Home
लॉगिन
रजिस्टर
सीखने की सामग्री
Loading...
The Limitless AI Lie. The Bubble Is Slowly BURSTING. - Video học tiếng Anh
श्रवण अभ्यास
श्रवण अभ्यास
/
Video
/
The Infographics Show
/
The Limitless AI Lie. The Bubble Is Slowly BURSTING.
The Limitless AI Lie. The Bubble Is Slowly BURSTING.
सीखने का मोड चुनें:
उपशीर्षक देखें
शब्द चुनें
शब्द को फिर से लिखें
Highlight:
3000 Oxford Words
4000 IELTS Words
5000 Oxford Words
3000 Common Words
1000 TOEIC Words
5000 TOEFL Words
उपशीर्षक (246)
0:00
You’ve probably heard the AI bubble is popping. That it’s all hype… a
0:03
gimmick… and the money is drying up. That’s not what’s happening. What’s
0:07
actually going on is a lot more serious… and almost nobody is talking about it.
0:12
The AI revolution didn’t hit a financial wall. It hit a physical one.
0:17
Right now, roughly 50% of American data center projects in major hubs are delayed or quietly
0:22
canceled. Not because companies ran out of cash. Because they ran out of power.
0:28
Now, Big Tech isn’t just making software anymore. They've become
0:31
some of the most aggressive investors in one industry you wouldn’t expect…
0:36
Nuclear power. This is the Limitless AI Lie.
0:40
Chapter 1 - The Ghost Clusters We are currently in the midst of
0:43
a global artificial intelligence arms race and it’s getting very expensive.
0:48
In 2025, The Wall Street Journal reported that just in the United States, the Big Four – Amazon,
0:53
Google, Meta, Microsoft – were collectively spending close to $400 billion annually,
0:59
with most of that money going on infrastructure to run giant AI data centers.
1:04
These companies are taking on heavy debt, but as The Journal pointed out, the “current revenue”
1:08
from AI was, and still is, “relatively tiny.” Even so, the reports of unprecedented
1:14
spending just kept coming. There’s a reason. And it’s why OpenAI,
1:18
Oracle, and SoftBank announced they would spend just under half a trillion dollars on
1:22
the Stargate initiative. They’re rolling out five new data centers in Texas, New Mexico,
1:28
Ohio and the Midwest. CEO Sam Altman promised that the investment would lead to future
1:33
breakthroughs in technology, adding that they are already making, quote, “historic progress.”
1:38
Big Tech and every other global enterprise with a focus on AI are expected to spend
1:42
$7 trillion on building or updating data centers in the next 5 years.
1:48
That’s the plan, but reality shows a very different picture.
1:51
Data centers aren’t easy to run. Of the almost 11,000 active ones in
1:56
the world in 2026, the US had the lion's share at around 4,000, well ahead of the UK (511),
2:03
Germany (507), China (368), and France (344). But just to run the ones we already have is
2:11
extraordinarily energy-intensive. Cooling them can take vast amounts of water, with the biggest
2:16
hyperscale sites using millions of gallons a day. Still, heat can be managed.
2:22
The problem is powering them. Reports in 2026 stated that 50%
2:27
of data centers being constructed in the US are currently either canceled or delayed,
2:32
with experts saying that number is going to rise. In major hubs like Northern Virginia – the region
2:37
known as “Data Center Alley” – projects are already slowing as the grid starts to buckle
2:42
under demand. The public sentiment has changed. Polls show just 35%
2:46
of Virginia voters give their thumbs up to the power-hungry projects compared to 69% in 2023.
2:53
The delays aren’t about a lack of money or ambition. Silicon Valley has plenty
2:57
of both. They’re about supply chains, engineering limits, and public/political
3:02
resistance…but most of all, power. The future Sam Altman talks about?
3:07
It’s sitting in a queue… waiting for electricity. Right now, America is facing a bottleneck caused
3:12
by unprecedented demand. So many of the data center projects are
3:16
in a traffic jam, waiting to connect. Before anything can plug in, utilities
3:20
and grid operators must run studies to see whether lines, substations and transformers can cope with
3:26
the extra load. They have to assess reliability. They have to look at the environmental impact.
