A Case on Micron

Plot: Over the past half a year, the best performing asset has not been crypto, the S&P or any other likely investment. Indeed, RAM Prices have seen a 3 times multiple on thier price since January. With the announcement of major players (Mircron, Sk Hynix, etc.) announcing their push towards STRICTLY enterprise customers, it has seemed like computer memory is now a luxury good.

It’s no suprise that subsequently, Micron, a leading manufacterer of computer memory has seen a rise in share price (20-30 %) YTD as of writing this.

Rising Action: Recent market conditions have seen a push away from debt in private markets. That is, we have seen withdrawal caps link yielding a slight panic in the public markets. It is clear these markets have strong correlation.

Although market conditions are due to change, that is not what concerns me. It is more of a capalyst in the smaller picture of memory markets.

Rising Action part 2: I would like to mention that I have been reselling RAM (random access memory) for the duration of this boom, I have made several thousand transations. The business is slow by design, but this is a different story entirely. In my practice, I’ve seen the RAM prices soar as people “panic buy”, but in the past weeks (maybe two weeks), I’ve seen what we might call flock behavior. Information has propogated to all corners of the internet and specifically to all those who own or were holding on to RAM.

More locally (across Ebay), we have that used RAM prices are stagnent. This link shows retail prices. Notice the lower GB modules which correlate more with the used markets like Ebay.

There are two cases we must consider.

First, we have higher supply, meeting demand. This means that specifically, suppliers are pumping more memory modules out to take a larger market share or whatever logic suits their hypothesis. As previously described, this is already happening in the used markets. As people find RAM under their desk or in their closet, demand is underweighted.

Second, consider we have lower demand. This is certainly the case in the consumer market as most people who work with computers cannot rationalize the higher prices especially because of the craze nature. That is, people are waiting for normality. I’ll leave it at that for the consumer market. The direct conclusion we can make is that enterprises will see a smaller consumer base. This yields a complete reliance on enterprise customers. (It should be noted that consumer markets already accounted for negligable amounts of business, but we use this more as a proxy for enterprise behavior)

Considering the enterprise market, we must understand how RAM is intellectualized. RAM is a fixed cost in datacenter hardware. Failure rates are often negligable, and these things last FOREVER. More specially, with larger models, we would like to have more RAM, but the cost is fixed in this regard (and innovation is slow, so there is no cycle of swapping to better products). As well, most major clients have made their most ambitious purchase orders over the last couple of months (OpenAI as an example link). There is certainly disbelief from the public in these never before seen multi-billion dollar orders, however, skepticism is hard to monitor, so we’ll estimate a small discount in the share price for such skeptisicm.

As it pertains to future orders, one can recall that inference has seen extreme dispersment in providers. See Ceribras for the major player. Thus, market share will likely see its tail end in the inference markets.

Considering training, this is where memory manufacturers will tend their argument. It is true that training, on the largest scale, requires terabytes of RAM. It would not be suprising to see a petabyte in the near future. We will contend with this argument. See link for stronger argument towards a less memory bottlenecked future.

Conclusion: Putting this all together, we have that RAM is relynig solely on enterprise customers for their business, yet a credit crash is causing private panic which has seen proxies in the public markets to feel the same impact. Those private customers are the buyers of memory, and indeed, will likely suspend buying orders just as manufacturers ramp up production. Lastly, as RAM is “forever lasting”, we will see a dump in the used markets for used RAM modules. Or, we are already there. Maybe, like The Eternal Sunshine, we’re already half way in this story. More specifically, memory has been seen as a luxury good over the past half year, yet the falling used markets indicate the change in tide from smaller players link who proxy the larger enterprise behaviors.

My Risk: I will be alloting 3 thousand dollars of risk capital in this regard. Implementation likely will involve long dated derivatives. Mostly in respect to a voiding of product orders. Liquid markets limit my choice to Micron Technologies (MU). This is inconsistent with my overall exposition as this relies on Micron not making a research effort and becoming an outlier.

Likely, other players in the computer hardware space will see second order effects, but those are harder to time. When the time comes, I hope to see this same logic play out. As an example, Western Digital.

Update (3/30): It seems like the markets responded. Micron is down 9 percent today as well as Western Digital. I was unable to capitalize in time, however, there is more price to capture in this move away from Memory and Storage hardware. I’m undercapitalized, so we’ll see.