How Green Innovation Is Turning Startups Into the New Climate Economy Powerhouses

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:05):
As we get started here, we're going to have each
panelists introduce themselves and then we'll jump right into our discussion.

Speaker 2 (00:13):
Perfect well, good afternoon, everybody. My name is David Hatley.
I think I'm on this panel because I'm a master
of many trades, hopefully of at least a few of them.
I'm the former CEO and vice chairman of a modular
manufacturing business called module X Solutions.

Speaker 3 (00:32):
And for you students.

Speaker 2 (00:33):
In the room or those who are not familiar, modular
manufacturing is in factory manufacturing of things that are deployed
out around the world. So our two biggest markets are
data center infrastructure and renewable power in battery storage infrastructure.
So we are serving both the supply side and the

(00:53):
demand side of the electricity and the power that we're
talking about in this conference today, and we see what's
happening on both sides of that. Prior to co founding
a business that became module X, I served in the
California State Legislature, and among my responsibilities there as I

(01:14):
was a member of the Utilities and Commerce Committee and
the Natural Resources Committee, so I had a very deep
dive in California public policy around all things electric and
then one final thing. I was an investment banker. I
was the senior telecommunications banker at BT Alex Brown back
in the late nineteen nineties when the dot com frenzy

(01:35):
was in full force. So as we look at this
investment wave, I compare with my previous experience.

Speaker 3 (01:42):
But thanks for being here.

Speaker 4 (01:43):
Hi, everyone, I'm Mark Cain.

Speaker 5 (01:46):
I am at Google, and I work on a team
that is responsible for Google's energy procurement and its engagement
with the broader electricity systems that provide us with our energy.
I'm going to speak a lot about that, so just
maybe a bit more about me. I spent much of
my career in and around government in policy, both in
the US and internationally in the UK and in Ethiopium,

(02:10):
and I currently serve also on a nonprofit board of
the Energy for Growth Hub, which is a DC based
organization that focuses on energy access and energy infrastructure in
the developing world.

Speaker 4 (02:22):
And I'll just say at the outset.

Speaker 5 (02:24):
It's so exciting to see a room full of people
focused on the business of energy. I am someone who
is excited about the business of energy and it is
just so heartening to see so many more people and
I just maybe of a frame up.

Speaker 4 (02:40):
Are we still We're still here. We can project if
we need to.

Speaker 5 (02:44):
Yeah, energy is a unique and special business. It is
not a business like any other. Depending on which economic
metrics you use, energy is about seven percent of USPDP,
but it is the seven percent that enables the other

(03:05):
ninety three percent in a very important way. And that's
true in the US, and that's true around the world.
And so I'm really excited that there's so many folks
here who are really engaged in that and excited for
the conversation today.

Speaker 6 (03:20):
Good afternoon everyone. My name is Nate Franklin.

Speaker 7 (03:22):
I'm the CEO of Pacifico Energy Group, and we are
essentially a power plant developer, so we develop.

Speaker 6 (03:30):
We've done a lot of renewable energy.

Speaker 7 (03:32):
We're actually the biggest renewable energy company in Japan, so
we've done a lot in Asia. We have an office
in Korea. We're doing offshore wind in Korea. We've done
some projects in Vietnam. One of the few companies US
companies have developed energy projects in Vietnam, which kind of
helps us punch above our weight there as far as

(03:53):
getting attention of government government officials there and even the
US for the closeness to China, and then in the
US we've uh started the company here about five years
ago and to target instead of doing the big power plants,
going to do the grid which we were.

Speaker 6 (04:11):
Doing in Asia.

Speaker 7 (04:13):
We thought with the US business, we wanted to sell
energy directly to customers, directly to the end users, and
so we started that business about five years ago and
just saw an opportunity there for supplying customers where the
utilities can supply them. And about a year and a
half ago so had financed a few portfolios of behind

(04:35):
the meter projects and then about a year and a
half ago saw the big opportunity and with data centers,
which you know is just a mega customer that we're
seeing that is having trouble and continue to have trouble
to get power from the grid. So we're developing a
number of projects at the supply behind the meter power
to the data centers. We and also too, I just

(04:59):
to the second the comments about energy for those students
looking to get into it, I think it's a great industry.

