Why Your Computer Forgets Everything When You Turn It Off

When you open a document on your computer, something specific happens in the hardware: the document's data moves from long-term storage — a hard drive or solid-state drive — into the computer's RAM, or Random Access Memory. The processor then reads from and writes to RAM to display, edit, and work with the document. This arrangement makes computation fast, because RAM is orders of magnitude faster than storage. It also makes computation precarious, because RAM is volatile — the instant power disappears, everything in RAM disappears with it.

The Physics of Volatility

Modern dynamic RAM — DRAM — stores each bit of data as an electrical charge in a tiny capacitor, paired with a transistor that controls access to it. A charged capacitor represents a 1; an uncharged capacitor represents a 0. The transistor acts as a gate, allowing the charge to be written or read.

The fundamental problem with capacitors is that they leak. A charged capacitor slowly loses its charge even without any deliberate discharge, due to tiny leakage currents that are unavoidable in real-world electronic components. Left alone, a DRAM cell that stores a 1 will gradually drift toward 0 as its capacitor discharges. This means DRAM cannot simply be written and left — it must be continuously refreshed, with each cell being read and rewritten thousands of times per second to maintain its stored value.

This refresh cycle requires power. When power disappears, the refresh cycle stops. Within milliseconds to a few seconds, the capacitors in DRAM discharge past the threshold where they can be reliably read as 1, and the data is gone. There is no gradual degradation or partial retention — the transition from functioning to data-lost is essentially instantaneous on human timescales.

Why Not Use Non-Volatile Memory for Everything?

If non-volatile storage — flash memory in SSDs, or magnetic domains on hard drives — retains data without power, why not use it for all computer memory? The answer is speed. DRAM can be read or written in nanoseconds — billionths of a second. Flash memory requires microseconds to milliseconds — thousands to millions of times slower. The processor in a modern computer executes billions of operations per second; if it had to wait for flash memory on every operation, it would slow to a crawl.

The speed difference exists because of the physical mechanisms involved. Reading a DRAM cell is a simple matter of opening a transistor gate and measuring the charge on a capacitor. Reading flash memory requires applying a voltage to a floating gate transistor and detecting whether charge is stored in an insulated layer — a more complex process involving quantum mechanical tunneling that inherently takes more time.

The result is that modern computers use a layered memory architecture that balances speed against persistence. The fastest memory — CPU registers — holds only a handful of values, measured in bytes. The L1, L2, and L3 caches hold kilobytes to megabytes of frequently used data in extremely fast static RAM (SRAM), which uses different circuitry than DRAM and is even faster but also more expensive and power-hungry. DRAM holds the active working memory of running programs, measured in gigabytes. Long-term storage on SSDs or HDDs holds everything else, measured in terabytes.

The Exception: Non-Volatile RAM

The tradeoff between speed and persistence has motivated decades of research into non-volatile RAM — memory that is as fast as DRAM but retains data without power. Several technologies have made commercial inroads, including Intel's Optane (based on phase-change memory) and MRAM (magnetoresistive RAM, which stores bits as magnetic states that are inherently stable without power).

These technologies offer intermediate tradeoffs — faster than flash but slower than DRAM, or DRAM-speed with persistence but at significantly higher cost per bit — and have found applications in specific markets. However, as of 2026, conventional DRAM remains the dominant working memory technology in most computers, meaning the fundamental situation has not changed: when the power goes off, what was in RAM is gone.

The persistence of this design reflects a deeper truth about engineering tradeoffs: the fastest solutions and the most durable solutions are often physically incompatible, and the architecture of modern computing is built around accepting that boundary and managing its consequences through software that saves state to durable storage before shutdown.