October 30, 2024 • 6 mins
Prepare to be shocked by the truth behind the AI-driven data center boom and its colossal impact on our electricity landscape. With predictions suggesting a staggering leap in power consumption by 2030, this episode promises to unpack the complexities of this rapid development. From exclusive insights gathered from the Electric Power Research Institute and various utility companies, you'll gain a comprehensive understanding of why this exponential demand for data center hookups is blindsiding even seasoned industry insiders. Journey with us through pivotal regions like Virginia and Texas, where the grid is under unprecedented pressure, and explore the surprising predictions from McKinsey that could redefine our expectations.
Join our expert guests as we explore the implications of this "power grab" for sustainability goals, our communities, and, inevitably, our electricity bills. As we dissect how AI is at the heart of this transformation, we'll visit the extraordinary developments in Ohio and consider what the future holds for energy consumption in the digital age. Discover why this surge in data center demand has significant climate ramifications, and why it might just be the biggest game-changer in recent memory. This episode, the first in our multi-part series, sets the stage for understanding the monumental shifts underway and the challenges that lie ahead.
Join our expert guests as we explore the implications of this "power grab" for sustainability goals, our communities, and, inevitably, our electricity bills. As we dissect how AI is at the heart of this transformation, we'll visit the extraordinary developments in Ohio and consider what the future holds for energy consumption in the digital age. Discover why this surge in data center demand has significant climate ramifications, and why it might just be the biggest game-changer in recent memory. This episode, the first in our multi-part series, sets the stage for understanding the monumental shifts underway and the challenges that lie ahead.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:01):
The past few weeks, every second story seems to be
about AI and data center load.
The topic has certainlydominated my recent weekly
videos, but since we're mostlygetting isolated data points
from the media, perhaps it'stime to construct the big
picture.
This week, instead of a summaryof top stories, we're going to
go deep in the first of amulti-part series on the
(00:21):
800-pound gorilla.
That's rather abruptly changedthe electricity supply and
demand landscape AI and datacenters.
It's not that electric vehicleload or power demand from
Bitcoin doesn't matter.
It's just that the potentialfuture demand from AI-driven
data centers is so much larger.
How big, that remains to beseen.
But let's consider somerecently available numbers.
(00:43):
In May, the Electric PowerResearch Institute, epri,
published a report saying thatdata centers might consume up to
9% of US electricity generationby 2030.
Epri then followed that up witha May to July survey of 26
utilities, of which 60% saidthey had requests for new
hookups of 500 megawatts orlarger.
(01:04):
48% of those utilitiesindicated they had requests
exceeding 1,000 megawatts andalmost half said that current
data center requests exceeded50% of current peak demand.
This load has come out ofnowhere.
None of the 26 utilitiessurveyed had current data center
connections over 500 megawatts.
(01:24):
Then there's the one utilitysitting in the Virginia data
epicenter, dominion.
Virginia is the world's datahub because of the high-speed
fiber backbone located there.
Google estimates 70% of theworld's internet traffic goes
through northern Virginia and in2022, dominion served about
2,800 megawatts of data centerload.
That was before the AI boom.
(01:47):
That number has since soared to4,000 megawatts by the end of
last year and Dominion will addanother 1,000 megawatts or more
this year, with totalconsumption in Virginia now
equaling one quarter of theentire state's consumption.
50,000 megawatts of additionaldata centers are waiting in line
, but many won't get built.
Dominion now says it won'tservice new interconnection
(02:08):
requests for data centers over100 megawatts for seven years.
Meanwhile, in Texas, encorefaces 59,000 megawatts of data
center connection requests.
For context, the Texas grid sawa record peak of 85,000
megawatts this past summer.
Then there's AEP Ohio, in themiddle of a tussle with data
center giants there.
(02:29):
As I noted last week, there'squite the dialogue going on in
Ohio and it gets to the heart ofthe issue.
As of May, when the utilityfiled a proposal for how to
address data center load, aepOhio served 600 megawatts of
data centers with about 4,400megawatts of interconnection
requests Fast forward to today,and those now exceed 40,000
(02:51):
megawatts.
One last note before we talkabout why this is happening and
what it all means.
Mckinsey thinks EPRI isunderestimating the demand and
projects an 11 to 13 percent oftotal load growth by 2030,
totaling 80,000 megawatts.
Hmm, encore, aap and Dominionalone tot up well north of 80
(03:11):
gigawatts.
So what's that all about?
More on that later as well.
It's pretty obvious that thisvery recent data center demand
development, which seeminglycame out of nowhere, has some
serious implications for ourclimate and sustainability goals
.
