If you are tracking global equities like most people do, there is a blind spot hiding in plain sight. Fortune describes an informal “memi” sector, a portmanteau of “memory” and “semiconductors,” that does not have its own ticker but is pulling a surprisingly large share of market gravity toward three companies: Micron Technology, SK Hynix, and Samsung.

And here is the specific fact that matters for investors who think they are diversified: all three have now hit $1 trillion or more in market capitalization this year. Micron’s market cap stands at $1.1 trillion, SK Hynix has surged after raising $26.5 billion in the largest U.S. listing ever by a foreign company, and Samsung is up 116% year-to-date on the Korea Exchange. The headline answer is simple. The “memi” engine is real, it is funded by the $700 billion AI buildout, and it is already showing up in stock performance.

Why is memory suddenly the center of attention? Because big tech’s data-center race has a bottleneck that is not just about the AI chips everyone headlines. Fortune points out that Amazon, Google, Meta, and Microsoft all face a key constraint: getting memory chips. Most of the supply comes from these three manufacturers, so when AI infrastructure spend accelerates, memory demand pulls forward with it. The logic is brutally physical. Nvidia’s chips run AI models alongside workhorse memory chips called DRAM, which store the data those models need to train. Translation: the “brains” can be powerful, but without the “memory storage,” the system still does not run the way builders need.

This is also why portfolio managers are talking about “memi” across parts of the market that usually do not move together. Fortune says memi is fueling growth in U.S. small-cap funds, international funds, and emerging market funds. In other words, a portfolio that looks spread out on paper can still be exposed to the same choke point. The article makes the case that “memi” exposure is not confined to the usual large-cap chip crowd. Harbor Capital’s head of multi-asset solutions, Spenser Lerner, connects that dots explicitly: U.S. small-cap, emerging market, and developed market ex-U.S. outperformed the U.S. market, and the pattern traces back to the same cluster of memory chip companies.

The performance math is notable. Among emerging markets, the index returned 43.51% over the trailing one year, with most of the performance due to Korea’s Samsung and SK Hynix, alongside Taiwan Semiconductor Manufacturing Company. Lerner also says most of that performance came from earnings growth rather than investors simply paying higher prices. Korea and Taiwan now make up 51% of the index. In developed markets outside the U.S., the relevant MSCI index returned 20.8% over the trailing year, and Lerner attributes much of it to Japan. MSCI’s Japan index returned 29.5% during the same period, with gains tied to chip-equipment makers and memory manufacturers, including Tokyo Electron and Kioxia holdings, which were Toshiba’s memory business before it was spun out in 2017.

Even within small caps, the AI link shows up. Fortune cites MSCI’s small-cap index posting returns of 30.2% over the trailing year, with Sandisk as the largest holding. Sandisk makes high-performance memory cards based on flash memory, a different type of memory. So the theme is consistent: wherever the bottleneck lives, equity markets reroute attention. Lerner frames it as a way to describe the year: look for where the bottlenecks are, where hyperscaler and neo-cloud cash flow is going, and who benefits from the resulting demand.

Now for the risk, because memory is famously cyclical. Manufacturers ramp up capacity to meet demand, and eventually produce a glut that drives prices back down. The key question Fortune raises is whether today’s AI-driven investment is different enough to break the boom-and-bust pattern. Harbor’s outlook report offers a grounded constraint. It says pricing for DRAM has remained “firm,” and that a larger supply of memory chips is “unlikely to become meaningful before 2028.” If you are mapping board-level risks, that time horizon matters: it suggests the pricing power of Micron, SK Hynix, and Samsung could last for the next two years.

Micron’s recent operating numbers in the article reinforce why investors are willing to pay up now. Micron CEO Sanjay Mehrotra called AI’s appetite for memory bandwidth “insatiable” at the company’s third quarter earnings call in June. Micron reported total quarterly revenue of $41.5 billion, up 74% from the previous quarter and 346% year-over-year. Revenue from DRAM was a record $31.3 billion, up 343% year-over-year and made up 76% of its total revenue for the quarter. These are not vibes. They are bandwidth and pricing power showing up in the financials.

Fortune also notes that Roundhill Investments launched the first-ever memory ETF called DRAM in April, with top holdings including Micron, SK Hynix, and Samsung. The ETF is less than a year old and is up 162% so far, after peaking in June of 180%. This matters beyond trading because it signals that the market is building product wrappers around the bottleneck.

So what should executives and board members take away? If you are responsible for portfolio decisions, risk posture, or capital allocation signals, “memi” is a reminder that AI’s supply chain is not one-dimensional. Even though headlines obsess over Nvidia, memory has carved out a standout role, and it is linking distant parts of the equity map. If big tech ever pulls back on AI spending, the cyclical nature of memory could reassert itself, potentially turning a powerful run into a painful return to earth. For now, though, the market’s “memi” signal is unmistakable: bottlenecks move markets, and this one is currently made of DRAM.