Analysts from the Milk Road AI service are questioning the recent forecast by Morgan Stanley, which predicted an imminent peak in the memory market. In their view, the bank is using an outdated methodology that fails to account for fundamental changes in demand structure. Contrary to these expectations, the market is not on the verge of a downturn but is in the midst of a powerful supercycle.
Methodology Error: Classics vs. Reality
Morgan Stanley predicts that the annual growth rate of the memory market will peak, and contract prices for DRAM will hit a high in the fourth quarter of 2026. However, as experts rightly note, the bank relies on a classic analysis method that has been used for decades to assess the cyclicality of consumer goods—PCs and smartphones. This approach, based on inventory fluctuations, is no longer adequate today.
The key issue is that Morgan Stanley is applying a thirty-year-old model to a completely new reality. The main product of the modern market is not ordinary DRAM but high-speed HBM (High Bandwidth Memory). And this is not a commodity that goes into oversupply every few years due to the launch of new factories.
HBM: A Strategic Asset with Shortage Until 2027
HBM is a strategic asset developed for specific AI chip architectures. Its production is so complex that simply "opening a new factory and starting production" is impossible. This is confirmed by the manufacturer itself. Micron has stated that all its HBM production capacity for 2026 is already fully sold out, and the shortage will persist at least until 2027.
Moreover, demand is so high that manufacturers can only meet a portion of it. According to analysts, in the medium term, Micron can only cover 50–67% of its key clients' needs. This indicates a structural, not cyclical, shortage.
Micron Financials: No Peak in Sight
Micron's financial results completely refute the thesis of an imminent peak. The company's revenue for the third fiscal quarter of 2026 reached $41.46 billion, exceeding expectations by 346%. Gross margin hit a record high of 84.9%.
The forecast for the next quarter is no less optimistic: Micron expects revenue of $50 billion, which clearly does not fit a downturn scenario. The company's strong position is also reflected in its contracts: over the past quarter, Micron signed 16 strategic agreements—long-term contracts with volume commitments and minimum prices, replacing previous short-term deals.
HBM Demand: Built into AI Accelerator Manufacturers' Plans
Demand for HBM is long-term embedded in the plans of AI accelerator manufacturers. This includes the roadmaps of Nvidia, AMD, and all major data center operators developing their own chips. The market's scale is confirmed by TrendForce forecasts: total memory market revenue in 2026 will grow by 134% to $552 billion, and the HBM market will reach $100 billion by 2028, up from $35 billion in 2025.
My conclusion as an analyst: The memory market is indeed experiencing a supercycle, not a cyclical peak. Those predicting an imminent downturn are simply using the wrong map. The current situation is not a repeat of past cycles but a completely new paradigm driven by insatiable demand for AI infrastructure. Investors should focus not on short-term fluctuations but on the long-term trend that will define the market at least until the end of the decade.