$GSI Tech(GSIT.US)$Everspin Tech(MRAM.US)
Recently, while looking at MRAM, I stumbled upon a company: GSIT.
At first, I thought it was just a small stock riding the MRAM hype, but after looking into it, I found its story quite interesting.
It originally made SRAM. SRAM can be simply understood as the very fast, expensive, low-capacity high-speed memory inside chips, typically used in places like cache.
The more imaginative part now is that it's working on something called Gemini APU. The core idea is: can we stop constantly moving data around and instead perform computation right near the memory.
I think this point is quite crucial.
Because in current AI computing, the real trouble isn't just "not computing fast enough," but also the significant time and power spent on moving data.
The process of moving data from memory to the computing unit and back is itself very resource-intensive. So, what GSIT is talking about isn't simply "I also make AI chips," but rather trying to solve part of the memory wall/power wall problem from a computing architecture perspective.
This is also why I look at it together with MRAM.
The appeal of MRAM lies in it not being a simple storage upgrade, but an attempt to combine some advantages of memory and storage. GSIT is somewhat similar; it talks about the changing boundary between computing and storage. One is "the form of storage has changed," the other is "the location of computing has changed."
Of course, GSIT is still very early stage.
It's not the kind of company that's already receiving massive orders from big clients with revenue exploding immediately. So my view on it is:
This isn't a stock you can mindlessly go all-in on, but it's the kind of small-cap hard-tech company well worth putting into a watchlist.
If the market later really starts hyping the AI memory, compute-in-memory, memory wall theme, GSIT might get a second look from capital. Its market cap isn't large now; if viewed only based on its old SRAM business, the valuation isn't cheap. But if the market starts believing it has a chance to break into AI inference, edge AI, or defense low-power computing, then the valuation logic would be completely different.
For now, I'm more inclined to view it as an early-stage technology option.
What it really needs to prove isn't whether it can rise today, but whether it can secure design wins later, and whether Gemini-II can turn into real revenue. As long as there's progress on these two points, the story will move one step from "concept" towards "validation."
Simply put:
MRAM's appeal is because it's like a breakthrough point in storage architecture.
GSIT's appeal lies in its potential disruption of computing architecture.
This kind of company carries high risk, but precisely because it hasn't been fully priced in yet, it's worth tracking.
@Miracle Trader cola
