George Gilder: Morgan Stanley’s Memory Problem

“Nobody will need more than 637 kb of memory for a personal computer.”

–Bill Gates

October 7, 2025 — Morgan Stanley says the market for wafer fab equipment (WFE) — the machines that turn silicon into circuits — is going to boom.

The firm raised its revenue estimate to $50 billion in 2026, up from $43.8 billion. Why? Memory. Their analysts’ argument is that higher dynamic random-access memory (DRAM) and NAND prices will support bit growth, justify more capital expenditures, and deliver upside to equipment suppliers like Lam Research (LRCX) and Applied Materials (AMAT).

Morgan’s analysts may have a memory problem of their own. Memory markets are notorious for their cyclicality. Average selling price (ASP) increases are often the result not of spontaneous demand but deliberate supplier actions to curb supply. Then when prices rise, the industry aggressively adds capacity. That leads to oversupply, inventory overhangs, and market crashes, not to mention rising debt burdens from all that capital expenditure (capex).

Here is a hot tip: stronger ASPs should be seen as a warning signal, not an invitation to flood the market with new equipment.

Morgan Stanley’s Bull Case

The Morgan Stanley analysts project DRAM bit growth of 22% in 2026, corresponding to $35 billion of DRAM equipment investment, and NAND bit growth of 26%, corresponding to $15 billion of NAND equipment investment. That adds up to their $50 billion WFE bull case.

Where will the money come from? They argue that higher ASPs will lift profitability for Samsung, SK hynix, Micron, and others, funding added capacity. They cite NAND supply tightness and the structural shift from HDDs to enterprise SSDs as reinforcing factors. For Lam Research and Applied Materials, this translates into roughly 7% and 5% EPS upside, respectively.

The critical point here is that Morgan Stanley interprets ASP increases as a demand signal, encouraging more tool purchases. But the recent pricing cycle shows that these ASP increases are not primarily demand driven.

The real story: price increases have been driven by supply discipline.

Memory ASP increases in 2025 are not the result of natural demand surges but deliberate supplier actions. Beginning in April 2025, Samsung joined Micron, SanDisk, and YMTC in announcing DRAM and NAND price hikes:

• Samsung raised DRAM and NAND prices by 3-5%, following negotiations with global clients.
• SanDisk hiked prices by over 10% effective April 1, then another 10% for new orders after Sept. 5.
• Micron also raised contract prices, citing tighter market conditions.
• YMTC followed suit with increases exceeding 10%.

TrendForce attributed these moves to a combination of factors: tariff-driven stockpiling, accelerating shipments of HBM3e 12-high products, recovery from inventory clearance in Q1, and strong demand from AI and data centers.

However, my analysis shows that much of the apparent tightness stems from capacity reallocation. Surging high bandwidth memory (HBM) demand has absorbed DRAM wafer starts, forcing suppliers to redirect production away from standard PC and server DRAM. This reallocation acts as supply discipline — tightening effective supply and allowing suppliers to sustain higher prices.

This matters because Morgan Stanley treats higher ASPs as a result of rising real demand, when they are the outcome of supply discipline and somewhat random factors such as tariff fears. If suppliers immediately respond to their own price increases with new capacity spending, they will undercut the very discipline they just imposed.

The Great Memory Blowout of 2018–2019

The memory industry has seen this playbook before. In 2018 and 2019, Micron, Samsung, and SK hynix aggressively expanded production capacity just as demand slowed. Micron in particular misjudged server NAND demand in mid-2018 and DRAM demand in 2019.

The result was catastrophic: DRAM ASPs fell 55% from their Q3 2018 peak, while NAND ASPs declined 44% from Q2 2018 onward. Inventories ballooned, profitability collapsed, and stock prices followed.

Chart 1 below shows how Micron significantly overestimated server NAND demand in mid-2018 and 2019 and DRAM demand in mid-2019. As a result, the company overshot production of both NAND and DRAM, creating an inventory overhang and a steep drop in ASPs during this period.

