Micron Technology: Is the Memory Supercycle Just Getting Started?

Today (9/22/2026), I decided to take a long position in Micron Technology (MU), despite the current macro environment and the recent wave of concerns surrounding AI.

We are currently facing several headwinds: inflationary pressure from oil prices, tighter monetary-policy conditions, and pressure coming from bond yields. At the same time, however, semiconductor demand continues to increase, while the gap between supply and demand remains significant.

The key question for me is whether the current concerns around AI represent a temporary sentiment-driven pullback or whether they reflect a fundamental deterioration in AI-related demand.

The Micron thesis

Micron is a semiconductor manufacturer focused primarily on memory products such as DRAM, NAND and HBM. The stock is currently trading below its all-time high.

The key point of the thesis is that memory demand remains extremely strong, while supply is not increasing fast enough to fully satisfy it. This creates a supply-demand imbalance that could keep memory prices elevated over the coming years.

That situation directly benefits Micron. If memory prices remain high while demand continues to grow, the company could continue to increase both revenue and profitability.

More importantly, if earnings and EPS continue to increase while the market maintains a broadly similar valuation multiple, the share price could theoretically follow earnings higher.

For example, if Micron were to return to its previous all-time high of approximately $1,200 per share while maintaining an 8.3x P/E multiple, it would require approximately $145 of annualized EPS.

Looking at the quarterly numbers, if EPS were to increase from roughly $31 to $38 per quarter, and we assume a broadly proportional relationship between EPS and revenue, this would imply approximately $9 billion of additional quarterly revenue, or around 22% growth versus the previous quarter.

In other words, if Micron continues to grow earnings and revenue at this pace, maintaining an approximately 8x P/E multiple would require the share price to increase alongside earnings. Otherwise, the P/E multiple would compress, meaning that the market would be assigning a lower valuation to the company despite continued growth in revenue and earnings.

Conversely, if EPS continues to increase while the stock price does not rise at the same pace, the P/E multiple would compress, meaning Micron would trade at an even lower valuation relative to its earnings.

The key question

The main dilemma with this thesis is determining whether we are looking at peak earnings or whether Micron still has significant room to grow revenue and earnings.

This is where analyst estimates become important. These are not my own forecasts, but projections from firms such as Citi.

Citi currently expects DRAM demand to grow by around 30% in 2027, versus approximately 19% supply growth. That would imply an estimated supply-demand deficit of around 8.7%.

For 2028, Citi expects DRAM demand to grow by approximately 35%, versus 22% supply growth, with the estimated deficit widening to around 9.7%.

If these estimates prove accurate, the memory market could remain structurally undersupplied, particularly as AI-related demand continues to increase.

The supply side is also difficult to expand quickly. DRAM and HBM manufacturing require massive capital investment, advanced technology, and significant time to build new capacity.

Micron is increasing its own capital expenditure to respond to this demand, but a significant portion of the additional capacity will not reach the market immediately.

Therefore, the real question is how long this supply-demand imbalance can persist and, more importantly, how long current pricing and margins can remain elevated.

The most important part of the thesis

If the market is currently valuing Micron as a company approaching peak-cycle earnings, but the memory-demand cycle — particularly AI-driven demand — still has several years to run, earnings could continue to grow and the current valuation multiple could ultimately prove too low.

Micron is already showing how strong this cycle has become.

In fiscal Q3 2026, the company generated $41.46 billion in revenue, compared with $23.86 billion in the previous quarter, and reported GAAP EPS of $24.67.

For fiscal Q4, management guided toward approximately $50 billion in revenue and GAAP EPS of around $30.73.

The main risk is therefore that we are looking at extraordinary earnings that are not sustainable.

The main catalyst, on the other hand, is that DRAM and HBM supply remains constrained while AI-driven demand continues to grow.

Valuation

Regarding the stock price, Micron’s intraday all-time high was $1,213.56, reached in June 2026.

With the stock closing at $1,043.96 on September 21, 2026, shares are approximately 14% below their all-time high.

Bottom line

The thesis ultimately depends on three variables:

• Memory demand continues to grow.
• Supply remains constrained.
• Memory prices and margins do not decline significantly before new capacity comes online.

For me, the key question is not simply whether Micron’s earnings are high today.

It is whether the current earnings cycle is closer to the beginning of a structural AI-driven memory cycle than to the end of a traditional semiconductor cycle.

That is the thesis behind my long position in $MU.