market microstructure

The Flash Crash: How $1 Trillion Vanished in 36 Minutes

June 21, 2026 · case-study · Archi·Finance

The day the market vanished

-998 pts~36 minutes~$1 trillion

May 6th, 2010. It's 2:32 in the afternoon, and the U.S. stock market is about to do something nobody has ever seen. In the next few minutes, the Dow falls almost a thousand points. Around one trillion dollars in value just evaporates. Some of the biggest companies in America trade for a single penny. And then, about 36 minutes later, most of it comes back, like it never happened. This is the flash crash. No bank failed. No war broke out. Nothing in the real world actually changed. The machinery underneath the market simply ran out of something it needs to work, and the whole thing buckled. So let's walk through it, almost second by second, and see how a market can disappear into thin air.

A thin market, and one giant order

$4.1B order75,000 contracts9% of volume

Start with the day itself. Markets were already jumpy over Greece's debt, prices drifting lower, and liquidity was thin. Thin liquidity just means few buyers and sellers are standing by, so any big order shoves the price further than normal. Into that nervous market, one firm pressed go. A large mutual-fund manager, Waddell and Reed, told an automated program to sell 75,000 E-mini futures contracts, a bet on the S and P 500 worth about $4.1 billion, to hedge its other positions. Here's the catch. The program was set to sell 9 percent of recent trading volume, with no regard for price, and no regard for time. A block that size would normally be fed out over five hours or more. The algorithm pushed the whole thing through in about 20 minutes.

The hot-potato feedback loop

millionths of a secvolume misreadno real bids

Now watch what the speed does. The contracts get snapped up by high-frequency trading firms, the machines that buy and sell in millionths of a second. But they don't want to hold this stuff. So they flip it to each other, over and over, the same contracts changing hands like a hot potato. And here's the trap. The selling program was watching trading volume to decide how fast to go. All that frantic flipping looked like real volume, so the algorithm sped up, pouring even more into a market with almost no genuine buyers left. The bids sitting under the price just vanished. Normally, lower prices pull in bargain hunters. That afternoon, the buyers backed away instead, the floor gave out, and the selling started feeding on itself.

When stocks traded for a penny

8 stocks to 1 centSotheby's $99,999.99P&G -37%

The futures market is wired straight into regular stocks and exchange-traded funds, so the panic jumped tracks. And individual stocks did something genuinely insane. Accenture, a giant consulting firm, had closed the day before at $42.17. At 2:47 and 53 seconds, 100 of its shares traded for a single penny. Eight big companies in the S and P 500 briefly hit one cent. Others flew the opposite way. Sotheby's printed at $99,999.99 a share. Procter and Gamble, a household name, fell about 37 percent in minutes. Why? When real buyers disappear, market makers leave behind what are called stub quotes, placeholder offers at absurd prices like a penny, never meant to actually trade. With nothing else left in the book, the market reached down and hit them.

The snap-back, and the invisible 60% line

~13 min recovery>60% trades cancelled36 min total

The strangest part is how fast it ended. The deepest plunge hit around 2:47. By roughly 3:00, just over ten minutes later, those same stocks had snapped back to near where they started. The machines that pulled out came back, saw the ridiculous prices, and bought. But that left a mess. Thousands of trades had gone off at crazy levels, so the exchanges went back and cancelled them. Only, they cancelled just the ones that printed more than 60 percent away from the pre-crash price. If your trade was 59 percent off, it stood. If it was 61 percent off, it disappeared. That line was invisible to traders in the moment, and it quietly decided who kept a windfall and who got erased. Thirty-six minutes, start to finish.

The trader in the bedroom

$200M fake orders~19,000 cancelscharged 2015

So who, or what, was to blame? The first reports pointed at that one big sell order, but the fuller answer took years. In 2015, U.S. prosecutors charged a futures trader named Navinder Sarao, working from his parents' house near London. For years he had been running a trick called spoofing. He would place huge sell orders he never intended to fill, around $200 million of fake selling pressure, then cancel them before they could execute. On the morning of the crash, his spoof orders were at times more than 20 percent of all the visible sell orders in that market. He didn't cause the crash by himself. But he helped tilt an already fragile market, and his orders were modified or pulled roughly 19,000 times. One man, a laptop, and a market that trusted its own order book a little too much.

What got fixed, and what didn't

circuit breakersLimit Up-Limit DownDodd-Frank

The flash crash rewired the market's safety systems. Regulators rolled out circuit breakers that pause trading in a stock when it moves too far, too fast. The version we use now is called Limit Up, Limit Down, and it puts guardrails around how far a price can run in a short window before trading slows. They banned the worst stub quotes, so market makers have to post real prices near the going rate. And spoofing became an explicit crime under the Dodd-Frank law. But the deeper lesson hasn't changed. Modern markets run on liquidity supplied mostly by fast machines, and those machines have no obligation to stay. When they step back at the same moment, the floor can drop out in seconds. The flash crash wasn't really a glitch. It was the system working exactly as built, just faster than any human could stop it.

Sources

Watch the full breakdown and subscribe for one new explainer at a time.

Watch on YouTube

flash crash2010 flash crashstock market crashhigh frequency tradingHFTmarket microstructurecircuit breakerslimit up limit downNavinder SaraospoofingE-mini S&P 500Waddell and Reedmarket liquidityalgorithmic tradinghow the stock market works