Home
Вход
Регистрация
Loading...
Практика аудирования
Практика аудирования
/
Video
/
The Infographics Show
/
AI Bubble vs Dot Com Crash. History is REPEATING
AI Bubble vs Dot Com Crash. History is REPEATING
Выберите режим обучения:
Посмотреть субтитры
Выбрать слово
Переписать слово
Highlight:
3000 Oxford Words
4000 IELTS Words
5000 Oxford Words
3000 Common Words
1000 TOEIC Words
5000 TOEFL Words
Субтитры (266)
0:00
We’re told that AI is a brand new technology led by a generation of geniuses.
0:04
But what if it’s not new at all… what if we’ve seen this exact story before?
0:09
Because behind the hype, the same billionaire class that rode the
0:12
Dot-Com Bubble of 1999 is back, just under a different name. Money is pouring in early,
0:17
long before anyone knows where the peak really is, because no one wants to miss out.
0:22
And through this hype, one phrase keeps getting repeated like a mantra:
0:26
“This time, it’s different.” Except… it’s not.
0:30
And it’s you that will be left to pick up the tab.
0:33
The Dot Com bubble promised global connectivity. Instead, it drained $5 trillion out of the
0:38
NASDAQ between March 2000 and October 2002. Ordinary Americans bore the brunt of that.
0:44
Retirement accounts loaded up with internet stocks lost about 78% of their value over 2 and
0:49
a half years. A Vanguard study found that by the end of 2002, millions of 401(k) accounts had lost
0:56
at least 20% of their value. Heavily tech-exposed portfolios were hit even harder. Across the
1:01
country, tens of millions of American workers were left holding the bag. These were people who pulled
1:07
cash out of their homes to chase Pets.com and Webvan and got margin calls instead of returns.
1:13
Foreclosures followed, destroying lives all across the country, most of them never made the news.
1:18
The NASDAQ peaked at just over 5,000 on March 10, 2000. Then the crash happened and it plummeted
1:24
to around 1,000. Once you account for inflation, it didn't get back to the peak level until 2018.
1:29
That is 18 years of lost growth for the people who clung to the Dot Com bubble.
1:34
And now, we’re staring down the barrel of the same gun.
1:37
Big Tech spending on AI data centers and chips is now over $300 billion dollars a year. The
1:43
value piled on top of that spending sits in fewer hands than at any time since the dot-com years.
1:49
Most people assume this is a brand new cast of characters.
1:52
It's not.
1:53
The AI movement is framed as a fresh rebellion led by hoodie-wearing newcomers in San Francisco.
1:58
The people setting the pace today are mostly the same people who set the pace last time. Only they
2:03
have 25 more years of contacts, government access, and investors lined up behind them.
2:08
During the original mania,Reid Hoffman made his fortune through Paypal. Now, he is an early backer
2:14
and former board member of OpenAI. Vinod Khosla rode his Sun Microsystems stake through the late
2:19
1990s hardware wave. He followed Hoffman into OpenAI. Marc Andreessen built Netscape and took
2:26
it public at 24 years old in August 1995. That IPO is what most people see as the official starting
2:32
point of the dot-com era. Today, he runs the venture firm Andreessen Horowitz, which has
2:37
poured billions into the current crop of AI labs. What looks like a technological revolution may be
2:42
something closer to a very expensive piece of theater. And the same
2:45
fingerprints keep showing up at every stage.
2:48
When Big Tech promised the internet would erase distance forever, it felt like a defining moment
2:53
in history. Money poured into anything with “.com” in the name. Between 1995 and the March 2000 peak,
2:59
the NASDAQ exploded roughly 400%. Investors stopped asking whether companies made money.
3:05
Revenue barely mattered. Profit was considered outdated. The only thing Wall Street cared
3:10
about was speed. They wanted to grow and attract customers. They wanted to dominate the sector.
3:15
The business model and logistics could come later. Then came the crash.
3:19
Now the AI Explosion is reviving that same energy, just with smarter machines instead of websites.
3:25
Massive data centers are burning through electricity to train models that get more powerful
3:29
every month. And again, nearly all the money is flooding into a select group of companies.