Speaker 6 (05:07):
I went to UCLA, I.

Speaker 7 (05:08):
Got my MBA there about fifteen years ago, and you
know a lot of my classmates were going to entertainment
and marketing, which is also great industries.

Speaker 6 (05:16):
But when I was looking what I wanted to do
coming out.

Speaker 7 (05:19):
I just looked at the dollars that were in the
energy industry compared to everything else, and just what a
huge part of every day personal and industrial life it is.
And you know, at the time it was wasn't seen
as cool as it is now. Is kind of one
of those boring industries where I thought, you know, that's
that's what I want to be in that you know,
boring but where you can make a lot of money.

Speaker 1 (05:43):
Well, thank you gentlemen for introducing yourselves, and we're going
to jump right into this space. Of course, you know,
this discussion is artificial intelligence, data centers, and energy, and
the big concern, of course is what is anticipated to
be x fonential growth and electricity demand at a time

(06:03):
when we're seeing electricity prices from coast to coast rising.
So my first question I actually opposed to David as
a developer, a designer and developer for data centers, can
you give us some context for all of this discussion

(06:26):
about we need a whole bunch more power. What are
you seeing as a practical matter in designing and developing
systems for data centers. Do you see this exponential growth?

Speaker 3 (06:40):
Absolutely.

Speaker 2 (06:43):
I won't speak about electricity specifically, I'll speak about the
infrastructure demand that's creating the demand for that electricity. But
just to give you a couple of frames of reference,
you know, fifteen years ago, the big tech companies were
all software companies. Fundamentally, today, the CAPEX of each of

(07:06):
the leading tech companies, whether it's Google or Meta or Oracle, Microsoft,
et cetera, exceeds exonmobiles capex in a year. We are
spending you know what We're once software companies are now
large scale infrastructure companies to meet the data needs of
their internal businesses and their customers. So the whole scale

(07:29):
of capex behind what is now becoming AI is massive.
That's that's one data point. Secondly, you know, my company,
module x is a private company. We don't release our financials,
but we have grown over five x in four years,
and our customer's biggest complaint is that we're not growing

(07:50):
fast enough. That we are scaling our business at every level,
from capital expenditures to adding people to labor saving productivity,
and we are not keeping up with the demand of
our customers. And then finally around just by a historical comparison,
the dollars we're talking about the capex the infrastructure spending

(08:11):
that's happening in this sector now dwarfs the dot com
maybe of the late nineties. The numbers are not even close.
And then just to make one comment about the durability
of this capital expenditures. For the older folks in the room,
you'll recall that that whole late nineties telecom dot com

(08:31):
infrastructure investment was really funded by capital markets. It was
venture capitalists, it was IPOs, it was high yield debt
directly or indirectly. The vast majority of the spending that's
happening now is happening out of the free cash flow
of the seven leading tech hyperscalers. So it's a much

(08:54):
more durable form of financing. This is not something that's
going to turn off off in twelve months because the
capital markets turn off. Now, these companies may conclude that
the ROI is not what they thought and they may
change their investment plans, but that's a very different discussion.

Speaker 3 (09:13):
Okay.

Speaker 1 (09:13):
So given that mark, so with the costs involved, especially
of the larger power generators, the long lead time involved
in these technologies.

Speaker 3 (09:30):
Last panels talking about nuclear power.

Speaker 1 (09:32):
Supply chain issues and those kind of challenges for the
resources like for natural gas plants. How is the tech industry?
I was Google preparing to meet electricity demand for this
exponential growth and power to support data centers and the

(09:54):
development of artificial intelligence, particularly.

Speaker 3 (09:57):
In more the immediate and over this next k M.

Speaker 5 (10:02):
So we Google are taking a really multifaceted approach to that,
and I think you're you correctly put your finger on
the timeline challenges. This is this is really rapid growth
that we're seeing now. And you know, we just had
a wonderful nuclear panel. I didn't hear anyone say that,
you know, we're going to have these plants online next year, unfortunately,

(10:23):
And so you know, we're in this interesting period where
there's a lot of really exciting new technology that is
coming down the pipeline, and there's also really significant demand
growth happening right here and right now. So a couple
of things that that we really focus on and prioritize
at Google. You know, the first is actually not on
the supply side at all. We are very focused and

(10:45):
having for a long time on operating our data centers
and all associated infrastructure as efficiently as possible, because the
more efficiently we can operate it, the less new power needs.