3:31
This can cost millions of dollars and lead to years of waiting around.
3:35
It doesn’t just affect Big Tech. It affects you.
3:38
These massive data centers burn through enormous amounts of electricity. When their energy bills
3:43
spike, that cost doesn’t just disappear. It gets passed on. To consumers. To communities.
3:49
And when people start seeing higher energy prices, they push back. That can be lethal
3:54
to the ambitions of anyone wanting to build a data center. And even if the projects get
3:58
past all those hurdles, the physical system might still not be ready to go.
4:03
According to Lawrence Berkeley National Laboratory, these queues add up to about
4:07
2,600 gigawatts of proposed generation and storage capacity. That’s about twice
4:12
the size of the current U.S. power system. It’s a backlog as big as the grid itself.
4:18
The amount of power these new data centers will require to keep running is astronomical.
4:22
Some of the new ones in the US will use 50 to 100 times more power than it takes to power the
4:28
Empire State Building… just to run and cool the servers. Reports say the biggest ones currently
4:33
under construction will collectively demand as much electricity as an entire country like Italy.
4:39
Big tech wants to spend trillions, but the “power crunch” is getting in the way.
4:43
It’s a problem they can’t code around. Chapter 2 - The Physics of the Bubble
4:48
On average, a single query to ChatGPT uses around 10 times more energy than a search
4:53
on Google. To put another way, one complex prompt can use about as much electricity
4:58
as leaving an LED lightbulb on20 minutes. The exact number is debated. And not every
5:04
query is the same. But the scale is. Because ChatGPT isn’t handling thousands of requests…
5:10
It’s handling billions. 2.5 billion queries per day,
5:14
it all adds up. The US alone experienced about 330 million ChatGPT queries per day in 2025
5:21
That’s just the beginning. Recent reports tell us about
5:24
1.1 billion people worldwide use AI but that number is growing fast. 56% of Americans use it
5:31
in their day to day lives. About 60% of American teachers have reported experimenting with it.
5:36
88% of global firms are now onboard, too. The experts say there are still about 5
5:41
billion people globally who don’t use it, but the adoption curve is rising fast. In
5:46
the next 5 or 10 years, we can expect many of those 5 billion will be regular users.
5:50
So as more and more people start experimenting with it,
5:53
the world will need to generate a lot more power. Goldman Sachs Research analysts have estimated
5:58
that global data center power demand will grow 160% by 2030. That’s about the equivalent of
6:04
adding a second 'Germany' to the global energy grid in just 6 years. Data centers already consume
6:10
about 1-2% of all the world’s electricity. If the experts are right, it will be more
6:15
like 3-4% by the time the decade is over. Where will all this new power come from?
6:21
In 2023 alone, new applications for interconnections surged by about 30%. The data
6:26
showed that close to 95% of requested capacity came from solar, wind and battery storage.
6:33
But here’s the problem. Renewables aren’t constant.
6:36
The wind slows down. The sun disappears. Even in the best solar regions in the U.S., you only get
6:42
about 3,500 to 4,000 hours of usable sunlight a year. That's out of almost 8,800 hours in total.
6:49
Less than half the time. In other words, solar and wind
6:52
won’t be enough to power the AI revolution. Companies are now fighting for whatever power
6:57
they can get, and whoever controls the power, controls the future of AI.
7:01
Chapter 3 - The Great AI Cleansing The delays in data centers look like
7:06
a failure of infrastructure. But they may also be creating something else: scarcity.
7:10
And scarcity, for the companies with all the money, isn’t always a problem.
7:14
It can actually be a very real advantage. The biggest tech firms are already making
7:19
deals with energy companies. Meta, Google, Amazon, Microsoft, have inked
7:23
deals with utilities firms all over the USA. They’ll get the power. The smaller players won’t.