3:34
The biggest winner so far is NVIDIA. Every serious AI company needs its chips. That demand pushed
3:39
NVIDIA’s valuation into territory that would have sounded insane just a few years ago. By 2024,
3:45
investors were throwing money at NVIDIA the same way they once piled into internet stocks before
3:51
the Dot-Com crash. It’s not quite as extreme as Cisco at the peak of the Dot-Com Bubble. But it’s
3:56
moving in a direction that feels very familiar. The cash this time is coming mostly from Big
4:01
Tech's own bank accounts, not solely venture capital. But it all ends up in the same place.
4:06
Every bubble sounds good while it's inflating, but which one is the ultimate
4:10
trap. To understand that, we need to look at various factors that shaped the bubbles.
4:15
In 1999, Cisco Systems owned 72% of the enterprise
4:19
routing and switching market. That means Cisco sold the physical boxes that made the internet
4:24
work. It was selling the backbone of the internet itself. Every company rushing online needed
4:29
Cisco’s hardware, and the money pouring in proved it. By fiscal year 2000, Cisco was generating
4:35
nearly $18.9 billion dollars in annual revenue. On March 27th, 2000, Cisco hit $80 a share.
4:42
Its market value surged past $555 billion. For a brief moment, Cisco became the most
4:48
valuable company on Earth, overtaking Microsoft. Investors weren’t just buying into a successful
4:54
company. At its peak, Cisco traded at a price-to-earnings, or P/E, ratio of 201.
4:59
Imagine paying $100 for a lemonade stand that only earns you $1.25 a year. The stand might be
5:06
incredible but the numbers border on fantasy. The whole thing rested on Venture capital
5:11
continuing to flow to the startups buying the routers. When that funding froze in the spring
5:15
of 2000, the orders dried up. Cisco couldn’t handle it.
5:19
The stock fell about 80% from its peak over the next 30 months. It took Cisco almost 26
5:26
to climb back to that $80 mark. The recovery hit in December 2025. Anyone who bought at the top and
5:31
held all the way still lost more than half of what their money could buy. Inflation ate the rest.
5:37
NVIDIA in 2024 looks eerily similar to Cisco at the peak of the Dot-Com era. Its chips are
5:42
shipped by the truckload. Data centers across the world are stuffing racks with NVIDIA GPUs
5:47
as fast as they can get them. The demand looks unstoppable.
5:51
But the reality is a lot more fragile.Many of NVIDIA’s biggest customers are AI labs
5:56
and startups burning through investor cash at historic speeds. The rest are tech giants
6:01
spending billions because they believe AI has to work, not because the profits already exist.
6:07
That’s the part that makes veteran investors nervous.
6:10
When analysts overlay NVIDIA’s 2024 valuation surge against Cisco’s climb before the 2000 crash,
6:16
the curves follow the same trajectory. When two bubbles separated by 25 years
6:21
begin drawing the same shape, people who lived through the first one tend to pay attention.
6:25
Most people assume NVIDIA is safe because, unlike the dot-com flameouts,
6:30
it has real hardware revenue. But Cisco had real hardware revenue and a dominant market share.
6:36
The 2000 crash didn't come because the routers stopped working. It came because
6:40
the people writing the checks ran out of money. Cisco's peak was actually sharper than anything
6:45
NVIDIA has touched so far. It should be a warning. The number tells you
6:49
how much further the current cycle could still inflate before the same demand cliff shows up.
6:54
So the machinery looks familiar. But the more revealing comparison is the people
6:58
making the decisions behind it.
7:00
During the late 1990s, executives at the biggest tech companies kept telling investors
7:04
the same story: the internet had changed everything. The old rules about profits,
7:09
and valuation no longer applied. Earnings would eventually catch up to the hype.
7:13
Meanwhile, behind the scenes, insiders were selling stock.
7:17
They were small sales. Just enough to avoid setting off alarms. At the time,
7:21
almost nobody paid attention. It only became suspicious years later,
7:25
after the bubble burst and someone looked closer. The numbers when they finally came out were ugly.
7:30
Between September 1999 and July 2000, dot-com insiders cashed out $43 billion
7:37
of their own company stock. That was twice the rate they had been selling at in 1997 and 1998.
7:44
February 2024 was a different animal. The camouflage came off.
7:49
In a single 9-day window that month, Jeff Bezos sold $8.5 billion of Amazon stock.
7:55
The Walton family trust dumped $1.5 billion of Walmart shares over the same stretch. Jamie Dimon,
8:01
the CEO of JPMorgan, sold $150 million of his own bank's stock. That was his first sale in 18 years
8:08
on the job. Leon Black, the Apollo co-founder, unloaded $172.8 million. His first sale ever.
