Trillion-Dollar AI Ambitions Are Colliding With a Very Real Power Shortage
When you read about companies being worth a lot of money and
people talking about plans for Artificial Intelligence you might think that
Artificial Intelligence is doing great.. There is a problem that nobody can fix
with just computers. It is a problem. It is not very interesting.. It is
becoming a big issue.
The problem is electricity.
Data centers are getting bigger and bigger. The energy
infrastructure that powers them is not growing fast enough. This means that
there is not electricity to meet the demand.
The Important Numbers
People think that by 2026 data centers will use 565
terawatt-hours of electricity. That is a 26 percent increase from 2025. This is
according to some research from a company called Gartner.
To put this in perspective 565 terawatt-hours is electricity
than most countries use in a year.. It is going up fast.
By 2030 Gartner thinks that data centers will need 290
gigawatts of electricity and they will use more than 1,200 terawatt-hours. That
is as much electricity as Japan uses in a year.
The Change in Hardware That Is Causing the Problem
The reason why electricity demand is going up fast is
because of a change in the hardware that Artificial Intelligence uses.
In 2025 the special servers that are made for Artificial
Intelligence used 20 percent of the total electricity in data centers.. Their
electricity use went up by more than 83 percent.
In 2026 these special servers will use 31 percent of the
total electricity and their electricity use will go up by another 84 percent.
In 2027 these special servers will use electricity than the
regular servers.
The Cooling Problem That Nobody Talks About
The hardware that Artificial Intelligence uses gets very
hot.. It takes a lot of energy to keep it cool.
Just the cooling systems will use 195 terawatt-hours of
electricity in 2026. That is a 22.6 percent increase from 2025. This is because
the new hardware that Artificial Intelligence uses gets very hot and because
data centers are getting bigger.
What Gartner Says About This
A person named Linglan Wang who works at Gartner said it
clearly:
"The demand for Artificial Intelligence is making data
centers use more and more electricity and now the amount of electricity that is
available is limiting how much Artificial Intelligence we can use. This means
that making sure we have electricity is the new challenge for companies that
want to be successful in the Artificial Intelligence business."
The report from Gartner says that if things do not change
there will not be electricity to power all the new data centers that are being
built.
What This Means for the Artificial Intelligence Industry
You can already see the effects of this problem in the
Artificial Intelligence industry:
* When companies are looking for a place to build a data
center they have to think about whether there is enough electricity available.
* Companies are signing long-term contracts to buy
electricity, which's just as important as buying land or getting permits.
* The people in charge of infrastructure and operations have
to make sure that their data centers are using electricity efficiently and that
they have a connection to the electricity grid.
Linglan Wang advises companies to invest in cooling systems
that use electricity and to use edge computing to reduce the demand for
electricity.
The Picture
This is not just a problem for data centers. It is a problem
for Artificial Intelligence. If we cannot get electricity then all the money
that is being spent on Artificial Intelligence will be limited by the amount of
electricity that is available.
The numbers, from Gartner do not say whether we can solve
this problem in time.. They make it clear that this is a problem that needs to
be solved.
For now it is clear that Artificial Intelligence needs a lot
of electricity and the electricity grid is not ready.