Your computer will not come up because power-on self-test (POST) is failing somewhere between the wall outlet and the operating system loader. The most common causes are a loose power connection, dead CMOS battery, failed RAM contact, or corrupted boot configuration. Roughly 60 to 80 percent of no-boot cases resolve with power cycling, cable reseating, or RAM reseating before any part replacement.

That answer assumes the machine is truly dead-to-display rather than showing a logo and stalling. The two failures look identical to a user but diverge at different stages of the boot chain. Diagnosing which stage failed determines whether you need a screwdriver, a USB stick, or a new power supply.

The boot chain and where it breaks

A computer does not "start" as a single event. It executes a fixed sequence, and each stage hands control to the next only if its own checks pass. Understanding the sequence converts a vague symptom into a location.

Stage one is standby power. The power supply unit (PSU) converts AC wall current to the DC rails the motherboard needs, typically 3.3 V, 5 V, and 12 V. A small standby rail stays live whenever the PSU is plugged in, which is why the motherboard LED glows even when the machine is "off." If that LED is dark with the cord connected and the rear switch on, the fault sits upstream of the motherboard.

Stage two is the power-good signal. The PSU asserts a signal called PWR_OK only after its output rails stabilize within tolerance, usually a delay of 100 to 500 milliseconds. The motherboard will not release the CPU from reset until it sees this signal. A PSU that delivers voltage but never asserts PWR_OK produces a machine that spins fans and shows nothing.

Stage three is POST, run by firmware stored in the motherboard's SPI flash chip. POST initializes the CPU, memory controller, and RAM, then enumerates PCIe and USB devices. Firmware beeps or displays a debug code here when a component fails its check. This is the stage where reseating RAM fixes a surprising share of dead machines, because a marginal contact prevents the memory training sequence from completing.

Stage four is boot device selection. Firmware reads the boot order, finds the EFI System Partition on the drive, and loads the bootloader. A machine that reaches this stage shows a logo, then either an operating system or an error message about a missing boot device.

Stage five is the operating system handoff. The kernel loads, mounts the root filesystem, and starts services. Failures here produce a logo followed by a spinner, a blue screen, or a black screen with a cursor, not a machine that never lights up.

The plain version: power must flow, stabilize, pass self-test, find a drive, and load software. Any break produces "it will not come up," but the break point tells you what to check.

How often each stage fails, and what the numbers mean

Published failure-rate data for consumer PCs is thin because manufacturers do not release return-merchandise statistics. What exists comes from warranty providers, repair-shop aggregates, and reliability studies of components rather than whole systems.

Hard drive and SSD failure rates are the best documented. Backblaze, which publishes annual failure rates for tens of thousands of drives in continuous service, has reported annualized rates typically between 1 and 2 percent for most models, with outliers above 5 percent. That figure describes drives in a datacenter, not desktop PCs, and it measures annual failure, not no-boot causation.

Power supply failure is widely cited by repair technicians as a leading cause of sudden no-power conditions, but no authoritative body publishes a percentage. Treat any specific PSU failure statistic you encounter as unverified unless it names its sample.

CMOS battery depletion follows a predictable schedule. A CR2032 coin cell under continuous trickle load typically lasts 3 to 7 years. When it dies, the machine may still boot, but it loses clock time and BIOS settings, and some boards refuse to POST until settings are re-entered. A desktop built in 2018 is a plausible candidate in 2025.

Memory contact faults are common enough that "reseat the RAM" is standard first-line advice, but they are intermittent by nature and resist quantification. They often appear after a move, a dust buildup, or thermal cycling.

The honest summary: storage failure is well measured, power and memory failure are well known but poorly measured, and firmware corruption is rare enough that no reliable rate exists. A diagnostic sequence that checks the cheap, common causes first is more valuable than a probability table.

Edge cases that mimic a dead computer

Several conditions produce a machine that appears dead but is not failing in the ordinary sense.

A monitor connected to the motherboard's video output on a system with a discrete graphics card will show nothing, because the integrated output is often disabled when a GPU is present. The computer boots fine; the display is simply wired to the wrong port. This is one of the most frequent false alarms.

A display that is powered off, on the wrong input, or connected by a cable with a bent pin produces the same blank screen. Check the monitor's own power LED and input selection before opening the case.

Fast Startup in Windows performs a hybrid shutdown that saves kernel state to disk. A corrupted hibernation file can prevent the next boot from completing even though the hardware is healthy. A full shutdown (holding Shift while clicking Shut Down) clears the saved state.

Peripheral devices can block POST. A failing USB device, a stuck keyboard key, or a shorted front-panel connector can halt firmware initialization. Disconnecting everything except power and display is a legitimate diagnostic step, not superstition.

Laptop-specific cases include a battery that will not accept charge and a display cable that has worked loose at the hinge. A laptop that powers on with an external monitor attached but shows nothing on its own panel has a display-path fault, not a boot fault.

Misconceptions that waste time and money

The most expensive misconception is that a no-boot machine needs a new motherboard. Firmware, RAM, and power faults all present identically and cost far less to address.

The second is that beep codes are universal. They are not. Beep patterns are defined by the firmware vendor (AMI, Award, Phoenix, and others) and then customized by the motherboard maker. A pattern that means "memory error" on one board can mean "video error" on another. Look up the code for your specific board model.

