The Capacitor Plague: How Stolen Electrolyte Killed a Decade of Hardware

Sometime around 2003, a measurable fraction of the world’s personal computers began dying of the same disease. The symptom was visible to the naked eye: small aluminum cylinders on the motherboard, normally flat-topped, swelling into little domes, splitting along their scored vents, and weeping brown crust onto the PCB. Machines that had worked perfectly for eighteen months started failing to POST, rebooting under load, corrupting memory. The cause traced back to a single act of industrial espionage gone wrong — a stolen electrolyte formula that was stolen incompletely, missing the additives that kept the chemistry from slowly manufacturing hydrogen gas inside a sealed can. The trade press called it the capacitor plague, and before it burned out it had touched Abit, Dell, Apple, HP, IBM, and Intel, cost Dell alone on the order of $420 million, and put the phrase “bulging caps” permanently into the vocabulary of anyone who fixes computers.

This is one of the great failure stories in computing — equal parts materials science, supply-chain economics, and corporate cover-up — and it is not purely historical. The wear-out mechanism that the bad electrolyte accelerated a hundredfold is the same one quietly running in every wet electrolytic capacitor in your homelab right now, just on a slower clock. If you buy decade-old enterprise gear (and if you read this blog, you probably do), the capacitor plague is the cautionary tale that teaches you how to evaluate it. So let’s do both: the history, properly told, and the practical residue — how capacitors age, how to spot the bad ones, and when a recap job beats the e-waste bin.

How an Aluminum Electrolytic Capacitor Works — and Why It’s the Part That Dies

Every capacitor is two conductive plates separated by an insulator. The aluminum electrolytic gets its absurd capacitance-per-dollar from two tricks. First, the anode “plate” is aluminum foil that has been electrochemically etched into a microscopic canyon landscape, multiplying its effective surface area by a factor of up to a hundred or more. Second, the insulator is not a sheet of plastic but a layer of aluminum oxide (Al₂O₃) grown directly on that etched foil by anodization — a dielectric layer on the order of a nanometer per volt of rating. Capacitance scales with area and inversely with dielectric thickness, so huge area plus vanishingly thin insulation equals thousands of microfarads in a can the size of a fingertip.

But there’s a catch hiding in that geometry: how do you make electrical contact with the other side of an oxide layer that follows every contour of a microscopically rough surface? No solid foil can do it. The answer is a liquid — the electrolyte — soaked into a paper separator wound between the anode foil and a second aluminum foil. The electrolyte is the true cathode plate; the second foil merely connects to it. This is the component’s genius and its death sentence in one design decision. The electrolyte is a liquid in a warm place, so it slowly diffuses out through the rubber end seal over years; as it dries, the contact area shrinks, capacitance droops, and equivalent series resistance (ESR) climbs. The electrolyte also performs a second job: when a flaw appears in the oxide layer, leakage current through the electrolyte re-anodizes the spot and heals it, consuming a little electrolyte and evolving a little gas each time. Both mechanisms make the wet aluminum electrolytic the canonical wear-out component on any circuit board — the one part with a datasheet lifetime measured in thousands of hours, surrounded by silicon rated for decades.

A well-made capacitor manages this gracefully. The electrolyte contains inhibitors and passivators that stop water from attacking the aluminum, and depolarizer additives that chemically absorb the hydrogen gas evolved at the cathode. The can has scored vent lines stamped into its top (the K, X, or T pattern you’ve seen) so that if pressure ever does build, the can opens in a controlled split instead of detonating. Take away the additives, and all of those safety margins invert.

              aluminum can (also the case)
              |
     .--------+--------.        <-- scored vent stamp (K / X / T)
     |  _______________ |           on a HEALTHY cap: dead flat top
     | |               ||
     | | anode foil    ||  etched, anodized Al (oxide = dielectric)
     | | ~~~~~~~~~~~~~ ||  paper separator soaked in ELECTROLYTE
     | | cathode foil  ||  (the liquid is the real second plate)
     | | ~~~~~~~~~~~~~ ||
     | |  ...wound...  ||
     |  --------------- |
     |   rubber bung    |
     '----+--------+----'
          |        |
        (+) lead (-) lead

