Hello, this is Pazz.
Over the past few days I have had the chance to travel around several regions (overseas) and look at the memory market.
And the memory shortage is far more serious than the situation I had only been reading about in the news in front of my PC.
It is no longer a matter of price; memory simply cannot be obtained, so many small and mid-sized manufacturers are facing the threat of insolvency, and it seems even large PC OEMs and smartphone makers cannot build products because there is no memory.
The problem is that the memory shortage will be even more extreme next year than this year.
Next year's supply is expected to grow only about 20-30% compared with this year, while demand is multiplying every year, so there is no answer.
Most of the supply increase is also allocated to servers, and there is simply no allocation to legacy memory areas such as smartphones, PCs, home appliances and set-top boxes, so existing manufacturers seem to have entered survival territory.
Channel market prices (spot price) are already approaching nearly twice the contract price, and even so the goods cannot be obtained; and even if they are obtained the unit cost does not add up, so producing devices is meaningless. Nobody would buy a cheap set-top box that used to cost tens of thousands of won if it were sold for hundreds of thousands. And who would buy a budget smartphone that used to sell for 100,000-200,000 won if it had to be priced above 500,000 won? They would buy a budget model from Samsung or Apple instead.
The problem is that the price of the low-capacity DRAM and NAND that went into budget smartphones has soared 10-20 times compared with last year, so if the memory alone costs 500,000 won... it makes more sense to stop production and wind down the business. As an alternative, some are producing small volumes using recycled memory and the like, but if this situation continues I think most small and mid-sized manufacturers will eventually halt production and go insolvent. I have even heard that large manufacturers cannot get memory and are hunting for it on the market. Just now there is news that Apple cannot hold its September new-product announcement because it is short of memory. You can imagine how desperate the world's foremost device maker must be to have reached out even to China's CXMT.
The situation is truly far more serious than reading about it in words in the news. Because of the severe memory shortage, the electronics ecosystem looks set to face enormous change.
As those who have seen my earlier posts and my book will know, I have consistently talked about the memory cycle, and taking the leading nature of share prices into account I had been saying the share price peak would come in the second half of this year or the first half of next year; but looking at the recent market situation my thinking has changed a great deal. In my view this memory cycle cannot end quickly and will likely last at least about 5 years.
My outlook for the memory market going forward is as follows.
The shortage in the memory market looks set to last at least until 2030 (5 years), and it could go on longer
By next year or the year after, most makers of budget PCs, smartphones and set-top boxes will go under, and the electronics market will be reorganized around the large players that can still source memory. Xiaomi appears to be cutting smartphone production by 20-30% this year, and makers in an awkward position such as Oppo and Vivo also seem to be struggling. From Qualcomm's standpoint it is bad news, since it will not be able to sell many budget APs
In the spot market, manufacturers have nearly reached the peak memory price they can bear, so I believe channel price increases are now close to their top. The memory market prices mentioned in the news are mostly channel prices. They will keep rising at around 10% QoQ in the third and fourth quarters this year, but rising 2-3 times as before looks impossible. If prices rose like that, device makers would give up production altogether. However, the contract market, which accounts for more than 70-80% of memory makers' revenue, is still at about half the spot price, so depending on how LTAs are struck there is still plenty of room for price increases, and I expect memory companies' earnings to keep improving accordingly. For HBM4 alone, talk is of a price increase of more than 80% over HBM3e. I do not think it entirely impossible for the two Korean companies to post 1,000 trillion won in operating profit next year.
I agree with the logic that a phase is coming in which operating profit grows through Q (volume) rather than P (price). Memory prices are now approaching the top of what the market can bear, so from here money has to be made by increasing Q; the problem is that the growth rate of Q supply cannot be raised suddenly because of the lead time for fab buildings and equipment.
Memory companies' share prices have corrected sharply recently, and I see this as fully reflecting the assumption that the memory cycle will enter a down cycle again from the second half of 2027, as in the past. But looking at the recent situation I think that possibility is very small, and if we assume earnings will hold up well for the next several years, memory companies' share prices should eventually break through their previous highs in line with earnings. Probably, as we head toward the end of this year, if a consensus spreads through the market that the memory shortage will not be resolved quickly, there is a very high chance of returning to a broad uptrend.
There has been talk of halving the amount of memory mounted in Rubin, but I think nVidia is reducing the amount because memory is so hard and expensive to obtain, not because of issues such as algorithmic efficiency. Even if algorithms are made more efficient, with memory more is always better, so giving that up and cutting memory is proof of how severe the shortage is.
Many people talk about the threat from CXMT, and about ten years from now CXMT may not be something to dismiss. But CXMT too can only produce if it brings in foreign equipment; EUV cannot be bought at all and even DUV equipment is hard to obtain, so it is difficult for them to expand capacity as they intend. Even if they somehow work hard to expand it, memory is in such short supply that the impact on the overall memory market is negligible. The DUV that China recently claims to have developed is at the level of ASML's 2008 equipment, and it has not even been verified whether it actually works. Lithography equipment is the biggest bottleneck in semiconductor production and is not something that can be developed quickly. It will take China 10 years to make a proper DUV.
Unlike DRAM, the NAND shortage may perhaps begin to ease a little from the end of 2027 or from 2028. Already, for certain product lines, price increases have stopped and some correction is visible. For one thing, manufacturing difficulty is much lower than for DRAM so production capacity is easy to expand, and China's YMTC is working hard to expand capacity after localizing much of its equipment. HBF still seems premature, and it remains uncertain whether it will really become mainstream. The fact that SanDisk's share price corrected more sharply than DRAM makers' recently also reflects the market's worries, I think.
Many of you in Damoang's Jusik Handang are probably stuck holding memory company shares and having a hard time of it; my view is that you should close the price screen, focus hard on your day job, and I would venture to predict that we will see good movement again from the end of this year or next year. A bubble is when the share price alone rises far too much while earnings do not follow or stagnate; no matter how I look at the memory market, I think it is still far too undervalued to be called a bubble. Look at a company like Cisco in the early 2000s, with a PER above 200, whereas the two Korean companies are currently at a PER in the 3.x range. And earnings do not look likely to break down any time soon. I do not think this can be called a bubble.
I too intend to keep increasing my position every time memory company share prices correct.
Thank you.