On May 30, 2018, Micron’s stock reached a high of $62.57, but DRAM and NAND bit shipments soon exceeded end-market requirements while server capex demand dropped precipitously in Q4 2018, bottoming out in Q1 2019. The stock retreated in tandem, reflecting the overshoot.

Thus, Micron’s share price during this period can be directly correlated with excess bit shipments, which drove ASPs lower. At the time, I warned in a series of Seeking Alpha articles that oversupply in DRAMs and NAND would force a downturn. The resulting cycle demonstrated how aggressive capex into rising ASPs can quickly invert into excess capacity, falling prices, and a market crash.

Chart 1. Server Capex vs. Micron’s DRAM and NAND Shipments (2018–2020)

In the chart, “server capex” refers to the investments cloud service providers and hyperscalers make in servers, storage systems, and networking gear. It reflects demand from companies like Amazon, Microsoft, Google, and Meta for the hardware that powers data centers and reflects demand for memory. In 2018 and again in 2019, memory shipments far outpaced server capex.

Table 2 underscores the connection between Micron expanding shipments between late 2018 and 2020 with sharply falling ASPs.

Morgan Stanley’s Circular Logic

Morgan Stanley interprets rising ASPs as an indicator of robust demand and a green light for capex expansion. That is just the fallacy the industry has fallen for so often before.

But the minute suppliers act on that interpretation, they destroy the very pricing power they engineered. The logic is circular: ASPs go up because supply is restrained; then equipment is purchased to add supply; then ASPs collapse.

The Hyperscaler Twist

Hyperscalers — Amazon, Google, Microsoft, Meta — compound the risk of the memory cycle. These companies are price sensitive and optimize their procurement. If DRAM and NAND prices spike, hyperscalers will slow purchases or stretch refresh cycles. They may also intensify efforts to reduce memory intensity per workload. ASP increases do not automatically unlock incremental demand; they risk curbing it.

Implications for Lam Research and Applied Materials

For Lam Research and Applied Materials, Morgan Stanley’s $50 billion bull case assumes memory producers translate higher ASPs into aggressive capex. But if suppliers stick to their pricing discipline, Lam and AMAT will not see the full order surge Wall Street is modeling.

Equipment demand will likely be selective — focused on technology transitions like DRAM 1β to 1γ or NAND string stacking — rather than broad-based capacity additions. Orders will be smaller, more targeted, and insufficient to deliver the $50 billion upside Morgan Stanley envisions.

Investor Takeaway

Overspending during periods of rising ASPs is self-destructive. For most products, today’s ASP increases result less from natural demand pull and more from supplier-enforced discipline. If memory makers treat them as justification for a capex binge, they will repeat past mistakes and trigger another collapse.

The $50 billion bull case for WFE in 2026 rests on a faulty assumption. Lam and AMAT may benefit from selective investments, but the cycle-defining upturn Morgan Stanley describes is unlikely.

Investors should temper expectations. If history repeats — and memory markets have a way of doing so — the companies that preserve pricing power will outperform, while equipment suppliers may find that the promised order boom never fully materializes.

George Gilder
Gilder Report & Grey Swan Investment Fraternity

P.S. from Addison: This is one of George’s recent essays, looking at some of the nuance needed in today’s rapidly-inflating AI stock bubble. We look forward to exploring that nuance when George joins us on Grey Swan Live! this Thursday.

George once handed President Reagan the first microchip, and now he says today’s tech wave dwarfs the original $6.5 trillion tech revolution of the 1980s.

Eight exponential technologies — AI, quantum computing, robotics, self-driving cars, blockchain, chips, advanced biotech, and even space — are no longer advancing in isolation.

They’re colliding, compounding, and accelerating into what could be the single greatest wealth-building event of our lifetimes.

The pace is staggering.

That’s why in Grey Swan Live!, we’ll show you how to navigate this convergence — and how early positioning could define not just your portfolio, but your legacy.

It’s not too late. Join us, won’t you?