Why the Stock Market Just Crashed — Transcript
Full transcript
- 0:00On Thursday, Nvidia lost more than four
- 0:03hundred billion dollars in a single morning,
- 0:05and it hadn't done anything wrong.
- 0:07Nobody had.
- 0:08No bank failed.
- 0:09No law passed.
- 0:11The economy looked fine.
- 0:12The S&P five hundred still lost more than two trillion dollars by midmorning.
- 0:17And the Nasdaq needed just six minutes to give back a rally that was one day old.
- 0:23So why did the stock market just crash?
- 0:25And was this the AI bubble popping?
- 0:29Best day after a midterm since nineteen eighty-two!
- 0:32Does this come with a receipt?
- 0:33The short answer is an election, six hundred thousand chips stuck in Hong Kong,
- 0:38and some very rich people locking in a very good year.
- 0:42The long answer starts a month earlier, on the sixth of October, in Santa Clara,
- 0:47California.
- 0:50That day, Nvidia closed at the highest price it had ever had.
- 0:54A month before, it was worth about five point four trillion dollars,
- 0:58roughly eight cents of every dollar in the whole S&P five hundred.
- 1:03If you own a plain index fund, the safe, boring kind everyone recommends,
- 1:08you own a lot of Nvidia.
- 1:10Add the rest of the Magnificent Seven, Apple, Microsoft, Amazon, Alphabet,
- 1:15Meta and Tesla,
- 1:16and about thirty-five dollars of every
- 1:19hundred in that fund sits on seven companies.
- 1:22I'm diversified.
- 1:23Five hundred companies.
- 1:25Seven of them are a third of the pie.
- 1:28...That is a big slice.
- 1:30Even at the dot-com peak, the market wasn't this lopsided.
- 1:34And most of those giants are betting on the same thing.
- 1:37That bet is expensive.
- 1:39This year,
- 1:40the four biggest cloud companies plan to spend over seven hundred billion dollars
- 1:45building data centers.
- 1:46Next year, more than a trillion.
- 1:49The biggest line on that shopping list says Nvidia,
- 1:52whose last quarter brought in ninety-six billion dollars,
- 1:55more than double the year before.
- 1:59By late October, investors were staring at four cracks.
- 2:03Crack one: the money goes in a circle.
- 2:06Nvidia invests in its own customers,
- 2:08the AI labs and chip-rental firms that then buy Nvidia chips.
- 2:12It owns a stake in CoreWeave, which rents those chips out,
- 2:16and agreed to buy up to six point three billion dollars of the computing power
- 2:21CoreWeave can't sell.
- 2:22Nvidia's own finance chief knows how that sounds: "we know some will call this
- 2:27circular financing."
- 2:29Here's a few billion dollars.
- 2:31Wow!
- 2:32What should I spend it on?
- 2:34Funny you should ask.
- 2:35Crack two: the chips are the collateral.
- 2:38Firms like CoreWeave borrowed huge sums to buy chips, rented them out,
- 2:42and pledged those same chips to borrow more.
- 2:45At one point CoreWeave owed over five dollars for every dollar of equity.
- 2:50Fine, while a used chip holds its value.
- 2:53But Nvidia brings out a new generation every year or two,
- 2:56and each one makes the last one worth less.
- 2:59And Michael Burry, who called the housing crash,
- 3:02says the giants book their servers as lasting longer than they do,
- 3:06which flatters profits.
- 3:07The nightmare: one borrower misses one
- 3:10payment, the lender grabs the chips to sell,
- 3:12and everyone learns that afternoon what a used AI chip is worth in a hurry.
- 3:17It hasn't happened.
- 3:18But every chip investor has thought about it.
- 3:21Crack three: who's paying.
- 3:23By the end of July,
- 3:24Nvidia was owed about sixty-three billion dollars for chips it had shipped but not
- 3:29been paid for yet.
- 3:30And this year, just three buyers made up more than two fifths of its sales.
- 3:35Picture a restaurant where three tables order two of every five plates,
- 3:40and a lot of the food goes on a tab.
- 3:42Crack four: the plug.
- 3:44You can buy a chip.