Speaker 4 (10:56):
To be built to meet that growing load.

Speaker 5 (10:59):
And we've really invested across the full stack, everything from
the hardware, the chips, the networking, the software architecture, the
system design, cooling, et cetera, et cetera, and it's really
yielded very significant results for us. So in the last
five years alone, we've been able to bring six times

(11:20):
more compute out of the same unit of energy as
we used.

Speaker 4 (11:24):
To, and so efficiency is really a big focus.

Speaker 5 (11:28):
It is a classic double bottom line kind of investment,
and it can really help us try to bend that
curve so that the amount of power that we actually
do need to bring in additionally is as little as possible.
We've also been applying that same kind of mindset to
AI itself, and we recently released a technical paper which

(11:50):
I'm really proud of and excited about, looking at the
energy consumption associated with a Gemini text query. And what
we found is that in the last year alone, through
a similar approach, this kind of full stack optimization across
the chips, the hardware, the software, the model architecture, et cetera.
We've been able to drive a thirty three x improvement

(12:14):
in the energy efficiency or the energy consumption for a
single median Gemini text query, getting us to a place
now where a single query we estimate to be equivalent
in energy consumption to watching around nine seconds of television,
and we are continuing.

Speaker 4 (12:32):
To push really hard on that.

Speaker 5 (12:33):
So thinking about how can we work across the entire
technical stack to minimize the amount of new power that
we do need, looking at new generation and how we're
going to bring new power online, but we're similarly taking
a really kind of casting a wide net. We have
invested in quite a lot of solar and wind. We've
done one hundred and seventy power purchase agreements for over

(12:55):
twenty two gigawatts of new capacity since twenty ten.

Speaker 4 (13:00):
Significant investments in.

Speaker 5 (13:01):
Storage, as well as early investments in some of the
newer clean firm technologies that are starting to get commercialized,
everything from enhanced geothermal to small modular nuclear reactors. We
recently announced our first long duration energy storage deal with
a company called energy Dome, and we're continuing to push
really hard to really use Google's kind of demand signal

(13:27):
to help bring these technologies out of the lab into
the market, help them get the project finance that they
need to build these early commercial projects and ultimately get
to scale and get these technologies down the cost curve
so that we can not only need that growing demand,
but also do it in a way that's affordable and
that brings broader benefits to the system.

Speaker 1 (13:47):
And you obviously emphasize efficiency, which obviously becomes a big
benefit to the grid and then all of us. But
how can the hyperscalers, how can the tech industry assure
residential consumers other businesses that their electricity costs won't increase

(14:12):
because of the ambitions of the tech industry around artificial intelligence.

Speaker 5 (14:19):
Yeah, it's a really important challenge, and it's something that
you've written about a lot recently, and if folks haven't
been reading Ivan's coverage on this, you owe it to
yourself to do so. I'm going to answer just with
a couple of examples that I think are illustrative of
the approach that we're taking. I can't speak for our
colleagues in the industry, but I understand that there's a

(14:40):
lot of thinking going into this across the industry. So
one thing that we've really leaned in on in the
last year on the technology side is thinking about how
we can make our own demand more flexible so that
less new generating capacity needs to be built.

Speaker 4 (15:00):
And there's a really great paper out.

