00:00:00
welcome and thank you for joining us
00:00:02
today I'm VJ vtis waren I'm the global
00:00:05
energy and climate Innovation editor of
00:00:07
The Economist and I'm here at the global
00:00:11
conference on energy and AI live
00:00:14
convened by the International Energy
00:00:15
agency in Paris I'm conducting a series
00:00:18
of interviews with leaders from industry
00:00:20
government Civil Society about what AI
00:00:24
will have in store for the global
00:00:26
economy in the years
00:00:27
ahead I'm here now with Kate Brent Chief
00:00:30
sustainability Officer of Google welcome
00:00:32
Kate thank you great to see you great to
00:00:34
see you and what a a a timely event
00:00:36
indeed absolutely uh let's jump right
00:00:39
into it since our audience is a
00:00:40
sophisticated one people have followed
00:00:42
uh the morning's proceedings with the
00:00:44
plenaries and uh the big picture we all
00:00:46
know AI for energy energy for AI this is
00:00:49
one of the hot topics of the last 12
00:00:51
months and of course at Google you're
00:00:52
right in the thi of it um people will
00:00:55
know that uh your company is uh a great
00:00:59
beneficiary of Ai and a promoter of AI
00:01:01
but we have seen also the impacts on
00:01:03
energy and your own emissions your
00:01:05
carbon intensity as well as your energy
00:01:07
consumption y uh has increased uh can
00:01:09
you tell us how are you thinking about
00:01:11
this how do you deal with this difficult
00:01:13
challenge yeah absolutely so I would
00:01:15
just start by saying we've set very
00:01:18
ambitious goals for how we want to
00:01:20
responsibly operate our business so our
00:01:22
goal is to be Net Zero across our
00:01:24
operations and value Chain by 2030 and
00:01:26
also to power all of our offices and
00:01:28
data centers with 24x7 carbon free
00:01:29
energy so very ambitious goals
00:01:31
especially in light of what you've noted
00:01:33
that we are seeing AI really at this
00:01:35
inflection point I think one thing that
00:01:37
was missed a little bit in the coverage
00:01:39
of you know when we and many others put
00:01:41
out our annual reports and there was a
00:01:42
lot of discourse around what was going
00:01:44
on with the footprint was we disclosed a
00:01:46
new metric around the percentage
00:01:48
increase of data center electricity and
00:01:51
what that showed was a 17% increase
00:01:54
which was actually in line with previous
00:01:56
years so I think there was a narrative
00:01:57
that wanted to be told when it was
00:01:59
actually a little bit Le more nuanced
00:02:01
and that you're seeing more of the
00:02:02
historical Trends holding around overall
00:02:05
data center expansion and also emissions
00:02:07
in the supply chain so I just think
00:02:09
important to note to your point of
00:02:10
having a more kind of in-depth
00:02:11
conversation about this and also very
00:02:14
important that we need to responsibly
00:02:15
manage that footprint going forward so
00:02:17
we're really focused on that in three
00:02:18
ways it's looking at first our
00:02:20
infrastructure that's all the way down
00:02:22
to the chip making our chips as
00:02:23
efficient as possible our sixth
00:02:25
generation TPU which is our AI chip is
00:02:27
67% more efficient than the pr previous
00:02:30
generation also overall data centers
00:02:32
making them much more energy efficient
00:02:34
we're getting 4X the compute out of the
00:02:36
same amount of electricity compared to
00:02:38
five years ago then it's how you're
00:02:39
training the models and we've been
00:02:40
sharing best practices for how you can
00:02:42
really much more efficiently train
00:02:44
models reducing emissions up to 1,000x
00:02:47
energy 100x and then lastly there is the
00:02:49
carbon free energy that it's very
00:02:51
important to us to get a lot of clean
00:02:53
firm power onto grids and I know we'll
00:02:55
talk about that more in this
00:02:56
conversation let just unpack that a
00:02:57
little bit because you've kind of
00:02:58
sketched out your your road map um uh
00:03:00
starting with uh the uh what what seems
00:03:03
to be a great race right that there's a
00:03:05
race between uh increasing uh
00:03:08
applications for AI we're increasingly
00:03:10
finding New Uses and uh that's certainly
00:03:13
broadly speaking good for the world uh
00:03:15
in Innovation and Science and improving
00:03:17
efficiency across sectors uh but at the
00:03:20
same time obviously the energy
00:03:21
consumption needs themselves can be made
00:03:23
more efficient on the chip uh as well as
00:03:25
in the data center design you've
00:03:27