   FAILURE SEQUENCE with the flawed electrolyte:

   2 Al + 6 H2O  ->  2 Al(OH)3 + 3 H2      (no inhibitor to stop it,
                                            no depolarizer to absorb H2)
           ____
         /      \      <-- top domes outward         ("bulging")
        |  ////  |     <-- vent splits, electrolyte
         \__  __/          dries to brown crust       ("venting")
            ||
   or: pressure extrudes the rubber bung out the bottom,
   tilting the whole can off the board

The Heist: A Formula Stolen Twice, and Incompletely

The origin story, as pieced together by the trade press in 2002–2003, runs like this. Around 2001, a materials scientist left Rubycon Corporation in Japan — one of the premier capacitor houses, maker of the highly regarded ZA and ZL low-ESR series — and took the water-based electrolyte formula for those parts with him to Luminous Town Electric in China, where he had previously worked. The theft, on its own, might have produced nothing worse than cheaper competition: water-based electrolytes were the hot technology of the moment, because water’s high conductivity is exactly what you want for the ultra-low-ESR capacitors that CPU voltage regulator circuits were beginning to demand. Rubycon had spent years learning how to use water aggressively and safely.

Then the formula was stolen a second time. Members of the scientist’s staff defected from the Chinese operation, copied the formulation, and began selling it to aluminum electrolytic manufacturers across Taiwan, undercutting Japanese electrolyte prices. And here the story turns from espionage caper into chemistry lesson: the copy was incomplete. As Dennis Zogbi, publisher of Passive Component Industry — the trade magazine that first reported the problem in September 2002 — put it to IEEE Spectrum, “it didn’t have the right additives.” The formula that propagated through Taiwan’s capacitor industry carried the high water content but lacked the proprietary inhibitors that keep water from corroding aluminum, and the depolarizers that absorb hydrogen gas.

The consequences were thermodynamically inevitable. Water in direct contact with unprotected aluminum drives a strongly exothermic hydration reaction — aluminum becomes aluminum hydroxide, and hydrogen gas comes off as a byproduct. Inside a hermetically sealed can, that hydrogen has nowhere to go. Pressure builds over months of operation. Eventually the scored vent does its job and splits, or the rubber bung extrudes out the bottom, or — in the dramatic cases that gave the plague its folklore — the can lets go with an audible pop and a spray of electrolyte. Meanwhile, even before the mechanical failure, the corroding foil and degrading electrolyte send capacitance down and ESR up, so the capacitor stops doing its job of smoothing the supply rails long before it visibly bulges. A CPU fed by a VRM with dying output caps doesn’t get a clean 1.5 volts; it gets ripple, droop, and crashes that look exactly like bad RAM or a flaky operating system. Most of the affected capacitors were manufactured between 1999 and 2003 and failed between 2002 and 2005 — a delayed-action defect, which is the worst kind, because by the time the failures surfaced, millions of boards had shipped and the parts had passed every incoming inspection.

A caveat that honest tellings of this story should include: the espionage narrative comes substantially from one chain of trade-press sourcing — Zogbi’s reporting, amplified by IEEE Spectrum’s February 2003 article “Leaking Capacitors Muck Up Motherboards” — and some details have never been independently confirmed in court or by the companies involved. The leaking parts pulled from failed boards in the US mostly carried the Tayeh label, with others from Taiwanese firms like Jackcon; the precise path from stolen formula to each affected brand is murkier than the clean story suggests. And critically, the stolen formula does not explain every failure of the era — as we’ll see, the biggest single casualty was felled by a Japanese manufacturer’s unrelated mistake. The plague was really two overlapping epidemics with one symptom.