8:15
The combined number for that one month came to $11 billion dollars.
8:20
But it didn’t stop there. Mark Zuckerberg offloaded roughly $2 billion
8:23
of Meta stock across the 4 months heading into that window. One at a time, the moves all looked
8:29
normal. They were nothing out of the ordinary. Stacked side by side, the people closest to the
8:34
numbers were cashing out at the same moment. The whole time, the public messaging from
8:38
those same executives stayed bullish. Belief in the project. Publicly, they talked about
8:43
decade-long opportunities and the future of AI. Privately, they were cashing out near the highs.
8:49
The interviews said confidence. The filings said take the money,
8:54
Fortune ran the headline "The Great Cash-Out" on February 27th, 2024. It was a fitting title.
9:00
When the people closest to the boom start taking money off the table, it usually means they
9:05
understand the risks better than everyone else. And unlike 1999, the selling is happening faster
9:10
and in larger amounts. The people building the boom increasingly look like people preparing
9:16
to survive the end of it. But if insiders are selling, who’s still buying enough stock to keep
9:21
prices floating at these levels?
9:23
In 1999, Webvan built refrigerated warehouses for customers who didn’t exist yet. Pets.com
9:29
made television commercials that turned out to be more memorable than its actual orders.
9:33
Both companies poured cash into buildings, trucks, and ad campaigns shaped around demand that never
9:38
showed up. Both became case studies in burning money because neither made it to its second
9:43
birthday on the public markets. Stability AI is the modern
9:46
version of these companies. In 2023, it spent roughly $99 million
9:51
renting compute power from AWS, Google Cloud, and CoreWeave. On top of that, another $54 million
9:57
went to salaries and running costs. Their total revenue for the year?
10:01
$11 million dollars. That’s a burn-to-revenue ratio north of 14 to 1.
10:06
By July 2023, Stability AI was already short on its AWS bill by $1 million. Internal reporting
10:13
later showed the company had no real plan to pay the $7 million August invoice either.
10:19
But the cash didn’t vanish into a black hole. It moved on a specific, traceable route. Venture
10:25
firms wired fresh capital into AI startups. The startups turned around and handed that capital
10:30
straight to NVIDIA for chips and to Microsoft Azure for cloud time. Big Tech then booked that
10:36
spend as their own revenue, pushing their stock prices higher. The higher stock prices justified
10:41
bigger venture commitments and the next round of money flowed back through the same pipe.
10:46
It’s what people inside the industry call the Circular Economy.
10:49
It might be the single most important trick in the current boom.
10:53
A dollar leaves a Silicon Valley account and lands in some AI startup's bank account.
10:58
But it doesn’t sit there for long. Within a few weeks, that same dollar usually shows up
11:02
on Jensen Huang’s earnings call as growth. It then helps push NVIDIA's stock higher. That
11:08
makes the next venture fund easier to raise. Then, another dollar gets sent through the same loop.
11:14
Most of the money isn’t coming from everyday customers buying AI tools because they can’t
11:19
live without them yet. Sure, companies like OpenAI have concrete revenue.
11:23
But a large part of the money doesn’t measure how many people actually use the products.
11:28
It is measuring the same pool of capital moving back and forth between 5 connected companies.
11:33
A good example is Inflection AI. In June 2023, it raised about $1.3 billion at a valuation of
11:40
roughly 4 billion. The investor list read like a who’s who of the AI boom: Microsoft, NVIDIA,
11:46
Bill Gates, Eric Schmidt, Reid Hoffman. Less than a year later, in March 2024,
11:51
Microsoft effectively absorbed the company. It paid around $650 million,
11:56
hired most of the team, and licensed the core technology.
11:59
Inflection, as a standalone business, was finished.
12:02
The investors, though, walked away with 1.5 times what they had put in. The cash
12:06
had already passed through NVIDIA's order book and Microsoft's cloud invoices on the way down.
12:11
The only people who lost out were the late buyers. The speed and the design of this cash loop go way
12:16
past anything the dot-com failures pulled off. Webvan was sloppy in a way the market eventually
12:22
figured out. What’s happening around AI feels different. It’s more coordinated.
12:27
It’s less of an accident and more of a system. So who benefits while it works and who is left
12:32
holding the losses when it stops?
12:34
(The Exit Liquidity) In 1999, day traders
12:36
opened online brokerage accounts for the first time and rushed into anything that was moving.