- 3:45You can't buy electricity that doesn't exist.
- 3:48In northern Virginia, the data-center capital of the world,
- 3:51the wait to hook a new building up to the grid had stretched to about seven years.
- 3:56And the power that does exist, somebody pays for.
- 3:59On America's biggest grid, across thirteen states,
- 4:03power plants get paid just to stand ready, and families cover that in their bills.
- 4:08Then the data centers piled in.
- 4:10Households there were paying an estimated twenty-five to thirty dollars a month more
- 4:15than in twenty twenty-four, and the grid's own watchdog blamed data centers first.
- 4:21Why is the electric bill up again?
- 4:23Hi, neighbor!
- 4:24Nobody reads a spending report.
- 4:27Everybody reads their power bill.
- 4:28Hold that thought.
- 4:29Wall Street had its own bad news.
- 4:32On October twenty-ninth, Google had its worst trading day ever.
- 4:36The next day Intel fell by a quarter, chip stocks slid into a bear market,
- 4:41and Apple, the boring one, passed Nvidia in value.
- 4:44By the second of November,
- 4:46the forty-seven biggest AI and chip stocks had lost about five trillion dollars
- 4:51since the top.
- 4:52That's about what the whole dot-com crash wiped out,
- 4:55and that took two and a half years.
- 4:57This took four weeks.
- 4:59Then came the election.
- 5:02Tuesday, the third of November.
- 5:05Democrats won the House.
- 5:07Then they won the Senate.
- 5:08Further down the ballot, in ten states,
- 5:10there were forty-one local votes on data centers.
- 5:14Voters went against them in forty.
- 5:16Twenty-one of those were in Ohio.
- 5:18The one town that said yes was Harwood, North Dakota.
- 5:21Ohio, no.
- 5:23Ohio, no.
- 5:25Ohio, no.
- 5:26Ohio...
- 5:27Hi!
- 5:28Yes!
- 5:29Over here!
- 5:30That's the power bill, showing up at the ballot box.
- 5:32In Ohio, Sherrod Brown beat Jon Husted,
- 5:36who'd been called "the face of data centers." In Texas, Elon Musk,
- 5:40the cycle's biggest donor with at least four hundred and seventy-four million
- 5:44dollars, lost both big races.
- 5:47His reaction:
- 5:48Data centers in space looking pretty appealing right now.
- 5:52Here's why a town vote moves Wall Street.
- 5:54A data center needs land, power and a permit, and all three are local.
- 6:00Goldman Sachs figured moratoriums like
- 6:02these could hit about forty-five percent of
- 6:05America's planned data-center expansion.
- 6:07And the bigger fear was Washington.
- 6:09For years, AI was mostly left alone to build.
- 6:13Now a Democratic Congress, elected the same night voters said no to data centers,
- 6:18might want rules on where data centers go, who pays for their power,
- 6:22and how fast the build-out runs.
- 6:24One congressman, Robert Garcia, had two words for AI executives: "Lawyer up."
- 6:32So.
- 6:33Data centers.
- 6:35We were just leaving.
- 6:36No rule has passed.
- 6:38Not one.
- 6:39It was the fear of a rule that did the selling.
- 6:41That trillion-dollar shopping list had picked up some small print: if they let us
- 6:46build it.
- 6:49And on Wednesday, Wall Street partied anyway.
- 6:52The Dow jumped fifteen hundred and sixty-nine points, about three percent,
- 6:56its best day after a midterm since nineteen eighty-two,
- 7:00and oil fell more than nine percent.
- 7:02Traders loved the cheap oil.
- 7:04They hadn't gotten to the new Congress yet.
- 7:06The party lasted one day.
- 7:08That morning, Vice President JD Vance had posted on X.
- 7:13The reason we lost is the war.
- 7:15The war with Iran.
- 7:17That raised hopes of a deal, and oil fell.
- 7:20By Thursday morning, traders doubted it, and oil climbed again.
- 7:24Then, before the opening bell, Reuters dropped the next problem.
- 7:28About six hundred thousand of Nvidia's H200 chips were sitting in Hong Kong,
- 7:33waiting for Beijing to let them in.