7:29
Venture-backed startups might be able to raise money and promise major market disruption,
7:33
but they won’t be able to compete for energy. They’ll have to rely on
7:37
Big Tech for renting and compute power. That could mean that every GPU shortage,
7:42
every power delay, every price rise will hit them first. And the state-of-the-art models
7:47
might become prohibitively expensive or limited in access. The biggest tech companies will
7:52
decide who gets access first, who gets the most computing power, and who gets the best versions.
7:57
Meanwhile the market itself is changing. Investors are losing interest in generic chatbots that
8:02
write poems and hallucinate history. Capital is moving toward specialized AI that can write code,
8:07
optimize logistics, detect fraud, perform medical analysis, and above all, increase profit margins.
8:13
But not everyone will receive the same treatment, and some might be left waiting. Countries have
8:18
anticompetition laws, but the gatekeepers of technology as they consolidate their power over
8:23
AI, might move faster than the law can keep up. A specialized startup might not ever truly be able
8:28
to compete, not if it's denied access. In the gold rush, fortunes weren’t made
8:33
by the miners, but by those who controlled the tools and the land.
8:37
This time, it isn’t just the tools. It’s the electricity that powers them.
8:41
For years now, people have been worrying about the rise of the machines, rogue AI ending the world,
8:46
but that’s all science fiction. The real danger is a handful of monopoly
8:50
powers who control all the key infrastructure. But to guarantee their monopoly, Big Tech knows
8:55
it can’t just rely on public infrastructure. So it’s planning to go off-grid.
9:00
Chapter 4 - The Nuclear Pivot For decades, tech companies rented power from
9:05
the grid like everyone else. Now, that’s changing. These days some of the largest firms with
9:10
seemingly bottomless reserves of money are getting into the
9:13
business of procuring their own energy assets. They’re not just software companies anymore.
9:18
In 2024, Microsoft made a deal to reactivate Three Mile Island nuclear power plant, a site forever
9:24
associated with a 1979 meltdown. One reactor was shut down permanently. The other ran until 2019,
9:30
when it was closed due to operating losses. Now, decades later, it’s being revived.
9:35
Not for history, but for demand. Under a 20-year power agreement,
9:38
Microsoft is securing 835 megawatts of electricity for Pennsylvania’s grid.
9:44
The deal wasn’t about powering Pennsylvanians’ personal computers or domestic appliances, but
9:49
achieving lift-off on its planned data centers. It’s enough electricity to power about 800,000
9:54
American homes, and all of it just for Microsoft's AI.
9:58
But will it happen? Well, firstly,
10:00
a nuclear plant has never in American history been fully shut down and sprung back to life.
10:06
Experts are saying that for Microsoft’s plan to come to fruition, there will be numerous
10:10
regulatory hurdles to cross. The permits are all still pending, and they might not ever come,
10:15
even if the plant owner, Constellation Energy, is trying to speed things up.
10:19
Nuclear power tends to make people uneasy. Even if the plant did get a second lease of life, it’s not
10:25
clear it could be integrated back into the grid. Mark Zuckerberg’s Meta also hit a problem when
10:30
trying to build a data center next to what was reported as an “existing nuclear facility.”
10:34
The media didn’t name the facility but it did report that Zuckerberg’s plan of operating a
10:39
nuclear-powered AI center crashed to the ground because of…bees. A rare species populated the
10:46
area where Meta wanted to build. It goes to show how Big Tech is
10:50
thinking outside of the box. Amazon has tried, too.
10:52
In 2024, the company signed a $650 million deal with Talen Energy in Pennsylvania where
10:58
it plans to open a number of data centers close to the Susquehanna nuclear power plant. The deal,
11:03
which is set to end in 2042, could be delivering almost double the
11:07
power needed for the city of Pittsburgh. That’s if it ever gets off the ground.
11:12
In 2024, the Federal Energy Regulatory Commission blocked Amazon from taking
11:16
power directly from a nuclear plant to feed one of its hyperscale data centers.
11:20
The reason? Diverting that
11:22
much power raised the risk of blackouts for the communities already relying on that same
11:27
nuclear supply.And even if the lights stayed on, there was another consequence… higher bills.