12:41
They were snapping up things like IPOs and internet stocks.
12:44
Many were buying on margin - borrowed money - so every rise felt amplified.
12:49
At the same time, the biggest institutions were backing off. But the market didn’t fall
12:54
immediately. It kept going, because there was still someone willing to buy at higher prices.
12:59
That someone was retail traders. Except they didn’t know that.
13:03
They just saw rising charts and didn’t want to miss out. Instead, they were absorbing the market.
13:08
The 2024 version is worse. Trading wasn’t just about
13:12
buying and holding stocks anymore. A huge share of activity was people making bets that expired
13:17
the very same day they were placed. Cboe Global Markets reported that this kind of
13:21
ultra-short trading became so common it was approaching half of all activity tied to the
13:26
S&P 500 options market on typical days. Even the 2021 meme-stock frenzy didn’t reach that level.
13:33
The market was being gamed in real time, minute by minute,
13:36
Robinhood spent 2023 and 2024 running television ads that pushed options trading
13:41
into the mainstream. Your cousin, your neighbor, the guy at the gym.
13:45
The platform was reporting more than 25.2 million funded accounts by the end of 2024. The user base
13:51
skewed heavily toward traders under 35 clearing more than 50 million contracts at peak times.
13:57
The favorite tool of the retail trader is no longer the stock itself.
14:01
It is a leveraged bet. A bet that the price will go up or down by closing time the same afternoon.
14:08
Most people assume the average investor in 2024 is just like the day trader from 1999,
14:14
just with a slicker app. But the truth is, it's not even close. Imagine a stadium full of
14:19
people betting their life savings on a single coin toss every hour. Then, they make another
14:24
bet before the previous coin has even hit the floor. That is roughly the speed of same-day
14:29
options trading in the current cycle. The public isn’t acting like a slow,
14:32
steady pool of long-term buyers anymore. It’s acting like a fast-moving crowd,
14:36
stepping in and out so quickly that it can absorb selling without even realizing it’s doing so.
14:41
That changes the whole system. In the late ’90s, retail was
14:45
loud but relatively simple. Today it moves faster and reacts instantly to price swings.
14:51
That means it can absorb a surprising amount of selling without the market immediately breaking.
14:56
So when early winners and insiders sell now, they don’t need a dramatic exit window. There’s already
15:02
a constant churn of buyers underneath them, stepping in and out quickly enough to take the
15:06
other side without noticing it in real time. But what happens if that flow of
15:11
buyers suddenly slows down?
15:13
In the late 90s, Big Tech was at war. Microsoft spent much of the decade locked
15:18
in an antitrust battle with the U.S. government. The fight was over its decision to bundle Internet
15:22
Explorer with Windows. The broader industry treated Washington as a problem to manage,
15:27
not a partner. Lobbying budgets existed mostly to keep federal hands off the fortunes being made.
15:33
By 2024, the stance had completely flipped. OpenAI's federal lobbying spend jumped from
15:39
$260,000 dollars in 2023 to $1.76 million in 2024. That’s close to a 7-fold rise in a single year.
15:48
Anthropic more than doubled its own spend over the same window. From $280,000 to $720,000.
15:54
According to OpenSecrets, 648 different companies spent money lobbying on AI issues in 2024.
16:01
It was a 41.5% jump from the previous year. The stated reason, in almost every case, is
16:07
responsible rollout. The effect, whether intended or not, is that the earliest and largest players
16:13
end up behind a kind of protective barrier. One that makes it harder for new
16:17
entrants to compete on equal terms. The clearest moment of all came in May 2023.
16:23
Sam Altman appeared before the Senate Judiciary Committee. He personally asked Congress to
16:27
license AI companies. The CEO of the leading AI firm was asking the United States government to
16:34
require permission slips to build advanced AI. That request lands very differently the
16:39
moment you ask who would qualify for one of those permission slips. And who would not.
16:44
Smaller companies don’t really get a seat at the table when these rules are being shaped.
16:48
None of them have the legal teams or the compliance budgets to fight back.
16:52
The rules are written around the needs of a company worth half a trillion dollars.
16:56
That is the whole point. The lobbying spend isn't an operating cost. It is the
17:01
price of permanently killing the competition. The framing dresses a protection racket up in
17:06
policy language. The big players pay the lobbyists. They help draft the rules.
17:11
They lock the door behind them and tell the public it’s for their own safety.
17:15
The same play is running in Europe, just with different paperwork.