- 7:35Washington had cleared the sales last winter, case by case,
- 7:39with a twenty-five percent tariff.
- 7:41In August, Beijing let companies use them in Hong Kong.
- 7:44Into mainland China, in bulk?
- 7:46Still waiting.
- 7:47You said we can sell them!
- 7:50You can.
- 7:50Twenty-five percent, please.
- 7:52They're in Hong Kong!
- 7:53Can they come in?
- 7:56Not yet.
- 7:57The strange part: Nvidia had already taken a four hundred million dollar hit on
- 8:02H200s that nobody bought.
- 8:04And the ones that did have buyers couldn't get through the door.
- 8:09Nine thirty.
- 8:10The bell rings in New York.
- 8:12It takes the Nasdaq six minutes to give back the whole rally.
- 8:16So much for Wednesday.
- 8:17Ten thirty-five.
- 8:18The Nasdaq is down four percent and the Dow seven hundred points.
- 8:23Across the S&P five hundred, more than two trillion dollars has vanished,
- 8:27and more than four hundred billion of it is Nvidia alone.
- 8:31That's about twice Amazon's entire building budget for the year.
- 8:35Eleven seventeen.
- 8:36The VIX, Wall Street's fear gauge, is up sixty-five percent.
- 8:41Think of it as the price of crash insurance, and everybody wants some.
- 8:45Eleven thirty-seven.
- 8:46The Dow is on track for its worst second day after a midterm in records going back
- 8:52to eighteen ninety-six.
- 8:53One thirty-eight.
- 8:55AMD's boss, Lisa Su, speaks up, with her stock down fifteen percent.
- 9:00I think the market can sometimes be a little bit short-sighted.
- 9:05Aaaaah!
- 9:06Two fifty-eight.
- 9:07And this is where ordinary people come in.
- 9:10Maybe you're at work, checking your phone under the desk.
- 9:13You own a plain S&P five hundred fund.
- 9:15If it holds a quarter of a million dollars,
- 9:18Nvidia alone took about nine hundred and ninety dollars out of it that morning.
- 9:23By now, small investors have sold, on net,
- 9:26three hundred and fifty-eight million dollars of stock,
- 9:29the most since March twenty twenty.
- 9:31Buy the dip...
- 9:34...or sell the dip?
- 9:35Viraj Patel at Vanda Research called it "a total capitulation by retail investors,"
- 9:41with "no signs of buying the dip."
- 9:44Three oh six.
- 9:45Charlie McElligott, a strategist at Nomura, explains who's really selling.
- 9:50They're shooting the generals.
- 9:52The generals are the year's biggest winners.
- 9:54Hedge funds had made a fortune in twenty twenty-six, and with two months to go,
- 9:59they sold the winners to lock it in.
- 10:01Computer-driven funds and funds on borrowed money were forced to sell too.
- 10:05And nobody stepped in to buy.
- 10:07Great year.
- 10:08Let's bank it.
- 10:09Sell the winners.
- 10:11Which ones?
- 10:13Yes.
- 10:13Single-name stock moves are absolutely BONKERS.
- 10:17Four o'clock.
- 10:18The closing bell.
- 10:19The Dow lost eighteen hundred and thirty-three points.
- 10:23The S&P fell over four percent.
- 10:25The Nasdaq fell five and a half, its biggest drop in nineteen months,
- 10:30and is now in a correction, at least ten percent below its high.
- 10:34Chip stocks had their worst day since March twenty twenty.
- 10:37Nvidia closed at a hundred and eighty-one dollars, down more than nine percent.
- 10:42And while everything burned, AppLovin rose almost thirty percent,
- 10:46and Goldman Sachs picked that day to name a hundred and nine new partners,
- 10:51its biggest class since twenty ten.
- 10:55Honestly?
- 10:56Great day to get promoted.
- 10:59It's easy to think a crash like this only hurts rich people.
- 11:03It doesn't.
- 11:04Around seventy million Americans save in a 401(k),
- 11:08mostly through the same index funds,
- 11:09and the pensions of teachers and firefighters hold them too.
- 11:13When the seven giants fall, they fall too.
- 11:15It goes deeper than stocks.
