Prática de Shadowing: The Limitless AI Lie. The Bubble Is Slowly BURSTING. - Aprenda a falar inglês com o YouTube
C1
You’ve probably heard the AI bubble is popping. That it’s all hype… a gimmick… and the money is drying up. That’s not what’s happening. What’s actually going on is a lot more serious… and almost nobody is talking about it.
76 frases
Se as frases estiverem muito curtas ou longas, clique em Edit para ajustá-las.
1
You’ve probably heard the AI bubble is popping. That it’s all hype… a gimmick… and the money is drying up. That’s not what’s happening. What’s actually going on is a lot more serious… and almost nobody is talking about it.
2
The AI revolution didn’t hit a financial wall. It hit a physical one.
3
Right now, roughly 50% of American data center projects in major hubs are delayed or quietly canceled. Not because companies ran out of cash. Because they ran out of power.
4
Now, Big Tech isn’t just making software anymore. They've become some of the most aggressive investors in one industry you wouldn’t expect… Nuclear power. This is the Limitless AI Lie.
5
Chapter 1 - The Ghost Clusters We are currently in the midst of a global artificial intelligence arms race and it’s getting very expensive.
6
In 2025, The Wall Street Journal reported that just in the United States, the Big Four – Amazon, Google, Meta, Microsoft – were collectively spending close to $400 billion annually, with most of that money going on infrastructure to run giant AI data centers.
7
These companies are taking on heavy debt, but as The Journal pointed out, the “current revenue” from AI was, and still is, “relatively tiny.” Even so, the reports of unprecedented spending just kept coming. There’s a reason. And it’s why OpenAI, Oracle, and SoftBank announced they would spend just under half a trillion dollars on the Stargate initiative. They’re rolling out five new data centers in Texas, New Mexico, Ohio and the Midwest. CEO Sam Altman promised that the investment would lead to future breakthroughs in technology, adding that they are already making, quote, “historic progress.” Big Tech and every other global enterprise with a focus on AI are expected to spend $7 trillion on building or updating data centers in the next 5 years.
8
That’s the plan, but reality shows a very different picture.
9
Data centers aren’t easy to run. Of the almost 11,000 active ones in the world in 2026, the US had the lion's share at around 4,000, well ahead of the UK (511), Germany (507), China (368), and France (344). But just to run the ones we already have is extraordinarily energy-intensive. Cooling them can take vast amounts of water, with the biggest hyperscale sites using millions of gallons a day. Still, heat can be managed.
10
The problem is powering them. Reports in 2026 stated that 50% of data centers being constructed in the US are currently either canceled or delayed, with experts saying that number is going to rise. In major hubs like Northern Virginia – the region known as “Data Center Alley” – projects are already slowing as the grid starts to buckle under demand. The public sentiment has changed. Polls show just 35% of Virginia voters give their thumbs up to the power-hungry projects compared to 69% in 2023.
11
The delays aren’t about a lack of money or ambition. Silicon Valley has plenty of both. They’re about supply chains, engineering limits, and public/political resistance…but most of all, power. The future Sam Altman talks about?
12
It’s sitting in a queue… waiting for electricity. Right now, America is facing a bottleneck caused by unprecedented demand. So many of the data center projects are in a traffic jam, waiting to connect. Before anything can plug in, utilities and grid operators must run studies to see whether lines, substations and transformers can cope with the extra load. They have to assess reliability. They have to look at the environmental impact.
13
This can cost millions of dollars and lead to years of waiting around.
14
It doesn’t just affect Big Tech. It affects you.
15
These massive data centers burn through enormous amounts of electricity. When their energy bills spike, that cost doesn’t just disappear. It gets passed on. To consumers. To communities.
16
And when people start seeing higher energy prices, they push back. That can be lethal to the ambitions of anyone wanting to build a data center. And even if the projects get past all those hurdles, the physical system might still not be ready to go.
17
According to Lawrence Berkeley National Laboratory, these queues add up to about 2,600 gigawatts of proposed generation and storage capacity. That’s about twice the size of the current U.S. power system. It’s a backlog as big as the grid itself.
