Micron (NASDAQ:MU) stock has risen 91% over the past six months, turning a £10,000 investment into roughly £19,100 (the pound’s pretty much flat versus the dollar). That’s a return most investors would take over a full decade, let alone half a year.
So what’s driving it, and is there still a case to be made at current levels?
The new Intel?
Think about what actually happens when a hyperscaler upgrades its servers. It’s not just buying a new Nvidia GPU. The memory’s physically attached to the die, so new hardware means new Micron chips too.
The upgrade cycle and the memory cycle are the same cycle.
That’s a remarkably strong position to be in. It’s reminiscent of where Intel sat in the 1990s — not because the businesses are identical, but because both companies found themselves embedded so deeply into the prevailing technology stack that growth elsewhere in the industry translated almost automatically into growth for them. Intel rose 7,000% over that decade.
A noteworthy pullback
Last month, Google announced TurboQuant — an algorithm designed to reduce memory requirements for AI workloads by up to a factor of six. Memory stocks fell sharply. The market read it as a demand killer.
It probably isn’t. The same logic was applied to DeepSeek’s efficiency gains last year, and what followed was more compute spending, not less. When you can do more with less, developers tend to simply do more. The Jevons paradox has a habit of making itself relevant.
Valuation looks even better today
This is where it gets genuinely interesting. On a forward non-GAAP price-to-earnings basis, Micron trades at around 6.3 times — versus a sector median of 21.8 times. That’s a discount of roughly 71%.
The forward price-to-earnings-to-growth (PEG) ratio sits at 0.05, against a sector average of 1.32. Across almost every metric, Micron screens as one of the cheapest large-cap technology companies in the market.
The issue is the sustainability of earnings. Due to a shortage, memory prices have surged. And this is why some analysts see Micron’s earnings falling from 2027 to 2028.
With the company spending billions on new factories and capacity, the question is whether higher volume can offset moderating prices. Some will highlight that the sector has been very cyclical in the past. I’d suggest that cyclicality may not be an issue with AI.
What if it’s still cyclical?
As noted, the current cycle’s being supercharged by AI infrastructure spending at a scale the industry has never seen — customers are locking in supply years out, which itself signals how tight things are.
Some analysts argue that, at some point, capex from the hyperscalers moderates, and when it does, the memory market will feel it.
As such, appropriate position-sizing matters here.
For what’s its worth, I disagree with this thesis. Memory’s soldered into servers and GPUs and typically lasts the life of the hardware. The replacement cycle will reflect that.
I think institutional analysts agree — the price target’s 40% above the current price.
The bottom line
It’s absolute worth considering. The valuation discount is too large to ignore for a company that sits at the centre of the AI infrastructure build-out. The sell-off looks like an overreaction.