- 11:17Data centers run on borrowed money,
- 11:20much of it raised from insurers and pension funds.
- 11:22Pimco bought a big piece of the bonds for Meta's giant Louisiana data center with
- 11:27savers' money, and Washington has moved to open 401(k)s to this kind of lending.
- 11:32So your retirement may be quietly lending to a half-built building.
- 11:37And some get hit without owning a share: the electricians,
- 11:40some making two hundred and sixty thousand dollars a year on data-center jobs,
- 11:45and the diners that feed them.
- 11:47Thursday didn't stop that work.
- 11:49But if the spending stops, it lands there first.
- 11:52If you're twenty-five, Thursday was a sale.
- 11:55If you're sixty-two and retiring in the spring, it was a very long day.
- 12:02So who do you get angry at?
- 12:04Every crash wants a villain.
- 12:06Twenty oh eight had the banks.
- 12:07Thursday doesn't really have one.
- 12:10I just didn't want the bill.
- 12:12I just wanted to keep my good year.
- 12:14I just said not yet.
- 12:16I just sold chips!
- 12:18Every one of those choices made sense on its own.
- 12:21They just all landed on the same morning.
- 12:23So here's why the stock market crashed.
- 12:26A market that's one-third AI had been sliding for a month.
- 12:30Then, in three days, voters elected a Congress that might rein in data centers,
- 12:35Beijing left six hundred thousand chips at the door,
- 12:38and funds sitting on a great year cashed out.
- 12:41Nothing broke.
- 12:42The fear did it.
- 12:43It isn't new, either.
- 12:45In the eighteen-forties, Britain went mad for railways,
- 12:49and at the peak it was pouring about seven percent of its economy into track.
- 12:53Railway shares lost nearly two thirds of their value,
- 12:56and a third or more of the approved lines never got built.
- 13:00But many of the lines that did get built are still part of Britain's railway today.
- 13:04Here's the catch.
- 13:05A railway can wait for its trains.
- 13:08A chip sitting in a dark building just gets older.
- 13:11So, was Thursday the AI bubble popping?
- 13:14Not yet.
- 13:15Both tariff-crash days in April twenty twenty-five were bigger,
- 13:19and together erased a record six point six trillion dollars.
- 13:24Jobless claims were near their lowest since nineteen sixty-nine.
- 13:27And no rule had passed.
- 13:29Bernie Sanders called for a national moratorium on data centers.
- 13:33Americans don't want Big Tech oligarchs wrecking the environment...
- 13:38A call, not a bill.
- 13:39A bubble doesn't pop when prices fall.
- 13:42It pops when the money stops.
- 13:44On Thursday, the money hadn't stopped.
- 13:47Nvidia's sales were still doubling.
- 13:49The giants were still expected to spend more than a trillion dollars next year.
- 13:53But here's the part almost nobody noticed.
- 13:56A week before the crash, one giant blinked.
- 14:00Microsoft cut its spending plan from a
- 14:02hundred and seventy-five billion dollars to a
- 14:04hundred and seventy.
- 14:06That's less than three percent.
- 14:07Tiny.
- 14:08But it's a foot moving toward the brake.
- 14:11And on Tuesday,
- 14:12the people who pay the electric bill
- 14:13elected a Congress that might want a say in
- 14:16where the buildings go.
- 14:17So Thursday wasn't the pop.
- 14:19It was the day the market noticed there's a pin, and that voters,
- 14:23Washington and Beijing all have a hand on it.
- 14:26Nvidia is now about a quarter below its October record.
- 14:30Its next report is due on the seventeenth of November,
- 14:33and the six hundred thousand chips are still in Hong Kong.
- 14:36The rally died in six minutes.
- 14:39What happens next gets decided much more slowly, by permits, power lines,
- 14:44and an electric bill on a kitchen table in Ohio.
- 14:47On Thursday, Lisa Su said the market can be a little bit short-sighted.
- 14:52That afternoon, it might have been the furthest it had looked in months.
About this transcript
This page contains the full transcript of Why the Stock Market Just Crashed by Paper Billions, generated from the public captions YouTube serves with the video. The transcript has 2,360 words across 313 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.