18
The amount of power these new data centers will require to keep running is astronomical.
19
Some of the new ones in the US will use 50 to 100 times more power than it takes to power the Empire State Building… just to run and cool the servers. Reports say the biggest ones currently under construction will collectively demand as much electricity as an entire country like Italy.
20
Big tech wants to spend trillions, but the “power crunch” is getting in the way.
21
It’s a problem they can’t code around. Chapter 2 - The Physics of the Bubble On average, a single query to ChatGPT uses around 10 times more energy than a search on Google. To put another way, one complex prompt can use about as much electricity as leaving an LED lightbulb on20 minutes. The exact number is debated. And not every query is the same. But the scale is. Because ChatGPT isn’t handling thousands of requests… It’s handling billions. 2.5 billion queries per day, it all adds up. The US alone experienced about 330 million ChatGPT queries per day in 2025 That’s just the beginning. Recent reports tell us about 1.1 billion people worldwide use AI but that number is growing fast. 56% of Americans use it in their day to day lives. About 60% of American teachers have reported experimenting with it.
22
88% of global firms are now onboard, too. The experts say there are still about 5 billion people globally who don’t use it, but the adoption curve is rising fast. In the next 5 or 10 years, we can expect many of those 5 billion will be regular users.
23
So as more and more people start experimenting with it, the world will need to generate a lot more power. Goldman Sachs Research analysts have estimated that global data center power demand will grow 160% by 2030. That’s about the equivalent of adding a second 'Germany' to the global energy grid in just 6 years. Data centers already consume about 1-2% of all the world’s electricity. If the experts are right, it will be more like 3-4% by the time the decade is over. Where will all this new power come from?
24
In 2023 alone, new applications for interconnections surged by about 30%. The data showed that close to 95% of requested capacity came from solar, wind and battery storage.
25
But here’s the problem. Renewables aren’t constant.
26
The wind slows down. The sun disappears. Even in the best solar regions in the U.S., you only get about 3,500 to 4,000 hours of usable sunlight a year. That's out of almost 8,800 hours in total.
27
Less than half the time. In other words, solar and wind won’t be enough to power the AI revolution. Companies are now fighting for whatever power they can get, and whoever controls the power, controls the future of AI.
28
Chapter 3 - The Great AI Cleansing The delays in data centers look like a failure of infrastructure. But they may also be creating something else: scarcity.
29
And scarcity, for the companies with all the money, isn’t always a problem.
30
It can actually be a very real advantage. The biggest tech firms are already making deals with energy companies. Meta, Google, Amazon, Microsoft, have inked deals with utilities firms all over the USA. They’ll get the power. The smaller players won’t.
31
Venture-backed startups might be able to raise money and promise major market disruption, but they won’t be able to compete for energy. They’ll have to rely on Big Tech for renting and compute power. That could mean that every GPU shortage, every power delay, every price rise will hit them first. And the state-of-the-art models might become prohibitively expensive or limited in access. The biggest tech companies will decide who gets access first, who gets the most computing power, and who gets the best versions.
32
Meanwhile the market itself is changing. Investors are losing interest in generic chatbots that write poems and hallucinate history. Capital is moving toward specialized AI that can write code, optimize logistics, detect fraud, perform medical analysis, and above all, increase profit margins.
33
But not everyone will receive the same treatment, and some might be left waiting. Countries have anticompetition laws, but the gatekeepers of technology as they consolidate their power over AI, might move faster than the law can keep up. A specialized startup might not ever truly be able to compete, not if it's denied access. In the gold rush, fortunes weren’t made by the miners, but by those who controlled the tools and the land.
34
This time, it isn’t just the tools. It’s the electricity that powers them.
35
For years now, people have been worrying about the rise of the machines, rogue AI ending the world, but that’s all science fiction. The real danger is a handful of monopoly powers who control all the key infrastructure. But to guarantee their monopoly, Big Tech knows it can’t just rely on public infrastructure. So it’s planning to go off-grid.