Nobody really saw this coming,with the exception of a few
industry insiders in theartificial intelligence space,
(03:33):
and even most of them have beensurprised at the speed with
which this has blossomed.
This dynamic is also likely tomeaningfully impact electricity
bills of anybody being served inmarkets where these data
centers are growing.
Why?
Let's dive into thecomplexities and find out.
First let's talk about AI andwhy the literal power grab.
(03:54):
In March of 2016, somethingvery unusual and somewhat
earth-shaking happened, nowknown as Move 37.
We already knew that computerswere getting better and perhaps
smarter when IBM's Deep Bluebeat world chess champion Garry
Kasparov in 1997.
But at some level, that wassimple math.
(04:14):
Then, in 2011,.
Jeopardy's reigning champion,ken Jennings, accepted a
challenge from IBM's Watson Atthe time.
Jennings expressed confidence,saying later that he'd taken
some AI classes, and quote Iknew there were no computers
that could do what you need todo to win on Jeopardy.
People don't realize how toughit is to write that kind of
(04:34):
program that can read a clue ina natural language like English.
To understand the puns, the redherrings, to unpack just the
meaning of the clue, I thought,yes, I will destroy the computer
.
Unquote.
Watson, however, was able todecipher the puns and nuances
and destroy Jennings, but itstill couldn't reason.
(04:56):
The next highly publicizeddemonstration of Ari's prowess
in a contest was AlphaGo's joustwith top Korean Go player Lee
Sedol.
Go is way harder than chess.
Chess has 64 squares, go has361, and you place black or
white stones strategically onthe board to capture territory.
In chess there are 30 possiblenext moves, while in Go you have
(05:18):
over 200.
So within two moves, in chessyou have about 400 potential
outcomes, versus around 130,000in Go.
Within just two moves, theAlphaGo team taught its computer
to play against humans and thenreleased it to play millions of
games against itself.
So on that much publicizedjoust in Korea, on move 37 of
(05:41):
that first game, alphago placeda stone in a location that
nobody would have expectedappearing to demonstrate
creativity, and it was the movethat eventually won the game and
demonstrated the growingcapabilities of AI.
But AI still couldn'tdemonstrate prowess on tests or
deal with any complexities oflanguage Until recent years.
(06:02):
Now it can, and it's gettingbetter with startling speed.
Consider two standard exams asproof.
In 2022, chat-gbt 3.5 scored inthe 40th percentile on the law
school LSAT exam.
In 2023, it jumped to 88th.
How about the standard SAT?
(06:23):
Ah, there we went from 87th to97th in a year, and within a few
years we were likely to achievewhat is known as
superintelligence.
That's where computers exceedthe thought processes of the
best of us.
How and why is this occurringso quickly?
Quite simply, the machines aretraining more rapidly on more
powerful chips.
(06:43):
Next week, we'll discuss thosechips and the electricity they
devour.
Thanks for watching and we'llsee you then.
The past few weeks, every second story seems to be
about AI and data center load.
The topic has certainlydominated my recent weekly
videos, but since we're mostlygetting isolated data points
from the media, perhaps it'stime to construct the big
picture.
This week, instead of a summaryof top stories, we're going to
go deep in the first of amulti-part series on the
(00:21):
800-pound gorilla.
That's rather abruptly changedthe electricity supply and
demand landscape AI and datacenters.
It's not that electric vehicleload or power demand from
Bitcoin doesn't matter.
It's just that the potentialfuture demand from AI-driven
data centers is so much larger.
How big, that remains to beseen.
But let's consider somerecently available numbers.
(00:43):
In May, the Electric PowerResearch Institute, epri,
published a report saying thatdata centers might consume up to
9% of US electricity generationby 2030.
Epri then followed that up witha May to July survey of 26
utilities, of which 60% saidthey had requests for new
hookups of 500 megawatts orlarger.
(01:04):
48% of those utilitiesindicated they had requests
exceeding 1,000 megawatts andalmost half said that current
data center requests exceeded50% of current peak demand.
This load has come out ofnowhere.
None of the 26 utilitiessurveyed had current data center
connections over 500 megawatts.
(01:24):
Then there's the one utilitysitting in the Virginia data
epicenter, dominion.
Virginia is the world's datahub because of the high-speed
fiber backbone located there.
Google estimates 70% of theworld's internet traffic goes
through northern Virginia and in2022, dominion served about
2,800 megawatts of data centerload.
That was before the AI boom.
(01:47):
That number has since soared to4,000 megawatts by the end of
last year and Dominion will addanother 1,000 megawatts or more
this year, with totalconsumption in Virginia now
equaling one quarter of theentire state's consumption.