17:18
The EU AI Act passed into law in March 2024 and started rolling out in 2025. A lot of the
17:25
strictest compliance requirements land hardest on smaller open-source developers and academic
17:30
groups. Meanwhile, the biggest US companies already have entire teams for exactly this
17:35
kind of thing. Mistral AI has become the clearest European challenger in this space, and it’s spent
17:41
a lot of time trying to influence how stricter rules apply to open models, with limited success.
17:46
The pattern is consistent on both sides of the Atlantic.
17:49
Once the rules become law, the story changes. Companies don’t need to keep selling the idea
17:54
of endless disruption at the same intensity. The system itself starts to lock in who can scale and
17:59
who can’t. Competition doesn’t disappear, but it becomes slower and more controlled.
18:04
That takes pressure off the narrative that everything has to grow forever.
18:08
What’s different this time is how intentional it feels. You can already
18:12
see pieces of the next regulatory framework sitting in draft form through 2025 and 2026,
18:18
waiting for the right political moment to move. The trap is built. The only
18:22
question left is when it springs.
18:24
So who actually wins when both booms run their course?
18:28
It isn’t the customers. They get cheaper tools, but not the upside. It isn’t the small investors,
18:33
who tend to arrive after most of the gains are already priced in. And it isn’t always
18:38
the companies in the headlines either. Many of them spend the peak years trying to justify
18:43
valuations that only make sense in the moment. The real winner is the system around the industry.
18:49
The mix of capital, infrastructure, and policy that doesn’t just take part in the cycle,
18:54
but shapes how it unfolds. The same forces that helped build the first wave didn’t disappear
18:59
after it ended. They adapted and scaled up. They’re now operating inside a second,
19:03
larger version of the same pattern. What’s changed is the scale and tolerance
19:08
for complexity. The buildout is bigger and the money is deeper. That doesn’t make the outcome
19:12
predetermined. But it does mean the system can absorb more stress before it breaks,
19:18
and keep running longer while it does. Most analysts can see what is happening.
19:22
The AI drawdown probably won't begin because the technology fails. The models are getting better.
19:28
The hallucination rates are dropping. But none of that is the trigger.
19:32
The trigger is the moment the rules get signed into federal law.
19:36
Once competition is legally locked out, the big players have permission to change stance. They
19:41
stop chasing growth and start chasing efficiency. That means mass layoffs.
19:46
Microsoft, Meta, and Google all announced cuts in the tens of thousands across 2024 and 2025.
19:52
That is a preview of the broader pattern. The story shifts from "spend whatever it takes"
19:57
to "responsible capital return." That’s when stock valuations drop. The architects keep
20:02
the cash they pulled out at the top. They walk out with a locked-in market
20:06
share and federal protection written into law. British investor Jeremy Grantham called both the
20:11
2000 and 2008 bubbles in advance. He’s been tracking this exact pattern for decades and
20:17
he doesn't sugarcoat any of it. Bubbles this size resolve through long, deep drawdowns measured in
20:23
years, not months. Cisco needed almost 26 years to climb back to its peak. That is the base rate for
20:30
the biggest stock at the top of a peaked bubble. The history books do not have a V-shaped recovery
20:35
on file for an unwind this dense. It doesn’t really look like
20:39
a broken system when you step back. It’s a system doing exactly what it evolved to do.
20:44
Money flows in from millions of ordinary accounts over long periods of time. It gets pooled and
20:50
concentrated into a small number of huge companies that dominate the market. The people who got in
20:55
early take money out along the way. The people who arrive later mostly ride whatever price is left.
21:01
And almost everyone is in it, whether they realize it or not.
21:05
Because retirement savings aren’t sitting on the sidelines anymore. They’re already inside the
21:10
same trade. They’re tied to the same handful of companies, exposed to the same outcomes.
21:15
Because the people running these cycles have been through this before. Most of the public hasn’t.
21:20
Or they were too young to remember what it actually felt like while it was happening.
21:24
That’s what makes bubbles so effective. By the time something feels obvious,
21:28
it already feels normal. And when the mood finally turns,
21:31
most people are still holding the same belief that existed at the peak of the dot-com era: that
21:36
this time the story is too important to slow down. But it’s not just the AI bubble that could trigger
21:42
a financial meltdown. Find out why analysts believe the next crisis may already be building
21:47
in “39 Trillion Dollars in Debt: Is the American Economy Doomed?” Or watch this instead.