The Body Count: 1999–2007

The first public reports came in late 2002, when Passive Component Industry flagged premature failures tied to Taiwanese raw materials, and enthusiast sites began documenting boards from premium manufacturers dying at twelve to eighteen months. IEEE Spectrum’s February 2003 piece pushed the story mainstream. The enthusiast motherboard maker Abit — whose boards of the era, including the wildly popular IS7 line of Springdale boards, became a byword for bulging caps — earned the dubious distinction of being the only affected manufacturer to publicly admit that defective Taiwanese capacitors had been used in its products. Everyone else stonewalled, replaced boards quietly under warranty, or litigated.

The Dell episode deserves its own paragraph, because it transformed a component defect into a corporate-governance scandal. Between May 2003 and July 2005, Dell shipped roughly 11.8 million OptiPlex desktops — the GX270 and GX280, the default corporate fleet machine of the era — carrying capacitors from the Japanese maker Nichicon that were at risk of failing. Note the nationality: these were not stolen-formula Taiwanese parts. Nichicon’s HM and HN series failures are generally attributed to a separate flaw in its own water-based electrolyte — a bitter irony, since buyers had specified Japanese caps precisely to dodge the Taiwanese problem. What made it a scandal was the response. Documents unsealed in 2010, in a lawsuit brought by customer Advanced Internet Technologies, showed Dell’s own internal study projecting that 97% of GX270s would fail within three years — and showed the company selling them anyway while instructing salespeople, in a manager’s now-infamous 2004 email, to “avoid all language indicating the boards were bad or had ‘issues.’” Dell took a roughly $300 million charge in 2005 against fixing and replacing the machines, with reporting on the unsealed documents putting the total cost of board replacements and triage logistics at about $420 million. The reputational charge was larger and is harder to quantify. If you ever need a case study for why blameless, public incident handling beats quiet containment, this is it — the same lesson we draw in the art of the postmortem, at nine-figure scale.

Apple’s chapter was smaller but instructive: first-generation iMac G5s sold from September 2004 to June 2005 suffered the same bulging-capacitor power-section failures, and Apple responded with a Repair Extension Program covering affected machines for three years from purchase, free even out of warranty. Same disease, materially better bedside manner.

It is worth pausing on what “millions of motherboards” actually means. A capacitor is about the cheapest active decision on a board’s bill of materials — fractions of a cent to a few cents in volume. The plague’s total damage plausibly ran into the billions of dollars across the industry, all downstream of a per-unit saving measured in pennies. Fifty years after Bell Labs gave away the transistor and the industry learned to mass-produce reliability in silicon, the thing that brought fleets of Pentium 4s to their knees was salt water in a can.

Spotting a Bad Cap

The plague’s one mercy is that its signature failure mode is visually obvious, and the diagnostic skills it forced a generation of technicians to learn still work today.

Visual inspection. Look at the tops of every can-style capacitor: they should be perfectly flat. Any doming — even a subtle one you confirm by sighting across the board edge-on — means internal pressure, which means gas, which means the chemistry has gone wrong. A split vent with brown or crusty residue is a corpse. Check the bases too: a capacitor sitting tilted, or lifted slightly off the board, often means the bung has been pushed out the bottom, and electrolyte may be pooling under the can where you can’t see it, eating traces. Around VRM circuitry, also look for electrolyte staining on the PCB.

Electrical inspection. Visual checks catch the late stage; the cap was electrically dead long before it bulged. The two numbers that matter are capacitance (drifts down as electrolyte dries) and ESR (climbs, often by an order of magnitude, and is by far the more sensitive indicator for the low-ESR parts used in power filtering). A rough field rule from the badcaps.net community: a capacitor reading two to three times its datasheet ESR is bad, full stop. For context, healthy motherboard VRM caps (820 µF and up) should read in the range of roughly 5–40 milliohms; PSU secondary-side caps of 1000 µF and up around 30–60 milliohms.