11:33
The regulators are standing firm. Amazon later asked for a re-hearing
11:37
and the result stayed the same. Google is also looking to nuclear
11:40
energy to power its future AI ambitions. It’s one of a few big tech firms currently
11:45
investing in next-generation nuclear energy in the form of small modular reactors (SMRs).
11:51
Smaller means less cost to build. It also means cooler, so ideally, less chance of a meltdown.
11:57
The company hopes to have the data center up and running by 2030. But long before any green lights
12:02
start flashing, Google will need approval from the U.S. Nuclear Regulatory Commission (NRC).
12:08
And that won’t come easy. As one analyst wrote,
12:10
the “path to commercial viability is still riddled with uncertainty.” SMR designs haven’t been proven
12:15
at scale and regulatory hurdles have set a high bar. With SMRs, there’s no easy path forward.
12:21
But there’s political will, and that’s important. The current administration has given the thumbs
12:26
up and has said it wants these small reactors to be functioning by 2030.
12:30
None of the new commercial SMRs are yet operating in the US, but China and Russia already have
12:36
working examples. In terms of the global AI race, this isn’t good news for Uncle Sam.
12:41
But even if the tech firms get what they want, the SMRs won’t solve all the power
12:45
demands of AI by themselves. They’ll rearrange the playing field but they won’t be a magic bullet.
12:51
They’ll rearrange something else, too. How we, the public, view the very
12:54
nature of what a tech company is. They don’t just make products to help you
12:58
be productive at work or search for a good place to eat dinner anymore. They’re actively trying to
13:03
become a player in critical US infrastructure. And once they have that kind of power,
13:08
they’re not going to let go of it. Chapter 5 - The Sovereign Cloud
13:12
Tech giants are not just buying up the existing spaces on the Monopoly board;
13:16
they’re creating new ones for themselves. But not every country in the
13:20
world will have nuclear power. They won’t all have abundant resources,
13:24
or a grid capable of supporting large-scale AI infrastructure. So, if nations want to remain
13:29
technologically competitive, many might find themselves relying on cloud platforms and compute
13:34
capacity controlled by foreign firms. What does that mean?
13:37
It means power. Political power.
13:40
That dependence nations might have on private companies could shift global power dynamics in
13:44
profound ways. Influence would no longer just rest with governments, militaries, or central banks,
13:50
but with corporations controlling energy contracts and digital systems that modern
13:55
economies will need to keep functioning. Businesses started in someone’s basement
13:59
or garage in Silicon Valley are becoming as powerful as sovereign states . Private
14:03
infrastructure powers with the ability to shape markets, policy, and holding the keys to access
14:09
the tools needed to be a modern economy. CEOs from Microsoft, Tesla, Amazon, Google,
14:14
Meta and others are now regular features at the White House. They are consulted on AI policy,
14:20
cybersecurity, competition, censorship, national security and economic strategy. In
14:25
earlier periods of history, hidden power and its invisible hands was often said to be intelligence
14:30
agencies, media tycoons, shadowy financiers, or military-industrial complex defense contractors.
14:36
Today, Big Tech can be added to the list. As Carnegie Europe said in 2025,
14:41
billionaire tech titans quote, “now rival nation-states in influence,
14:46
shaping the rules of the global digital order.” Their technologies, and increasingly AI,
14:51
dictate how we communicate, how we trade with each other. As the recent “all lawful use” argument
14:56
between Anthropic and the US government shows, they also play a major role in how we wage war.
15:02
Political realists often write that the global balance of power largely relies on the size of
15:06
an economy and its military. Well, we can now add the size of data centers to that equation, too.
15:12
In the years ahead, real power may belong not only to those who govern nations,
15:17
but to those who own the electricity and computing power behind artificial intelligence. Some of
15:22
the biggest firms are not just richer than many small countries. In the future, they might well
15:27
have more influence on how the world is run. For some critics, that’s a major concern.