Speaker 5 (15:02):
From researchers at Duke University earlier this year which basically
modeled the US electricity system and asked the question how
much new generating capacity could be avoided if large energy
consumers were willing to curtail their demand a certain amount

(15:22):
over the course of the year. And so a couple
of the numbers that they reported was that if a
subset of large users were to curtail their demand by
an average of zero point two five percent per year,
that would unlock seventy six gigawatts of capacity that they
wouldn't you know, otherwise, you know, be there in the

(15:42):
system and would need to be built. Additionally, if those
large consumers didn't flex their demand, if you get up
to one percent curtailment over the course of the year,
it adds another fifty gigawatts to one hundred and twenty
six gigawatts. So, you know, motivated by that kind of analysis,
we've been really focused on thinking about and operationalizing how

(16:06):
can we make the consumption at our data centers more
flexible so that for those one hundred two hundred hours
a year where there really is peak demand on the system,
we can actually help meet that gried wide need through
a demand side flex instead of building new generating capacity

(16:27):
to be able to meet a higher peak. And the
economics of that are really good. We've implemented that and
are working closely with Indiana, Michigan Power, and TVA on
two initial pilots to really build that capability into their
resource planning and their long term power generation investment strategies.

(16:51):
And so we're really excited about that, and I hope
and expect to see quite a lot more happening in
that space, both from Google and hopefully from others. But
then to to the broader point on making sure that
that local rate payers are protected, it's really important, you know,
as I said in my opening, you know, energy is
really the lifeblood of the economy as well as society,

(17:13):
and the more people have to pay for it, you know,
the less they have available for whatever other needs they
have in their life. And the same applies to businesses.
Money that they're spending on their energy bill is money
that they're not paying their employees are investing in new production,
and so it's really a high priority for us to
be a good grid citizen in the place where we operate,

(17:35):
to make sure that we are helping offset any local
costs that they might be escalating because of the increased
demand from our power.

Speaker 4 (17:44):
There are a number of forms that that takes.

Speaker 5 (17:45):
We've we've made a lot of direct investments in new
generation just to make sure that that we're financially responsible
for the new capacity that's coming online. We've also done
a number of things on the regulatory front. We have
a a regulatory model that we've built and have now
rolled out in West Virginia, Georgia, and Indiana that we

(18:06):
call the Capacity Commitment Framework, And this is basically a
regulatory compact that we create that puts us on the
hook for any necessary grid and infrastructure updates that need
to happen to be able to meet the load that's
coming online at our data centers, and it protects the
financial stability of the utilities in addition to the rate payers.

Speaker 4 (18:29):
And so that's just one example.

Speaker 5 (18:31):
We have a few other examples like that, really deep
in the kind of very wonky regulatory space, so I
won't get too into it. But then there are a
few other programs that we've done that we'd like to
do more of and see more of across the industry,
just around working in these communities on some really high
value investments that can be made in things like home weatherization,
distributed resources on homes and small businesses. One of the

(18:55):
pilots that we've done, or it's more than a pilot,
that's a big program with SOUL Systems in North Carolin
in South Carolina. We've just gotten the initial results back
from the first year, and we found that there was
an average of five hundred and eighty dollars of annual
savings on utility bills from the homes that were participating
in that program. And so you know, it's going to
be a combination of things ranging from technology to utility

(19:19):
regulatory work, ranging to actually just getting into the communities
and figuring out what's needed to be able to kind
of decouple the growth in this data center and AI
demand from the impacts that that might have on ratepayers.

Speaker 3 (19:36):
So let's draw naden here. You're as a power plant.

Speaker 1 (19:43):
Developer, you are actually offering some solutions apart from mid connections.

Speaker 3 (19:52):
What does that look like?

Speaker 7 (19:54):
Yeah, so it'd be interesting to see how the market develops.
Our view is just I mean, the demand that we're
we're witnessing, you know, even today and kind of what
it's forecast in the future is you know, unprecedented and
kind of just like a shift change in the market.
And our view is that, like there's definitely going to

(20:15):
be some solutions as far as you know, something like
that you mentioned or familiar with that du the Duke paper.

Speaker 6 (20:21):
I think some of that will happen.

Speaker 7 (20:22):
But we just think that the util the supply is
not going to keep up with the demand. And it
really is a race, right, Like we can't like pause
and will like it will wait for these kind of
solutions to materialize or to pan out and for a
lot of people to work together and make things happen,
and wait for utilities. I feel like nothing against utilities.

