What at first glance looks like a classic demand shock in the semiconductor market is, on closer inspection, nothing more than a strategically calculated supply bottleneck brought about by the two dominant players in the global DRAM business: Samsung Electronics and SK Hynix. Together they control around 70 percent of the global market, and they currently show no interest in jeopardizing this market position by generously expanding production. Quite the opposite: instead of investing in new production capacities or aggressively ramping up existing plants, both companies are focusing on a decidedly conservative investment regime with the aim of maximizing profitability per bit. In plain language, this means less supply but higher prices.
This stance has recently been clearly expressed at several investor conferences. Samsung, for example, made it unmistakably clear at a briefing with Morgan Stanley that it wanted to deliberately avoid the risk of oversupply. Instead, the company wants to ensure long-term profitability through targeted CAPEX initiatives that are closely aligned with market dynamics. The picture is the same at SK Hynix: There, the company announced that it would switch production more strongly to advanced sixth-generation 10nm D-RAM (1c) in the future. But here, too, there was no sign of any urge to expand; instead, it was conceded that even with these measures, the existing undersupply could hardly be solved before 2027. A diplomatically packaged declaration of bankruptcy for the term “security of supply”.
Why this reticence? Quite simply: demand, particularly from the high-growth AI server segment, is exploding, while at the same time the market is undergoing a technological upheaval. Old products such as DDR4 and LPDDR4x are being phased out, while new standards such as DDR5 and HBM (High Bandwidth Memory) are increasingly dominating production planning. This changeover takes time, results in lower yield rates and is technically demanding. In this mixed situation, manufacturers are consciously deciding against risky investments and in favor of a controlled supply shortage that keeps margins at a high level. Long-term supply contracts? Are currently rather undesirable. Any major customer who wants to commit for years to come runs the risk of being seen as an annoying contractual brake on price increases. Accordingly, both Samsung and SK Hynix rely on quarterly adjusted agreements, which means flexibility for manufacturers and planning uncertainty for customers.
This strategy is already having an effect: DRAM prices have risen by up to 60 percent in recent months, with an upward trend. Not only traditional PC manufacturers are particularly affected, but also smartphone manufacturers, cloud services and automotive suppliers. Wherever DRAM is a critical resource in production, costs are rising rapidly. And because pricing is increasingly short-term, smaller suppliers in particular are coming under pressure as they cannot afford to hold long-term stocks. Those who relied on just-in-time are now being overwhelmed by the reality of a fragmented spot market.
There is also a geopolitical aspect: while Western countries are trying to build up their own semiconductor production via “chip acts”, this shows how difficult it is to achieve genuine independence. Capacity alone is not enough, access to state-of-the-art production technologies, machines and know-how is required. And as long as this market is dominated by two South Korean giants, any form of sovereignty will remain wishful thinking. The USA, Europe and even China are on a drip in this matter, a fact that Samsung and SK Hynix obviously understand very well.
As a result, the so-called “supercycle”, which analysts like to interpret as a sign of a technological upturn, is turning out to be an asymmetrical market dynamic with oligopolistic structures. What we are experiencing is not a classic boom, but a calculated shortage situation with maximum profit skimming. The parallels with the commodities industry are obvious: there, too, supply is often artificially reduced in order to control prices. Now also in the memory segment.
Conclusion: The DRAM market is facing a structural bottleneck that is likely to last until at least 2027. Anyone hoping that prices will fall again soon is misjudging the strategic interests of the market leaders. Because from Samsung and SK Hynix’s point of view, things are going perfectly: high demand, scarce goods, rising prices. The rest? Must pay, or wait.
Source: Hankyung
11 Antworten
Kommentar
Lade neue Kommentare
Veteran
Urgestein
Veteran
Urgestein
Mitglied
Mitglied
Urgestein
Mitglied
Mitglied
Neuling
Urgestein
Alle Kommentare lesen unter igor´sLAB Community →