15:33
Chapter 6 - The Real Revolution The real power of AI isn’t what
15:37
it might seem at first sight. Products that make deepfake
15:40
videos of a pop star or produce videos of the U.S. declaring war on Great Britain.
15:44
No, the power is in all the processes you don’t see. The grease that will oil the wheels
15:50
and cogs of the future global economy. And it happens mostly behind the scenes
15:54
receiving very little credit for its genius. That is, agentic AI systems for complex tasks
16:00
used by B2B companies. The focus now is on Multi-Agent Orchestration, where teams of AI
16:05
agents working together can handle anything from workflows to financial risk management.
16:10
These systems don’t just answer questions, they collaborate with staff and optimize procedures at
16:15
every step of the way. Current reports are saying the most effective ones can improve
16:20
efficiency by a massive 30% to 50%. When something breaks, they flag it,
16:25
and that’s when the human overseers step in. The goal is to save time, save money, and on a
16:31
scale that is changing entire industries. Think about global logistics.
16:35
Some of the biggest logistics firms can lose huge amounts of money through delays and empty trucks.
16:41
If AI cuts waste by just 2%, that still could be worth billions. The proof? Recent reports
16:48
say the market for AI in logistics reached $12 billion in 2026, up from $8.2 billion in 2024.
16:55
The reason was automation. Not replacing people in warehouses
16:59
with clever robots, but adding an AI layer to operations that streamlined many of the
17:04
processes. Industry reports say in just two years AI has cut down on operational costs by 20-30%.
17:11
Drug discovery and biotech are also expected to experience an AI windfall. The cost of bringing a
17:16
successful drug to market can run into billions and take years. But if AI can help scientists
17:22
better understand how diseases work, design better drug compounds and run smarter clinical trials,
17:28
it could completely change the very big business of sickness and health.
17:32
The potential is huge. But potential right now is the operative word.
17:36
Countless analyst reports tell us AI will make drug companies billions over the next 10 years
17:41
or so, but it’s not nailed on. If the predictions are correct, the pharmaceutical industry stands to
17:47
increase operating profits by more than 10%. The first AI boom entertained us. We’ve had
17:52
a lot of fun. But all the hype behind it and overvalued companies will probably soon die out.
17:58
What’s happened is a cleansing. When the gimmicks have been washed away, the serious money will be
18:03
invested in specialized B2B applications in just about all the major industries.
18:08
Industries that will rely on a handful of companies with all the biggest data centers.
18:12
This brings us to the terrifying evolution of the tech industry.
18:15
Chapter 7 - The Private Empire The largest technology firms are
18:19
no longer just software vendors. They run global cloud networks
18:23
that make the world’s economy tick over. They maintain private cyber-defense divisions,
18:27
armies of the future. They issue cloud credits that function like economic
18:32
incentives inside their own ecosystems. It’s a kind of currency in their digital
18:36
kingdoms that businesses will need to function. And as you’ve seen, they’re moving into power
18:41
itself, securing nuclear contracts, gas generation and dedicated electricity supplies for data
18:47
centers. They are effectively ending the nation state’s monopoly on critical infrastructure.
19:18
Some of those data centers might one day stand empty, graffitied and broken. A
19:22
monument to investor overconfidence abandoned before a single server was even switched on.
19:27
But Big Tech will by then have already consolidated its power because it
19:32
secured all the energy deals. The speculative frenzy will
19:33
burn itself out. Weaker players will collapse. But then what remains is
19:33
infrastructure concentrated in fewer hands. The AI boom won’t be remembered as a passing
19:33
mania people laugh at. It will be remembered as the moment a handful of companies secured
19:34
control over compute, energy, and the digital infrastructure everything else depends on.
19:34
Not platforms. Not products. Infrastructure.
19:34
Private empires, positioned underneath the next era of technology, charging
19:38
rent on everything built above them. But empires don’t just run on power.
19:42
They run on cash. What happens when the money required to sustain it starts to run out? Find
19:48
out in ‘$115 Billion Burn Rate. The AI Bubble Just POPPED’. Or click on this video instead.