36
Chapter 4 - The Nuclear Pivot For decades, tech companies rented power from the grid like everyone else. Now, that’s changing. These days some of the largest firms with seemingly bottomless reserves of money are getting into the business of procuring their own energy assets. They’re not just software companies anymore.
37
In 2024, Microsoft made a deal to reactivate Three Mile Island nuclear power plant, a site forever associated with a 1979 meltdown. One reactor was shut down permanently. The other ran until 2019, when it was closed due to operating losses. Now, decades later, it’s being revived.
38
Not for history, but for demand. Under a 20-year power agreement, Microsoft is securing 835 megawatts of electricity for Pennsylvania’s grid.
39
The deal wasn’t about powering Pennsylvanians’ personal computers or domestic appliances, but achieving lift-off on its planned data centers. It’s enough electricity to power about 800,000 American homes, and all of it just for Microsoft's AI.
40
But will it happen? Well, firstly, a nuclear plant has never in American history been fully shut down and sprung back to life.
41
Experts are saying that for Microsoft’s plan to come to fruition, there will be numerous regulatory hurdles to cross. The permits are all still pending, and they might not ever come, even if the plant owner, Constellation Energy, is trying to speed things up.
42
Nuclear power tends to make people uneasy. Even if the plant did get a second lease of life, it’s not clear it could be integrated back into the grid. Mark Zuckerberg’s Meta also hit a problem when trying to build a data center next to what was reported as an “existing nuclear facility.” The media didn’t name the facility but it did report that Zuckerberg’s plan of operating a nuclear-powered AI center crashed to the ground because of…bees. A rare species populated the area where Meta wanted to build. It goes to show how Big Tech is thinking outside of the box. Amazon has tried, too.
43
In 2024, the company signed a $650 million deal with Talen Energy in Pennsylvania where it plans to open a number of data centers close to the Susquehanna nuclear power plant. The deal, which is set to end in 2042, could be delivering almost double the power needed for the city of Pittsburgh. That’s if it ever gets off the ground.
44
In 2024, the Federal Energy Regulatory Commission blocked Amazon from taking power directly from a nuclear plant to feed one of its hyperscale data centers.
45
The reason? Diverting that much power raised the risk of blackouts for the communities already relying on that same nuclear supply.And even if the lights stayed on, there was another consequence… higher bills.
46
The regulators are standing firm. Amazon later asked for a re-hearing and the result stayed the same. Google is also looking to nuclear energy to power its future AI ambitions. It’s one of a few big tech firms currently investing in next-generation nuclear energy in the form of small modular reactors (SMRs).
47
Smaller means less cost to build. It also means cooler, so ideally, less chance of a meltdown.
48
The company hopes to have the data center up and running by 2030. But long before any green lights start flashing, Google will need approval from the U.S. Nuclear Regulatory Commission (NRC).
49
And that won’t come easy. As one analyst wrote, the “path to commercial viability is still riddled with uncertainty.” SMR designs haven’t been proven at scale and regulatory hurdles have set a high bar. With SMRs, there’s no easy path forward.
50
But there’s political will, and that’s important. The current administration has given the thumbs up and has said it wants these small reactors to be functioning by 2030.
51
None of the new commercial SMRs are yet operating in the US, but China and Russia already have working examples. In terms of the global AI race, this isn’t good news for Uncle Sam.
52
But even if the tech firms get what they want, the SMRs won’t solve all the power demands of AI by themselves. They’ll rearrange the playing field but they won’t be a magic bullet.
53
They’ll rearrange something else, too. How we, the public, view the very nature of what a tech company is. They don’t just make products to help you be productive at work or search for a good place to eat dinner anymore. They’re actively trying to become a player in critical US infrastructure. And once they have that kind of power, they’re not going to let go of it. Chapter 5 - The Sovereign Cloud Tech giants are not just buying up the existing spaces on the Monopoly board; they’re creating new ones for themselves. But not every country in the world will have nuclear power. They won’t all have abundant resources, or a grid capable of supporting large-scale AI infrastructure. So, if nations want to remain technologically competitive, many might find themselves relying on cloud platforms and compute capacity controlled by foreign firms. What does that mean?