(20:44):
I've been working with uties my whole life. I have
like PTSD from working with the utilities, just with the
you know, because it's difficult and I get it they
have a tough job. But if the counterparty there's they wrote,
you have no recourse against what they say or what
they say they're going to do or timelines that they
commit to, and so it can be a really of
shreading process. But yeah, we we just see that, you know,

(21:05):
demand's going to outstrip supply for the foreseeable future or
outstripped the supply that the utilities are going to be
able to to to give. And so where we see
that gap, what's filling that gap is behind what you
call it behind the.

Speaker 6 (21:20):
Meter generation or off grid.

Speaker 7 (21:22):
What we're really calling it is like private grid because
at the size that these data centers are are getting
developed that now when you're like two three, five hundred
megawats data centers, you know, gigle WoT sized data centers
or of power of power need I mean at five
hundred megawats, you're bigger than eighty percent of the utilities
in the US. So just to kind of give you

(21:42):
a sense of the scale this, you know, we are
working with the utility where they're actually going to sleeve
power where it's like an off grid configuration, but they're
going to stand in between just for a regulatory issue.
You know, you look at their tiff and for new
load interconnects, their terriff tops out at two megawats, So
like you know, there's all the different levels, and then

(22:04):
at two megawots, I mean then it's anything above two
megawats is a different one.

Speaker 6 (22:08):
So when you're going.

Speaker 7 (22:08):
To utility and say I need five hundred megaots of power,
I mean, that's just that's like it just hasn't been done,
you know. I mean it's been done in very few instances,
you know, a refinery or chemical plant, but very few
instances in the US. So when you're needing like that
much power in one place, we just don't think that
the response is going to be really quick to supply it. Meantime,
the race is on, you know, for for s. So

(22:31):
we just see that the gap being filled with essentially
what we're calling private grids, where you know, you develop
gas co locate it a lot of energy storage. You're
gonna need to get the reliability numbers you need.

Speaker 6 (22:46):
I think, uh, you know, one of the one of
the concerns with building.

Speaker 7 (22:51):
You know, first of all, I feel like it's a
concern because it's something kind of different or new. Right,
everyone's usually getting power from the utility. It's known, we
follow the interconnect you can trust that they're going to
give the power. You know, eventually this is something new.
But if you look at the math and the engineering,
like our reliability could be better than the utility. I mean,
we're building, we're designing these plants that what we call

(23:13):
N plus two configurations, so essentially two extra turbines than
what is needed. I mean, anyone that gets on an
airplane puts their life on the line. At L plus one,
you know, you have one, you're having one extra turbine
on the on the airplane. So the plants are really reliable,
you know, more reliable than the utility, and we just
see it as being a solution going forward.

Speaker 1 (23:32):
So then you're you're mentioning natural gas. How much of
a role do you see natural gas playing in this
in this immediate moment, until we get to small modular reactors,
until we get to some of the other technologies that
that are well, we'll get to constraint issues in a minute.

(23:53):
But how much of a role do you see natural
gas playing immediately?

Speaker 7 (23:57):
Yeah, I mean we definitely see it plan a big,
a big role because especially you know in the in
their turn i want to say, in their term, in
the next five to seven years to meet this this demand, Uh,
just because I think someone mentioned today you know, it's
not a kill what hour issue we have, it's a
kill what issue like so in other words, like it's
capacity mark I think you mentioned that, right, Yeah, so

(24:19):
it's it's it's a capacity issue. And renewables are great.
We've developed a lot of renewables you know, in different
markets in our company, but it doesn't give you the
it's not that's doesn't give you the capacity you needs.

Speaker 6 (24:30):
The Really the only way to get that is.

Speaker 7 (24:31):
You know, energy storage will play a role, but you know,
firm dispatchable power is what's needed to meet this demand
because we're seeing across the industry or across you know,
in all markets in the US, like they're out of capacity.
I mean, you look at the capacity prices in PGM
just recently, you know, they're ten x from what they
were a year ago. And some of the regulators are

(24:54):
coming or the ISOs are are starting to implement rules.

Speaker 8 (24:57):
You know.

Speaker 7 (24:57):
We know SB six is a new law in Texas
that hasn't been the details. It's a lawn now but
the details haven't been finalized. But essentially, any new load
over seventy five nagalots has to bring their own generation
because they just they know, like they don't have the
capacity to serve it. There was a board proposal in
PGM just last month that said any new load has

(25:19):
to be curtillable, you know, And there's a bunch of pushback,
but it's essentially saying the same thing that we just
don't have the capacity to connect this load.

