opinion, is one of the most interesting things about the current situation.

Samsung, SK hynix and Micron are trying to reduce their dependence on the traditional model, where a large part of memory was sold at current market prices and companies were extremely sensitive to every fluctuation in supply and demand.

Long-term contracts with customers are starting to play an increasingly important role.

Micron has announced 16 strategic customer agreements across data centres, the consumer segment and automotive. A typical agreement runs for approximately five years, i.e. from 2026 to the end of 2030. These agreements represent approximately 20% of the company's DRAM volume and a third of its NAND volume over that period. Moreover, key contracts involve take-or-pay commitments, so the customer undertakes to purchase a specified volume.

SK hynix is moving in a similar direction. In its second-quarter results, the company announced that it had completed long-term agreements with approximately 10 customers and continues to negotiate with other major players. The aim is precisely greater supply stability, better production planning and greater predictability of the entire business.

And Samsung is not lagging behind either.

According to Korean media, Samsung has entered into long-term agreements with five of its largest data-centre customers and is in talks with five other AI companies. According to these reports, Samsung has the ambition to gradually cover up to 60–70% of planned memory production with long-term contracts.

That, in my opinion, is an extremely important change!!!

Does it mean the end of cyclicality?

Definitely not.

The memory industry will, in my view, continue to be cyclical. Manufacturers invest enormous sums in new factories, and history has repeatedly shown how quickly a shortage can turn into an oversupply.

But personally, I sense that the degree of cyclicality could be lower than in the past.

The reason is a combination of several factors: long-term contracts, structurally growing demand for AI infrastructure, high barriers to entry, the technological complexity of HBM and a very concentrated market.

Interestingly, SK hynix's own CEO talks about this similarly. According to him, a potential future downturn might not resemble the sharp declines of past cycles, but could rather be a gradual slowdown or a period of stagnation.

Long-term contracts are, of course, no guarantee that memory makers' profits will never drop significantly again. But they can provide some protection against the worst parts of the cycle and, above all, significantly improve visibility of future revenues and production.

And this is where the valuation starts to get very interesting

When I did my own analysis of SK hynix, Micron and Samsung, I arrived at a forward P/E of roughly 4.5 to 5.5 using expected future earnings.

The PEG ratio appears even more extreme, coming out at around 0.1 for me.

At first glance, these are absurdly low values given how quickly these companies' earnings are growing today.

And this, in my view, is where the biggest question of the entire investment thesis lies.

Are these companies truly extremely cheap, or does the market simply expect that the current record earnings are not sustainable in the long term?

A forward P/E of around 5 can indeed look like a huge opportunity. But for cyclical companies, a low P/E often arrives precisely at the moment when their earnings are near the peak of the cycle. If memory prices subsequently fall, earnings can plummet and a seemingly cheap stock suddenly isn't so cheap.

On the other hand, if SK hynix is right and the current shortage truly lasts until the end of the decade, while long-term contracts simultaneously reduce the volatility of future results, the current valuation could indeed be very attractive.

And that is exactly why the whole sector seems fascinating to me today.

We have three companies that practically control the global DRAM market, benefit from one of the strongest technological trends of our time, and are gradually changing the way memory is sold.

At the same time, however, we are investing in an industry that has decades of brutal cycles, price wars and periods of overproduction behind it.

Therefore, I see the current situation in two ways:

Either we are looking at one of the most interesting valuations in the entire semiconductor sector, or at a perfectly set cyclical trap.

Personally, I lean towards the view that the structural changes in the industry are real and that long-term contracts can genuinely reduce some of the historical cyclicality. But I would definitely not count on cyclicality having completely disappeared.

And with a forward P/E of around 4.5–5.5 and a PEG of around 0.1, the answer to this question is absolutely crucial for investors.

How do you see it? Is SK hynix, Micron or Samsung a brutal investment opportunity at current valuations, or are we looking at the peak of the memory cycle and a TRAP?

$SNDK $SKHY $MU $005930.KS $NVDA $GOOG

Personally, I have tremendous faith in memory:)