54
It means power. Political power.
55
That dependence nations might have on private companies could shift global power dynamics in profound ways. Influence would no longer just rest with governments, militaries, or central banks, but with corporations controlling energy contracts and digital systems that modern economies will need to keep functioning. Businesses started in someone’s basement or garage in Silicon Valley are becoming as powerful as sovereign states . Private infrastructure powers with the ability to shape markets, policy, and holding the keys to access the tools needed to be a modern economy. CEOs from Microsoft, Tesla, Amazon, Google, Meta and others are now regular features at the White House. They are consulted on AI policy, cybersecurity, competition, censorship, national security and economic strategy. In earlier periods of history, hidden power and its invisible hands was often said to be intelligence agencies, media tycoons, shadowy financiers, or military-industrial complex defense contractors.
56
Today, Big Tech can be added to the list. As Carnegie Europe said in 2025, billionaire tech titans quote, “now rival nation-states in influence, shaping the rules of the global digital order.” Their technologies, and increasingly AI, dictate how we communicate, how we trade with each other. As the recent “all lawful use” argument between Anthropic and the US government shows, they also play a major role in how we wage war.
57
Political realists often write that the global balance of power largely relies on the size of an economy and its military. Well, we can now add the size of data centers to that equation, too.
58
In the years ahead, real power may belong not only to those who govern nations, but to those who own the electricity and computing power behind artificial intelligence. Some of the biggest firms are not just richer than many small countries. In the future, they might well have more influence on how the world is run. For some critics, that’s a major concern.
59
Chapter 6 - The Real Revolution The real power of AI isn’t what it might seem at first sight. Products that make deepfake videos of a pop star or produce videos of the U.S. declaring war on Great Britain.
60
No, the power is in all the processes you don’t see. The grease that will oil the wheels and cogs of the future global economy. And it happens mostly behind the scenes receiving very little credit for its genius. That is, agentic AI systems for complex tasks used by B2B companies. The focus now is on Multi-Agent Orchestration, where teams of AI agents working together can handle anything from workflows to financial risk management.
61
These systems don’t just answer questions, they collaborate with staff and optimize procedures at every step of the way. Current reports are saying the most effective ones can improve efficiency by a massive 30% to 50%. When something breaks, they flag it, and that’s when the human overseers step in. The goal is to save time, save money, and on a scale that is changing entire industries. Think about global logistics.
62
Some of the biggest logistics firms can lose huge amounts of money through delays and empty trucks.
63
If AI cuts waste by just 2%, that still could be worth billions. The proof? Recent reports say the market for AI in logistics reached $12 billion in 2026, up from $8.2 billion in 2024.
64
The reason was automation. Not replacing people in warehouses with clever robots, but adding an AI layer to operations that streamlined many of the processes. Industry reports say in just two years AI has cut down on operational costs by 20-30%.
65
Drug discovery and biotech are also expected to experience an AI windfall. The cost of bringing a successful drug to market can run into billions and take years. But if AI can help scientists better understand how diseases work, design better drug compounds and run smarter clinical trials, it could completely change the very big business of sickness and health.
66
The potential is huge. But potential right now is the operative word.
67
Countless analyst reports tell us AI will make drug companies billions over the next 10 years or so, but it’s not nailed on. If the predictions are correct, the pharmaceutical industry stands to increase operating profits by more than 10%. The first AI boom entertained us. We’ve had a lot of fun. But all the hype behind it and overvalued companies will probably soon die out.
68
What’s happened is a cleansing. When the gimmicks have been washed away, the serious money will be invested in specialized B2B applications in just about all the major industries.
69
Industries that will rely on a handful of companies with all the biggest data centers.
70
This brings us to the terrifying evolution of the tech industry.
71
Chapter 7 - The Private Empire The largest technology firms are no longer just software vendors. They run global cloud networks that make the world’s economy tick over. They maintain private cyber-defense divisions, armies of the future. They issue cloud credits that function like economic incentives inside their own ecosystems. It’s a kind of currency in their digital kingdoms that businesses will need to function. And as you’ve seen, they’re moving into power itself, securing nuclear contracts, gas generation and dedicated electricity supplies for data centers. They are effectively ending the nation state’s monopoly on critical infrastructure.