Behavioral symptoms that should send you hunting for caps: machines that won’t start when cold but run once warm (or vice versa), spontaneous resets under load spikes, failure to POST with a known-good PSU, USB devices browning out, and memory errors that move around between sticks. On running gear, rail telemetry is your friend before you ever pick up a screwdriver:

1
2
3
4
5
6
7
8

# IPMI-equipped servers: check rail voltages, ideally under load
ipmitool sdr type Voltage

# Desktop boards: same idea via lm-sensors
sensors | grep -Ei 'in[0-9]+|12v|5v|3.3v|vcore'

# Kernel-visible fallout that often tracks rail sag on aging boards
journalctl -k | grep -iE 'mce|machine check|whea'

A 12 V rail that reads fine idle but sags hard the moment you load the CPUs is the classic signature of bulk capacitance that has quietly evaporated.

Why It Still Matters: Capacitor Math for the Used-Gear Era

The plague ended around 2007, but the physics didn’t. Every wet electrolytic ever made is on the same clock the bad ones were on — just running at the design rate instead of a runaway one. If your homelab includes a decade-old Dell R720, a secondhand UPS, or a stack of retired switches (and per our homelab hardware guide, it probably should), then capacitor aging is the dominant reliability question about that hardware, and you can actually do the math.

Electrolytic capacitor life follows the Arrhenius relationship, which collapses in practice to the “10-degree rule”: every 10°C reduction in operating temperature roughly doubles life, as L = L₀ × 2^((T₀ − T)/10). A garden-variety capacitor rated 2,000 hours at 105°C sounds alarming — that’s twelve weeks — until you apply the rule: at 75°C it’s 16,000 hours, and at 55°C it’s around 64,000 hours, or seven-plus years of continuous duty. Premium parts rated 5,000 or 10,000 hours at 105°C, run at sane temperatures, plausibly outlive the platform. Two corollaries fall straight out of the math. First, airflow and ambient temperature are capacitor-life multipliers: the difference between a server breathing 22°C air and one cooking in an unventilated closet at 35°C is, by the rule, roughly a 2.5x difference in capacitor lifespan — one of several reasons proper power and cooling planning, like sizing your UPS correctly, pays compounding dividends. Second, placement matters as much as rating: the caps that die first are the ones parked next to heatsinks and in PSU enclosures, where local temperature, plus self-heating from ripple current, runs far above ambient. The rule is an approximation — the doubling factor holds best in roughly the 60–95°C range — but it is accurate enough to be the working tool the industry actually uses.

The other thing that changed after the plague is the parts themselves. Motherboard makers, burned badly, migrated VRM filtering to solid polymer capacitors — the “all solid capacitor” badges plastered on boards from about 2006 onward were direct marketing fallout from the plague. Polymers replace the liquid electrolyte with a conductive solid polymer, which removes the dry-out mechanism and the gas-generation failure mode entirely.

Note the trade-offs: polymers are not a universal upgrade. Their voltage ceilings rule them out on PSU primary sides, their leakage current is worse, and their ultra-low ESR can actually destabilize older switching regulator feedback loops that were designed expecting some resistance in the output filter. This is why wet electrolytics still fill every power supply you own — including the ones in your “all solid capacitor” era servers. The plague-vulnerable component class never left; it just retreated inside the PSU casing where you don’t look at it.

Recap or Recycle: The Practical Decision

So you’ve bought, inherited, or unearthed aging hardware. What do you actually do?

Get an ESR meter. This is the single tool the plague era canonized. The classic is Bob Parker’s ESR meter design; the cheap modern options are the MESR-100 or any of the ubiquitous transistor-tester component checkers (LCR-T4, TC1 and clones), which read capacitance and ESR for under $30. ESR meters test at around 100 kHz with millivolt-level signals — too low to forward-bias semiconductor junctions — which means you can often test capacitors in circuit without desoldering, at least to the resolution of “fine” versus “obviously dead.” For borderline readings, pull one leg. Always discharge capacitors first, and treat PSU primary-side caps (those 400 V cans hold genuinely dangerous charge) with respect: check with a meter, bleed through a resistor, never a screwdriver.

