The Vice President of Micron stated that demand from AI data centers now accounts for 50%-60% of the DRAM market, fundamentally reshaping the supply and demand structure and serving as the core reason for the current shortage. Despite significant industry investment, capacity expansion faces systemic bottlenecks due to manufacturing complexity and product specification fragmentation. New production lines require several years to stabilize supply after initial operation. The tight situation in the memory market is expected to persist until around 2028, with little prospect of rapid improvement.
$Micron Technology (MU.US)$ Senior executives stated that despite significant investments, the memory shortage issue will not improve before 2028.
Christopher Moore, Vice President of Marketing for the Mobile and Client Business Unit at Micron Technology, one of the world's largest memory manufacturers, stated in an interview that the current tight situation is not due to manufacturers actively adjusting their client mix. Instead, it results from the explosive growth in demand for AI data centers combined with the continuously increasing complexity of memory manufacturing, which together have rapidly tightened the industry’s supply boundaries.
Against this backdrop, even with high levels of capital expenditure in the industry, a substantial improvement in memory supply will still require a long time horizon, and the related tight situation may persist until around 2028.
AI Consumes DRAM Capacity, Data Center TAM Expands Sharply
The structural change in current DRAM demand is key to understanding the shortage issue. The proportion of data center and AI-related demand in the overall DRAM market has rapidly increased from the previous 30%-40% to 50% or even 60%. In this context, the entire industry faces the reality of 'insufficient capacity,' rather than a resource allocation issue for a single manufacturer.
Moore noted that this is not a challenge faced by Micron alone but rather an industry-wide constraint affecting all major memory manufacturers. During the phase of concentrated expansion of AI data centers, no manufacturer can afford to ignore this shift in demand. He commented:
“This is not a problem specific to Micron; it is an industry-wide issue. Both we and our peers or competitors are doing everything we can to serve these niche markets, but the supply falls far short. It is indeed a regrettable situation.”
Consumers Are Not Being Neglected; Only ‘Channel Dynamics Have Shifted’
In response to market concerns that Micron’s exit from the Crucial brand implies an “exit from the consumer market,” Micron emphasized that its consumer-side presence continues primarily through OEM channels. The company still directly supplies LPDDR and DDR modules to original equipment manufacturers such as Dell and Asus, reaching end-user markets via device integration.
In Moore’s view, the rise in AI-driven demand does not equate to consumers being marginalized. Instead, it reflects the inability of supply to keep pace after the total addressable market (TAM) expands.
Capacity expansion is not simply about 'buying more equipment,' but rather a systemic constraint imposed by manufacturing complexity.
Micron Technology explicitly refuted the notion that 'simple capacity expansion can alleviate shortages.' The primary bottleneck in current DRAM production is not the number of machines, but the losses incurred from frequent line switches due to product specification fragmentation.
As smartphone and PC manufacturers increasingly demand multiple capacities such as 8GB, 12GB, and 16GB concurrently, wafer fabs must frequently switch between different designs (DID), which directly reduces effective output per unit time. Following the surge in AI-driven demand, manufacturers are now required to reduce the variety of specifications and extend the continuous production cycles of individual products to maximize bit output.
Moore said:
"Imagine this: if you have a fab with many different machines producing one type of chip, and then you have to stop those machines to run another type of chip, your yield will definitely drop. Of course, it’s not that simple in reality, but that’s the best way I can explain it. What we’re doing now is trying to minimize the number of chips and reduce the variety of DID chips as much as possible to maximize yield, understand? So we’re working closely with our customers."
The ramp-up of new capacity is proceeding slowly, with the critical inflection point expected around 2028.
Moore revealed that the ID1 fab, which Micron began constructing three years ago in Idaho, USA, has moved its scheduled production start date forward from the end of 2027 to mid-2027. However, from equipment installation, process ramp-up, to customer qualification, it will still take time for the new capacity to achieve scaled shipments.
In his view, only after the new production lines complete full qualification and achieve stable operation will there be any substantial change in the supply-demand dynamics of the memory market, with this milestone more likely occurring in 2028. Moore further stated:
Memory manufacturers are racing to build new production lines, but process limitations have ultimately forced them to pull their timelines forward by several quarters. This implies that for ordinary consumers, the DRAM shortage may persist for an extended period or at least until AI-driven demand begins to taper off.
Editor/joryn
