DRAM
Volatile semiconductor memory using capacitors and transistors for high-density storage.
Dynamic random-access memory (DRAM) is a type of volatile semiconductor memory that stores data in cells made of capacitors and transistors. It requires periodic refreshing to maintain charge and is widely used for main computer memory due to its low cost and high capacity compared to static alternatives.
- Uses one transistor and one capacitor per bit for high density
- Requires periodic refreshing to prevent data loss from charge leakage
- Dominant suppliers are Micron, SK Hynix, and Samsung Electronics
- Replaced magnetic-core memory as the standard in the early 1970s
- AI demand caused over 200% price increases in early 2026
What it is
Dynamic random-access memory, known as DRAM, is a form of semiconductor storage used extensively in digital electronics. Each memory cell typically holds one bit of data using a microscopic capacitor and a transistor based on metal-oxide-semiconductor technology. The capacitor stores an electric charge to represent binary values, but this charge leaks away over time. Consequently, the system must periodically rewrite the data to restore the original charge levels. This refresh process distinguishes it from static RAM, which does not need such intervention. DRAM is volatile, meaning it loses all stored information when power is removed, unlike non-volatile flash memory.
How it came about
Early dynamic memory appeared in World War II cryptanalytic machines like Aquarius, which used capacitors to store data temporarily. Toshiba introduced a bipolar dynamic RAM for calculators in 1965. In 1966, Robert Dennard at IBM invented the modern single-transistor MOS DRAM cell, offering a cheaper alternative to six-transistor SRAM designs. Advanced Memory Systems commercialized the first MOS DRAM chip in 1969. Intel followed with the 1103 in 1970, which became the first widely available DRAM despite initial yield issues. Mostek later introduced multiplexed addressing in 1973, reducing pin counts and costs. Japanese manufacturers dominated the market in the 1980s, leading to trade disputes with US firms. Samsung developed synchronous DRAM in the early 1990s.
Inside the chip
A DRAM chip arranges memory cells in a rectangular grid. Each cell connects to a word-line horizontally and bit-lines vertically. Reading data involves precharging the bit-lines to a midpoint voltage. Activating a word-line connects a cell capacitor to its bit-line, causing a tiny voltage change. Sense amplifiers detect this difference and amplify it to full logic levels using positive feedback. This process also recharges the capacitor, effectively refreshing the data. Writing data forces the sense amplifier to a specific state, charging or discharging the capacitor accordingly. The structural simplicity of one transistor and one capacitor per bit allows for very high densities and lower costs per bit compared to SRAM, which requires four or six transistors per bit.
How it was made
DRAM manufacturing has consolidated among a few major suppliers, including Micron Technology, SK Hynix, and Samsung Electronics. These companies control most of the global production capacity. In late 2025 and early 2026, demand from the artificial intelligence sector drove significant price increases for commodity DRAM. High Bandwidth Memory production consumes wafer capacity that might otherwise produce standard DDR5 chips, creating supply constraints. Other firms like Kingston Technology assemble modules but do not manufacture the underlying DRAM chips. China's CXMT announced design progress in October 2026, aiming to increase data density and compete with established manufacturers.
Where it went
This memory type serves as the main system memory in personal computers, laptops, and servers. It is also essential for graphics cards, where it functions as video memory. Portable devices and video game consoles rely on DRAM for their primary storage needs due to its balance of cost and capacity. Supercomputers use stacked DRAM variants for high-speed operations. While SRAM is preferred for processor caches due to its speed, DRAM remains the standard for large-capacity applications where cost efficiency is critical. The need for a memory controller to manage timing and initialization adds complexity to system design but enables the high densities required by modern software.
Versions and family
Synchronous DRAM (SDRAM), developed by Samsung, synchronized memory access with the system clock. The first commercial SDRAM appeared in 1992. Double Data Rate (DDR) SDRAM followed in 1998, transferring data on both clock edges. Recent market shifts show HBM crowding out commodity DRAM capacity. Variants include DDR4 and DDR5, which face supply pressure from AI-driven HBM demand. Stacked DRAM is used in exascale supercomputers. Some manufacturers integrate DRAM directly into other products, such as Fujitsu in CPUs or Nvidia in GPUs. Early variants included bipolar DRAM and three-transistor cells before the single-transistor MOS design became dominant.
Why it matters
DRAM replaced magnetic-core memory as the dominant technology in the early 1970s due to lower cost and higher performance. Its development enabled the rapid growth of personal computing by providing affordable high-capacity memory. Trade tensions in the 1980s between US and Japanese manufacturers highlighted its strategic economic importance. Price-fixing investigations in the early 2000s revealed market manipulation by several computer makers. The concentration of production among three major suppliers in the 2010s allowed them to control capacity tightly. Recent AI demand has reshaped the market, causing unprecedented price jumps and shifting focus toward high-bandwidth variants.
Problems and reception
US semiconductor makers accused Japanese competitors of dumping in 1985, citing drastic price drops that threatened their survival. The US Commerce Department ruled in favor of the complaint. Later, the US Department of Justice investigated price-fixing among computer manufacturers, finding several guilty in the early 2000s. In 2017, the market saw its largest price-per-bit increase in thirty years. By 2026, compounded price rises exceeded 200 percent for some types, driven by AI sector demand. Concerns about competition arose when Chinese firm CXMT announced new high-density designs in October 2026.




Questions people ask about DRAM
Why does DRAM need to be refreshed?
DRAM stores data as electric charge in microscopic capacitors. This charge gradually leaks away over time. Without intervention, the data would be lost. An external circuit periodically rewrites the data to restore the original charge levels, ensuring the information remains intact.
How is DRAM different from SRAM?
DRAM uses one transistor and one capacitor per bit, making it denser and cheaper but slower. It requires refreshing. SRAM uses four to six transistors per bit, making it faster and more expensive. SRAM does not need refreshing and is typically used for processor caches.
Who are the main manufacturers of DRAM chips?
The market is dominated by three major suppliers: Micron Technology, SK Hynix, and Samsung Electronics. These companies control most of the global production capacity. Other firms may assemble memory modules but do not manufacture the actual DRAM silicon chips.
Good reading and sources
- 1966: Semiconductor RAMs Serve High-speed Storage Needs | The Silicon EngineComputer History Museum, Article
- JAPAN CHIP 'DUMPING' IS FOUND (Published 1985)The New York Times, Article
- DRAM pricing investigation in Japan targets Hynix, SamsungEE Times, Article
- Dynamic random-access memory on WikipediaEncyclopedia
- More photos on Wikimedia CommonsPhotos
- Wikidata recordOpen data
The facts on this page come from the Wikipedia article "Dynamic random-access memory", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryChip rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.
It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.
Like DRAM
All memory chipsMore from Micron Technology
All of them
🇺🇸 United StatesRapidCAD
A 486DX in a 386 socket, with a dummy FPU chip to match.CPU · x86
WorldMOS Technology 8502
An 8-bit CPU from MOS Technology that powered the Commodore 128CPU · 8-bit
🇰🇷 South KoreaISOCELL
Samsung family of CMOS image sensors for mobile and digital camerasSensor