50,000 megawatts of additionaldata centers are waiting in line
, but many won't get built.
Dominion now says it won'tservice new interconnection
(02:08):
requests for data centers over100 megawatts for seven years.
Meanwhile, in Texas, encorefaces 59,000 megawatts of data
center connection requests.
For context, the Texas grid sawa record peak of 85,000
megawatts this past summer.
Then there's AEP Ohio, in themiddle of a tussle with data
center giants there.
(02:29):
As I noted last week, there'squite the dialogue going on in
Ohio and it gets to the heart ofthe issue.
As of May, when the utilityfiled a proposal for how to
address data center load, aepOhio served 600 megawatts of
data centers with about 4,400megawatts of interconnection
requests Fast forward to today,and those now exceed 40,000
(02:51):
megawatts.
One last note before we talkabout why this is happening and
what it all means.
Mckinsey thinks EPRI isunderestimating the demand and
projects an 11 to 13 percent oftotal load growth by 2030,
totaling 80,000 megawatts.
Hmm, encore, aap and Dominionalone tot up well north of 80
(03:11):
gigawatts.
So what's that all about?
More on that later as well.
It's pretty obvious that thisvery recent data center demand
development, which seeminglycame out of nowhere, has some
serious implications for ourclimate and sustainability goals
.
Nobody really saw this coming,with the exception of a few
industry insiders in theartificial intelligence space,
(03:33):
and even most of them have beensurprised at the speed with
which this has blossomed.
This dynamic is also likely tomeaningfully impact electricity
bills of anybody being served inmarkets where these data
centers are growing.
Why?
Let's dive into thecomplexities and find out.
First let's talk about AI andwhy the literal power grab.
(03:54):
In March of 2016, somethingvery unusual and somewhat
earth-shaking happened, nowknown as Move 37.
We already knew that computerswere getting better and perhaps
smarter when IBM's Deep Bluebeat world chess champion Garry
Kasparov in 1997.
But at some level, that wassimple math.
(04:14):
Then, in 2011,.
Jeopardy's reigning champion,ken Jennings, accepted a
challenge from IBM's Watson Atthe time.
Jennings expressed confidence,saying later that he'd taken
some AI classes, and quote Iknew there were no computers
that could do what you need todo to win on Jeopardy.
People don't realize how toughit is to write that kind of
(04:34):
program that can read a clue ina natural language like English.
To understand the puns, the redherrings, to unpack just the
meaning of the clue, I thought,yes, I will destroy the computer
.
Unquote.
Watson, however, was able todecipher the puns and nuances
and destroy Jennings, but itstill couldn't reason.
(04:56):
The next highly publicizeddemonstration of Ari's prowess
in a contest was AlphaGo's joustwith top Korean Go player Lee
Sedol.
Go is way harder than chess.
Chess has 64 squares, go has361, and you place black or
white stones strategically onthe board to capture territory.
In chess there are 30 possiblenext moves, while in Go you have
(05:18):
over 200.
So within two moves, in chessyou have about 400 potential
outcomes, versus around 130,000in Go.
Within just two moves, theAlphaGo team taught its computer
to play against humans and thenreleased it to play millions of
games against itself.
So on that much publicizedjoust in Korea, on move 37 of
(05:41):
that first game, alphago placeda stone in a location that
nobody would have expectedappearing to demonstrate
creativity, and it was the movethat eventually won the game and
demonstrated the growingcapabilities of AI.
But AI still couldn'tdemonstrate prowess on tests or
deal with any complexities oflanguage Until recent years.
(06:02):
Now it can, and it's gettingbetter with startling speed.
Consider two standard exams asproof.
In 2022, chat-gbt 3.5 scored inthe 40th percentile on the law
school LSAT exam.
In 2023, it jumped to 88th.
How about the standard SAT?
(06:23):
Ah, there we went from 87th to97th in a year, and within a few
years we were likely to achievewhat is known as
superintelligence.
That's where computers exceedthe thought processes of the
best of us.
How and why is this occurringso quickly?
Quite simply, the machines aretraining more rapidly on more
powerful chips.
(06:43):
Next week, we'll discuss thosechips and the electricity they
devour.
Thanks for watching and we'llsee you then.
Popular Podcasts
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Therapy Gecko
An unlicensed lizard psychologist travels the universe talking to strangers about absolutely nothing. TO CALL THE GECKO: follow me on https://www.twitch.tv/lyleforever to get a notification for when I am taking calls. I am usually live Mondays, Wednesdays, and Fridays but lately a lot of other times too. I am a gecko.