introduced things like a more efficent
00:03:29
data center design I think at least in
00:03:31
affiliate of yours sidewalk
00:03:33
infrastructure Partners I think has done
00:03:34
so but can you talk what is the
00:03:36
potential of that great raise because
00:03:37
typically what we find with in energy is
00:03:39
the jeans effect right if you make
00:03:41
things more efficient or cheaper people
00:03:42
tend to use more of energy as happened
00:03:45
with coal a few hundred years ago uh
00:03:47
could this happen that our our uses that
00:03:49
we find for AI uh outstrip the
00:03:51
efficiency gains that you're able to
00:03:53
come up with yeah so I think a few
00:03:55
things I would say about that the data
00:03:57
center sector as a whole has a very
00:04:00
long-standing track record of driving
00:04:02
really deep efficiency and of course why
00:04:04
is that well yes we all very much care
00:04:06
about environmental sustainability but
00:04:07
also for cost reasons and so there's a
00:04:09
piece of research that John Kumi did
00:04:11
many years ago that looked at that
00:04:12
period of time between 2010 and 2018
00:04:15
when there was a big data center boom
00:04:17
electricity consumption for data centers
00:04:19
during that time only increased 6% while
00:04:21
compute capacity increased by 550x so
00:04:25
this is a sector that knows how to
00:04:26
operate really efficiency and
00:04:27
efficiently and has a major business in
00:04:29
incentive to do so so I'm not
00:04:31
necessarily saying that Trend we will
00:04:32
see again but there's a lot of muscle
00:04:34
memory and a lot of understanding how to
00:04:35
keep driving that efficiency for
00:04:37
business reasons and then also I think
00:04:38
you're pointing to you know the net
00:04:40
benefit what are all the solutions that
00:04:42
are going to be possible that AI is
00:04:44
going to enable to modernize electricity
00:04:47
systems which is a lot of what we've
00:04:48
been talking about here at the Ia and
00:04:50
that's what we're seeing huge potential
00:04:51
for whether that's you know using AI to
00:04:54
better understand the solar potential of
00:04:55
rooftops to enable developers to deploy
00:04:57
more quickly uh or working with grid
00:04:59
operators to help them better model
00:05:01
their grids and understand what it would
00:05:02
look like to pull off a coal plant how
00:05:04
could they deploy more wind energy uh
00:05:06
and then also even understanding extreme
00:05:07
weather events more effectively which we
00:05:09
know impact clean energy Supply being
00:05:11
able to predict weather much more
00:05:13
effectively um or even being able to
00:05:16
predict flooding so I think we're seeing
00:05:18
um both a sector that really understands
00:05:20
how to drive towards deep efficiency and
00:05:22
so many use cases that will be essential
00:05:24
to driving innovation in the energy
00:05:26
sector let's pick up on the piece about
00:05:28
the actual energy that's used um as you
00:05:30
said you've made some very strong
00:05:32
commitments as a company towards uh
00:05:34
decarbonization uh quite aggressive
00:05:36
goals that are now obviously being
00:05:37
challenged by this booming Market um you
00:05:41
and uh other technology companies
00:05:43
pioneered uh the use of power purchase
00:05:45
agreements In America which really
00:05:46
helped cement the rise of renewable
00:05:48
energy in America at scale is that
00:05:51
sufficient is that tool fit for purpose
00:05:54
or I'm seeing you put forward new kinds
00:05:56
of commercial agreements uh can you tell
00:05:58
us just very briefly what what
00:06:00
differently now needs to happen given
00:06:02
that the scale of energy now we're
00:06:04
talking
00:06:05
hyperscaling uh data centers the size of
00:06:08
nuclear power plants worth of energy
00:06:10
what do you need to do differently yeah
00:06:12
so of course yes we've been um executing
00:06:15
on ppas for renewable energy since 2010
00:06:17
we have a portfolio of 14 gws so we are
00:06:20
very comfortable with that approach and
00:06:22
we need new Tools in our toolbox so one
00:06:24
thing that we did uh earlier this year
00:06:26
was We partnered with Envy energy who's
00:06:28
the utility in Nevada on a new model
00:06:30
called a clean transition tariff this
00:06:33
enables us to do direct planning with
00:06:34
the utility so they can better
00:06:36
understand what our load profile or any
00:06:38
large energy users load profile will
00:06:40
look like do that planning with us
00:06:42
understand the kind of clean firm
00:06:43
capacity we're going to need and not put
00:06:45