The third is that a spinning fan proves the PSU is good. Fans draw little current and will spin on a rail that collapses under the load of a booting CPU. A PSU can supply 5 V standby and fail to hold 12 V under load. Only a load test or substitution proves a PSU healthy.

The fourth is that reseating components is placebo. It is not. Oxidized or thermally cycled contacts genuinely lose continuity, and the mechanical action of removal and reinsertion restores it. The caveat: reseating without grounding yourself can introduce electrostatic damage, so touch bare metal on the case first.

The counterintuitive detail professionals rely on: a computer that powers on, spins fans, and shows nothing may be protecting itself. Modern boards detect a short or an overcurrent condition and refuse to complete POST rather than risk damage. The silence is a feature, not a failure.

Diagnostic sequence and decision thresholds

Work from the wall outward. Each step eliminates a class of causes before you spend money.

Red flag: any burning smell, scorch marks, bulging capacitors, or a PSU that trips a circuit breaker. Stop, unplug, and do not power it again. Replace the PSU or have it serviced. Red flag: repeated clicking from the PSU or drive, or a machine that powers on for one second and shuts off. This suggests a protection circuit or a failing component, not a software problem. Verify the outlet with a known-good lamp or another device, then confirm the rear PSU switch is on and the voltage selector (if present) matches your region. Disconnect all peripherals except power and display. Try a different display cable and a different monitor input. Open the case, ground yourself, and reseat the RAM, pressing until the retention clips snap. Try one stick at a time in the primary slot. Clear CMOS by removing the coin cell for 30 seconds or shorting the clear-CMOS jumper, then retry. If the clock resets to an old date, replace the CR2032 (about USD 2 to USD 5). If the board has a debug LED or code display, read it and look up the code for your exact model.

Thresholds to reconcile: a machine with no standby LED and no fan spin points to PSU or wall power. A machine with standby LED but no fan spin points to the front-panel power switch or motherboard. A machine with fans but no display points to RAM, GPU output, or firmware. A machine with display but no OS points to storage or boot configuration. These four outcomes map cleanly onto the five boot stages and should be resolved in that order.

Symptom Most likely stage First action Typical cost if part needed
No lights, no fans Wall power or PSU Test outlet, swap power cord USD 50 to USD 120 for a PSU
Standby LED lit, no fans Front panel or board Jump the power switch pins USD 0 to USD 100
Fans spin, no display RAM or GPU output Reseat RAM, move display cable USD 30 to USD 90 for RAM
Beeps or debug code POST component check Look up code for board model Varies
Logo, then no OS Boot device or config Check boot order in firmware USD 40 to USD 100 for an SSD
Logo, then blue screen OS or driver Boot into recovery environment USD 0

Prices reflect typical US retail for replacement parts, not service labor.

Timeline of a normal boot versus a stalled one

The time course separates a slow boot from a failed one. A machine that is merely slow will still reach the desktop; a stalled machine will not progress regardless of how long you wait.

Elapsed time Normal behavior Stalled behavior
0 to 0.5 s Standby rail live, power button signal received Nothing, or brief fan twitch
0.5 to 1 s PSU asserts PWR_OK, fans spin up Fans spin, no display
1 to 3 s POST runs, firmware logo appears Blank screen persists past 10 s
3 to 10 s RAM training, device enumeration Beep code or debug LED holds
10 to 30 s Bootloader loads, OS kernel starts "No boot device" or firmware menu
30 to 90 s Services start, desktop appears Spinner hangs or blue screen
Beyond 2 min Machine idle at login No change; fault is not timing-related

If a machine has shown nothing for two full minutes with fans running, waiting longer will not help. The fault is in POST or earlier.

Practical limits of self-diagnosis

Some failures require equipment you probably do not own. A PSU load tester, a POST card that reads port 80 codes, and a known-good spare of each component are the professional toolkit. Without them, substitution is the only reliable test, and substitution requires spare parts.

Firmware corruption sits at the edge of DIY repair. Some boards support BIOS flashback from a USB stick without a CPU installed; most do not. A failed firmware flash can brick a board permanently, which is why manufacturers warn against interrupting it.

Data recovery is a separate problem from boot repair. A machine that will not boot may still have an intact drive. Removing the drive and connecting it to another computer through a USB adapter (USD 15 to USD 30) preserves your data before any repair attempt. Do this first if the data matters.

The limit of what can be concluded without inspection: no article can tell you which component in your specific machine failed. The sequence above narrows the candidates, and the symptom table maps symptoms to stages, but the final identification requires physical testing.

For anyone weighing repair against replacement, the economics have shifted. A 2020-era desktop that fails POST may need a USD 60 PSU or a USD 80 motherboard, both cheaper than a new system. A laptop with a failed motherboard often costs more to repair than its resale value, which changes the decision.

This article is general educational information about computer hardware behavior, not a diagnosis or repair plan for your specific device. It does not promise that any step will restore your machine or recover your data. For hardware under warranty, contact the manufacturer before opening the case, since self-repair can void coverage. If a device shows burning smells, scorch marks, or repeated breaker trips, stop using it and consult a qualified repair technician or your local fire safety guidance.