(Constantin K4/Shutter Stock)
The story of AI is often told in terms of GPU, achievements, and billionaire parameter models. However, each chat boot and AI application is a more basic force: electricity. Without it, even the latest AI stops grinding. At a blind speed on which AI is scaling, electricity is becoming both a currency and a barrier.
Rising demand and rigorous grid obstacles, policy makers and tech leaders are entering to resolve the scale. For years, nuclear energy has remained, which moves forward due to political risk and cost fluctuations. Now, as the AI background utility transforms power into frontline strategic concern, nuclear power is returning. This is not surprising, however, very few options are similar to the combination of nuclear energy density and reliability.
To see why the nuclear time has returned to the table, it helps to look at the sheer energy scale, which is now demanding the AI system. Just last week, Google shared a technical report on how much energy is taken to act on one of the AI prompts. It is still the most transparent estimate with a popular AI product by a large tech company. Although it only reflects a type of conversation, it offers a clear glimpse of the demand for energy that we are moving towards when AI embedded fast in daily life.
According to Google Requires 0.24 watt hours of electricity. This figure not only reflects the power used by AI chips itself, but also through its associated ingredients such as CPU, memory, backup system and cooling.
(Shutter stock AI generator)
Google reports that its AI models are becoming more efficient. This may be true on the basis of everyone, but it does not change a bigger picture. Since the use of AI spreads almost every industry, the total number of questions is increasing from the sky. Energy impressions are increasing rapidly than the benefits of any performance.
Multiply this single instant number from Google across the billions of questions a day, and you will start to see why the energy planners are playing the alarm and why AI companies are saying that power is the real obstacle to development.
A recent Goldman sex analysis It is estimated that the demand for global data center by 2030 may increase by 165 %. McKenis has predicted In 2020, consumption in the United States can reach 400 trivial hours a year, which is less than 100 -100 -hour hours in 2020. It will exceed the current power demand of the whole country like Mexico.
ٹیک کمپنیاں جوہری پر لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی لمبی تاریخ جس سے زیادہ اضافہ کر رہی ہیں۔ Many firms will need reliable, massive strength in the next three to five years. Traditional nuclear plants can take about a decade to build, and even the next generation options, such as small modular reactor (SMR), are still several years away from their first commercial deployment to the use of AI.
Some SMRs are already working abroad. Russia’s floating academic Lumonosov has been in power since 2020, and China’s Ling Long-one lives at the end of 2024. Canada has recently approved the construction of the next General BWRX-300 in Darlington, which is expected to be changed by 2030. However, none of them are reactors, so far none of them are reactors.
In fact, most AI companies will still rely on foam fuel for years. They are running coal plants and building new gas. The construction of nuclear facilities may be long -term, but it is not ready yet. In the meantime, tech companies are turning to existing nuclear assets, signing contracts with existing plants and supporting the resume to gain clear power soon.
(Ole.cnx/shutter stock)
Over the past year, Microsoft, Google, Amazon, and Meta have announced all nuclear energy measures. Some are straightforward contracts to buy electricity from existing plants, while other SMRs and other modern designs are focused on development.
Microsoft already exists Proceeded with a 20 -year contract With the energy, the 3 -mile -Island unit with energy to resume the planning of the reactor from the reactor. Amazon had recently signed a 1.9 gigwatts contract to buy atomic electricity With 2042, Tailon Energy’s Social Facility, a scale that makes it one of the largest corporate nuclear deal in the record.
Meta has also turned to nuclear, which has been earned a long -term contract Source Nuclear Power from Clinton Clean Energy Center Starting in 2027 in Illinois, obtaining a permanent supply of carbon -free electricity for your data centers.
Google recently signed a To deal with Keros Power to make SMR Meeting its energy needs. Michael Terral, the senior director of Google, for energy and climate, compiled the deal as a wider pressure part to secure reliable and clean strength for the company’s data centers. “We feel that the nuclear can play a vital role in helping us meet our demands, and to help us meet our demand, and to meet our demands, and to meet us 24 hours 24 hours,” he said.
Some companies are moving even more, targeting old nuclear plants that have been closed. For example, the palacedes in Michigan is a reactor On the way to resume in 2025 with a $ 1.5 billion federal loan. In Iowa, the closed Divine Arnold Plant can be resurrected with a new SMR hybrid project.
Other plants are working to advance the life and production. By 2035, 24 US reactors will be ready to renew the license. Each one can get a 20 -year extension. The Department of Energy also sees the capacity to increase the capacity in existing plants, estimating that the additional power between 2 and 8 GW can be added only by promoting performance. These stop gap arrangements may not get the headline, but these are some of the nuclear options available in the next few years.
Another challenge with nuclear power is less technical and more root in the past and credibility. Old disasters, wasted fears and weapons associations still cloud public understanding. This hesitation can slow down, delay investment, and even make safe projects politically dangerous.
“I am really excited about nuclear, but this is a technology that contains many myths and misinformation,” said Woodoff Professor Anna Ericson at the George W. Woodoff School of Mechanical Engineering (ME).
“The concerns about nuclear weapons, accidents and waste have shadowed nuclear energy capabilities as a clean, carbon -free technology,” he added.
However, AI power needs are not going anywhere. If the AI is increasing, the grid needs to grow with it. One way or the other, the tech companies and governments will have to solve the power problem, and now, and nothing like nuclear power offers the same mixture of reliability and scale.
Relevant
