What DRAM is
DRAM stands for dynamic random-access memory. It is the working memory of a computer, the fast, temporary storage that holds the data a processor needs right now, as opposed to the permanent storage of a hard drive or a flash drive. When you open a program or load a webpage, the active data lives in DRAM.
The “dynamic” part matters. DRAM cells store each bit as a tiny electrical charge that leaks away quickly, so the chip has to refresh itself thousands of times per second just to hold onto its contents. That constant refresh is a design trade-off. It keeps the cells small and cheap enough to pack billions of them onto a single chip, which is what makes DRAM the workhorse memory of nearly every computing device on earth.
Why AI and self-driving cars need it
The reason DRAM keeps coming up in AI discussions is bandwidth. A modern AI accelerator processes data far faster than ordinary storage can deliver it. If the processor has to wait for data, all that compute goes to waste. DRAM, especially the stacked high-bandwidth variety, closes that gap by feeding the chip a continuous stream of data.
This is the mechanism behind one of the “Elon Musk’s 10X Project” claims. The pitch argues that fully autonomous driving needs “500 trillion operations per second,” a rate it says is only possible with DRAM. The specific number is marketing, but the direction is correct. A self-driving system processes a relentless stream of camera and sensor data in real time, and that workload is memory-hungry. The full teardown walks through where the pitch gets this right and where it stretches it.
DRAM versus storage
DRAM is easy to confuse with the flash storage in a phone or laptop, but the two do different jobs. Storage, like a solid-state drive, keeps data safe when the power is off. DRAM holds only what a processor is actively using, and it loses everything the moment the power goes out.
That difference is why DRAM is so much faster than storage. It trades permanence for speed. A computer uses DRAM as a scratchpad between the processor and the slower storage, shuttling data back and forth so the processor never has to wait on the disk. The more demanding the workload, the more DRAM it needs, which is why AI systems and self-driving cars consume so much of it.
Who makes DRAM
The DRAM market is one of the most concentrated industries on earth. Three companies, Micron Technology (MU), SK Hynix, and Samsung, control essentially the entire market. Two of them are South Korean. Micron is the only major US-based manufacturer, vertically integrated from design through fabrication.
That concentration is both a strength and a risk. On one hand, three suppliers have enough pricing discipline to keep DRAM profitable during normal times. On the other, it means the whole industry is a handful of decisions away from oversupply. When one player decides to grab share by ramping capacity, the others tend to follow, and prices fall. Our HBM explainer covers how the higher-margin stacked segment changes that math.
The cycle that never goes away
DRAM has spent decades in a boom-and-bust rhythm. Shortages push prices up, high prices justify new factories, new factories eventually oversupply the market, and prices collapse until the weakest players cut back. Then the cycle starts again.
Investors who ignore this cycle tend to buy DRAM stocks at the wrong moment. The pitch’s “+911%” figure, for example, measures where Micron traded when the presentation was recorded against where it trades now. It is a scoreboard for people who acted months ago, not a forecast for a new buyer. Understanding where the industry sits in the cycle, expanding or contracting, matters more than any single teaser number.
How to think about DRAM stocks
The honest frame for DRAM is as a cycle bet on a durable technology. The technology itself is not going anywhere. Every phone, computer, server, and AI accelerator needs it, and the AI buildout has only increased the demand. The question for an investor is timing and price, not whether DRAM matters.
A reasonable approach is to watch the three suppliers as a group. When they are disciplined about capacity, margins hold up. When one breaks ranks and builds, the whole sector pays for it. Micron’s position as the only US pure-play adds a strategic angle, but it does not remove the cycle. For context on the publisher running this pitch, see our Brownstone Research profile.
The bottom line
DRAM is the unglamorous memory that makes every AI and self-driving claim possible. It is real, it is essential, and it is controlled by just three companies. The 10X Project pitch leans on all of that correctly, then wraps it in a “secret supplier” story the market has long since priced in.
Ready to see the research? Click here to access the Near Future Report.
NewsletterVetter is an independent publication. We receive compensation from some of the services we review through affiliate links. Nothing on this site is investment advice. Always do your own research.