Decide if a recap is rational. Recapping — desoldering every suspect electrolytic and fitting new ones — is a few dollars in parts and a few hours of soldering on multilayer boards with big ground pours that fight your iron. The honest decision matrix: recap when the board is irreplaceable or expensive — retro hardware, a discontinued controller, audio gear, a motherboard whose socket generation has cult value, an LCD monitor whose PSU board needs four caps and thirty minutes. Skip it when the board is a commodity: a $40 used OptiPlex-equivalent is not worth two hours of bench time plus the risk of lifted pads, and a swollen-cap consumer PSU should simply be replaced, since by the time caps fail, the fan bearings, fuses, and design margins are usually equally tired. When you do recap, match or exceed the originals: same capacitance, equal or higher voltage and temperature rating, equal or lower ESR (with the polymer caveats above), and reputable Japanese brands — Rubycon, Nichicon, Panasonic, Chemi-Con — bought from real distributors like Digi-Key or Mouser, because the counterfeit capacitor market will happily sell you the plague all over again.

For used enterprise purchases, fold caps into your evaluation ritual: pop the lid, sight across every visible can top, look for staining around VRMs and inside the PSU, check the manufacture date (a 2012 board has had its bulk caps in service or storage for fourteen years), pull SEL logs and voltage sensor history over IPMI before money changes hands, and discount accordingly for gear that lived in a hot closet. None of this takes ten minutes, and it routinely catches the one machine in the lot that will otherwise eat a weekend.

Verdict

The capacitor plague is the best kind of engineering story: a microscopic cause with an industrial-scale effect, fully explicable after the fact, and preventable at every step. A formula stolen without the additives that made it safe; capacitors that passed every spot check and failed in the field eighteen months later; vendors who mostly chose silence, and one — Dell — whose silence converted a supplier defect into a $420 million scandal and a permanent reputational scar. The deeper lesson is about wear-out components generally: every system has one part that is consumed by operation rather than merely used by it, and the wet electrolytic capacitor is that part on nearly every board ever made. The plague just compressed twenty years of aging into two, so an entire industry got to watch the failure mode in fast-forward. Learn the visual inspection, buy the $30 ESR meter, run your gear cool, and treat capacitor age as a first-class line item when evaluating used hardware. The chemistry has not changed; only the timetable has.

Sources

• https://en.wikipedia.org/wiki/Capacitor_plague
• https://spectrum.ieee.org/leaking-capacitors-muck-up-motherboards
• https://www.pcgamer.com/the-capacitor-plague-of-the-early-2000s/
• https://thetech.com/2010/11/19/capacitors-v130-n55
• https://venturebeat.com/2010/11/19/unsealed-lawsuit-reveals-dell-lied-about-millions-of-faulty-computers/
• https://money.cnn.com/2010/07/01/technology/dell_lawsuit/index.htm
• https://www.computerworld.com/article/1349376/how-a-capacitor-popped-dell-s-reputation.html
• https://en.wikipedia.org/wiki/Aluminum_electrolytic_capacitor
• https://www.rutronik.com/article/water-based-electrolytic-capacitors-from-plague-to-indispensable-component
• https://www.osti.gov/biblio/5118443
• https://www.chemi-con.co.jp/en/faq/detail.php?id=alLifetime
• https://www.xppower.com/resources/blog/electrolytic-capacitor-lifetime-in-power-supplies
• https://www.badcaps.net/forum/troubleshooting-hardware-devices-and-electronics-theory/general-electronics-technical-discussion/12712-using-an-esr-meter-and-reference-chart
• https://hackaday.com/2019/04/12/ask-hackaday-experiences-with-capacitor-failure/
• https://www.instructables.com/Imac-G5-DIY-capacitors-repair/