72
Some of those data centers might one day stand empty, graffitied and broken. A monument to investor overconfidence abandoned before a single server was even switched on.
73
But Big Tech will by then have already consolidated its power because it secured all the energy deals. The speculative frenzy will burn itself out. Weaker players will collapse. But then what remains is infrastructure concentrated in fewer hands. The AI boom won’t be remembered as a passing mania people laugh at. It will be remembered as the moment a handful of companies secured control over compute, energy, and the digital infrastructure everything else depends on.
74
Not platforms. Not products. Infrastructure.
75
Private empires, positioned underneath the next era of technology, charging rent on everything built above them. But empires don’t just run on power.
76
They run on cash. What happens when the money required to sustain it starts to run out? Find out in ‘$115 Billion Burn Rate. The AI Bubble Just POPPED’. Or click on this video instead.
Baixar aplicativo
Pontuação por IA para cada frase que você fala
TRENDING
Populares
Por que praticar a fala com este vídeo?
Praticar a conversação em inglês com vídeos é uma das melhores maneiras de aprimorar suas habilidades linguísticas. Este vídeo, que discute as complexidades e os desafios da revolução da inteligência artificial, oferece um contexto rico e atual. Ao assistir e repetir as falas, você não apenas aprende sobre um tópico relevante, mas também se expõe a uma variedade de vocabulários e expressões, podendo aplicar isso em situações reais. A técnica de shadowing em inglês — que envolve imitar instantaneamente a fala do locutor — pode ser extremamente útil para melhorar sua fluência, entonação e compreensão auditiva.
Gramática e expressões em contexto
No vídeo, várias estruturas gramaticais e expressões são utilizadas que podem enriquecer seu vocabulário e sua compreensão. Aqui estão algumas delas:
- "O problema é alimentar eles." - Essa frase é um exemplo claro do uso coloquial do inglês, utilizando a forma de necessidade. Ao aprender tais construções, você pode se expressar de forma mais autêntica.
- "O futuro aguarda..." - Essa expressão é frequentemente usada para falar sobre expectativas e previsões. Incorporá-la no seu discurso pode demostrar um nível avançado de fluência.
- "Não é sobre falta de ambição." - Aqui, a estrutura negativa é usada para enfocar os contrários, uma técnica que pode ser valiosa para argumentação.
Essas estruturas tornam a conversa mais dinâmica e ajudam na construção de frases complexas em seus próprios diálogos.
Armadilhas de pronúncia comuns
Ao assistir ao vídeo, você também notará algumas palavras que podem apresentar desafios de pronúncia:
- "Infrastructure" - O som da combinação "str" pode ser complicado; pratique pausadamente, enfatizando cada parte da palavra.
- "Bottleneck" - A parte "neck" pode ser facilmente confundida se não for pronunciada claramente. Tente repetir em voz alta para dominar a entonação.
- "Energy" - Esta palavra pode ter variações dependendo do seu sotaque, portanto, ouça atentamente e tente reproduzir o som mais fielmente ao que você escuta no vídeo.
Identificar e praticar esses sons desafiadores pode ajudar a aprimorar sua articulação e fluência na conversação em inglês.
O que é a Técnica de Shadowing?
Shadowing é uma técnica de aprendizado de idiomas com base científica, originalmente desenvolvida para o treinamento de intérpretes profissionais. O método é simples, mas poderoso: você ouve áudio em inglês nativo e repete imediatamente em voz alta — como uma sombra seguindo o falante com 1-2 segundos de atraso. Pesquisas mostram melhora significativa na precisão da pronúncia, entonação, ritmo, sons conectados, compreensão auditiva e fluência na fala.
☕ Pague-nos um café
O ShadowingEnglish permanece 100% gratuito graças ao seu apoio. Os custos de servidor e IA são altos — seu café nos mantém ativos! 🙏