I Was Wrong Again on Memory Chip Stocks

Here's how to play it

I have memory chip stock PTSD.

Long ago earlier in my career, I recommended memory stocks for a fund. It did not go well.

The commodity memory industry is known for its boom-and-bust cycles. The stocks always look incredibly attractive at the worst possible times. Value-oriented investors (like me in the past) pile in at the peak because the price-to-earnings ratios look incredibly cheap relative to growth rates, just in time for the inevitable implosion in memory chip prices as demand falters as new fab capacity comes online.

Every time investors believe “this time is different.” Like Lucy holding the football for Charlie Brown, promising she won’t pull it away, but then she does.

I have the scars of getting my head handed to me losing tons of money in memory stocks. I vowed never to recommend them again, especially when buying more proprietary non-commodity chip stocks like Intel and Nvidia had performed so much better for me in the past.

So that is the background heading into this memory cycle. Let’s bring it back to what is happening now.

At first, I made a decent call. In mid-December, Micron posted blowout earnings and gave incredible guidance, forecasting a revenue range with a midpoint of $18.7 billion, above analysts’ expectations of $14.3 billion.

“Outside of NVIDIA this was likely the best revenue/$ net income upside in the history of the U.S. semis industry,” Morgan Stanley analyst Joseph Moore wrote at the time.

Moore is referring to the Nvidia earnings report from May 2023, when Nvidia provided a quarterly revenue outlook of $11 billion versus the $7.2 billion consensus, sparking the first wave of AI trade for stocks. I wrote about it and called it the “Big Bang” in my book.

Again at the time in mid-December, chip stock sentiment was weak and I posted on X that Micron’s incredible guidance meant chip stocks would no longer go down.

It worked. Chip stocks surged for a few weeks with Micron leading the rally, rising more than 40%. Then I remembered my PTSD and wrote a skeptical column in early January cautioning investors on memory chip stocks, arguing they shouldn’t forget the history of overcapacity problems and that Chinese companies are a disruption risk as they were likely to move more aggressively into the memory space.

The piece was too cute and wrong. Because this time may really be different.

Smart investors like Lux Capital’s Josh Wolfe and Atreides Management’s Gavin Baker have been adamantly bullish - and right - on memory stocks for a while. They called out how memory makers aren’t just producing commodity chips anymore, the dynamic random-access memory (DRAM) for computers and consumer electronics. They have also become key suppliers of high-bandwidth memory (HBM) for AI servers, which are far more advanced and complex, dare I say, proprietary.

So, the memory stocks kept rising heading into Big Tech fourth quarter earnings season in late January. And then the industry’s capex outlooks came out.

I’ve been very bullish on the AI infrastructure buildout, but the new capex guidance from the hyperscalers during the last few weeks were on another level.

Alphabet raised its 2026 capex guidance to $175–$185 billion versus the $115.6 billion estimate. Amazon raised its capex outlook to $200 billion versus $146.6 billion. Meta guided to $115–$135 billion versus $111 billion.

These unprecedented numbers will lead to MUCH higher spending for AI servers, which means more spending on CPUs (more on this in the future), more spending on GPUs, and much more spending on memory.

The facts have changed. This memory cycle is going to be far stronger than anything in the past. My PTSD has been a detriment that you shouldn’t let affect you. Memory chips are going to keep going higher.

So how should investors play it? It’s not Micron, but something else.