that burden on the rate payers so tools
00:06:46
like that are going to be really
00:06:47
critical also we want to see more clean
00:06:50
firm power deployed and so we've been
00:06:53
utilizing the advanced Market commitment
00:06:54
model and we announced a pilot of an
00:06:56
advanced Market commitment with both
00:06:58
Microsoft and new cor also this year
00:07:00
where we put out an RFI to the industry
00:07:02
showing our demand for clean firm power
00:07:04
and we're really excited to see that
00:07:05
there's been a very strong response to
00:07:07
that uh we're going to see more advanced
00:07:09
nuclear we also uh signed agreements
00:07:11
with chyos kyos is here with us today
00:07:13
here today we'll be speaking with them a
00:07:14
little bit later and you did a a deal
00:07:16
for multiple reactors in advance so it's
00:07:18
a form of advanced Market commitment
00:07:20
yeah that's an order book model so this
00:07:21
is something that the US Department of
00:07:23
energy has suggested is really
00:07:24
productive for the deployment of
00:07:25
advanced nuclear and so we were able to
00:07:27
use that model and say we have a demand
00:07:29
for 500 megawatts in 2030 uh and Beyond
00:07:33
and that's what chyos will be delivering
00:07:35
to us through that partnership so we
00:07:36
also see that as a really nice
00:07:37
Innovative model as well now why does
00:07:39
that make sense for you we all know you
00:07:41
know when there's an NF1 the first
00:07:42
nuclear power plant tends to be
00:07:44
extremely expensive just look at the
00:07:45
recent history of nuclear power the last
00:07:47
few decades and this is a Innovative new
00:07:49
technology small modular reactors uh the
00:07:52
sixth one may be a lot cheaper one hopes
00:07:55
uh how do you think about the Financial
00:07:57
Risk to your company from taking on
00:07:59
novel technology risk for the first
00:08:00
couple of plants how do you justify it
00:08:02
yeah we've been certainly very impressed
00:08:04
with chyos and the approach they've been
00:08:06
taking we've been looking at the sector
00:08:07
for a long time and we really see them
00:08:09
doing a lot of great Innovation that's
00:08:11
driving a lot efficiency and really
00:08:12
bringing cost down and also that ability
00:08:15
to use the order book model where we're
00:08:16
not just saying we're going to have one
00:08:18
PPA for a single facility but actually
00:08:20
to have that order book that also
00:08:21
creates more efficiencies in economies
00:08:23
of scale from a cost perspective and
00:08:25
this is what we need this is why we need
00:08:27
things like Advanced Market commitments
00:08:29
for clean for power we need buyers to
00:08:31
come forward who have an appetite to be
00:08:33
early adopters but also where we can
00:08:35
then get creative with the structures
00:08:36
that we use so that it can be cost-
00:08:38
effective for us as businesses too no
00:08:40
another application of this you hinted
00:08:41
at this with your comment about Nevada
00:08:44
uh is geothermal there's a lot of
00:08:46
innovation uh in this space and the Ia
00:08:48
in fact this has an important report
00:08:49
coming out next week on uh Advanced uh
00:08:52
geothermal but um can you tell us about
00:08:55
your relationship with fvo one of those
00:08:56
Innovative companies uh that you've uh
00:08:59
done a deal with what is it how are you
00:09:01
structuring that yeah absolutely so yes
00:09:03
fvo um has been a great partner of ours
00:09:05
and uh our second agreement with them
00:09:08
was actually off the back of the CTT the
00:09:10
clean transition tariff um in Envy
00:09:12
energy we're now being able to deploy
00:09:14
more um Advanced geothermal through that
00:09:17
model so we've really found fvo to be
00:09:20
excellent Partners we are very bullish
00:09:21
on Advanced geothermal we're excited
00:09:23
about the ia's report we've also been
00:09:25
doing work with project interpace around
00:09:27
mapping Advanced geothermal capability
00:09:29
in different parts of the world so we're
00:09:30
very bullish on it it is indeed as
00:09:32
you've said uh the decisive decade and
00:09:35
so we'll see if we meet in 2030 or 2035
00:09:38
how things change uh thank you so much
00:09:40
for joining us and this conversation uh
00:09:43
live from the ia's global conference on
00:09:45
I uh Ai and energy of course uh will be
00:09:48
able uh to be replayed at this website
00:09:50
at this link up next I'll be speaking
00:09:52
with uh ashis Kumar Dash who is
00:09:55
Executive Vice President at infosis we
00:09:57
hope you'll join us then thank
00:10:01
[Music]
00:10:28
you for