Speaker 1 (25:28):
Well, so, and I'd like to hear from all three
of you a bit on that constraint issue. What do
you all see as the constraints to meeting this growing
demand for data centers, both from a developer standpoint and
then also from the technology industry position.

Speaker 3 (25:51):
Well, I'll go first.

Speaker 2 (25:52):
I do want to make one comment on a previous
question before I go to that, and that goes back
to the affordability question. And I appreciate what Mark and
Google and the other big tech companies are doing to
manage their demand and reduce costs and be efficient in
as many ways as they can, but we shouldn't hide
the ball. Public policy is the dominant factor in the

(26:17):
electricity cost equation Californians. Californian rate players pay over two
times Texas rate payers per kill a lot hour for
electricity California commercial rate payers pay over three times Texas
rate payers for commercial energy. Florida is very similar to

(26:40):
Texas on residential, not as much on commercial. So I
don't want to let utility companies. I don't want to
let independent power providers. I don't want to let tech
companies off the hook. But it is state public policy
that's determining electricity rates and affordability issues. But you know,

(27:01):
throughout the United States, on the question of how we
meet this capacity that AI is generating. You know, the
business my module X, our module manufacturing business. Our work
is typically in the twelve months before a renewable power
facility is commissioned, so we see kind of the late stage,

(27:23):
permitting has been done, planning, We help on the planning,
but we don't actually turn a wrench until the last
twelve months before something is about to be commissioned. And
notwithstanding the changes in the One Big Beautiful Bill relative
to the Inflation Reduction Act and so on, we have
a very robust pipeline of renewable power projects. And I

(27:45):
believe part of the reason is that, you know, permitting
and so on has been a recurring theme in today's discussions.
These projects are underway, they are able to be built,
and to meet the needs of the data center companies
and utilities, they are going to be built. So I
think the ones that can be built in the next
couple of years are overwhelmingly renewable power, battery storage, and

(28:09):
natural gas. You know, you get to the nuclear, the geothermal,
and some of the larger scale recommissioning of nuclear cliants
and so on, it's just not going to happen quickly.

Speaker 5 (28:21):
Helpful to weigh in on the constraint question of just
what would help get all this built faster? You know
what we see and this has been talked about quite
a lot today, certainly permitting interconnection. It takes a long
time to build new infrastructure in this country, and there

(28:43):
are a number of different potential fixes that can happen,
many of them at the policy level. But it's not
just federal policy, it's state policy. There's a lot of
local veto points that come in when a new infrastructure
project is being proposed, and it adds really material delays
to the actual you know, build out of all this stuff.

(29:09):
And so I think, you know, we've talked a lot
and we've heard a lot about you know, the unprecedented
nature of the demand growth that we're seeing now, and
I'm just going to, you know, provoke the panel by
pointing to some data that we were talking about before,
which is that, you know, even in a pretty maximalist
projection that was recently public publicized by the Rodium Group,

(29:34):
which is an economic and energy consultancy, they estimate that
we'll see about one point four to one point nine
percent annual demand growth in the electricity system out to
twenty forty now compared to the last twenty years. That
is unprecedented because we have barely grown electricity demand in
this country in all over.

Speaker 4 (29:53):
The last two decades.

Speaker 5 (29:55):
But if you look back to the nineties, which wasn't
so long ago, we sustained two point two percent annual
growth in electricity demand and we're able to meet that
with new supply and other solutions in a way that
didn't generate the same kind of feeling of crisis that
we're in now. And now a lot has changed in

(30:16):
the energy system and elsewhere, and of course, you know,
the demand that we're seeing now tends to be pretty
geographically concentrated in certain places, and so you may see
more than one point four to one point nine percent
in certain geographies that are having a lot of data
centers built, for example. But I think it's a really
important question that we all need to ask ourselves. You
know what, what has changed in the last couple of

(30:38):
decades that has made it so hard to build new
infrastructure in a timely manner, whether it's new generation or
transmission lines or or you know, distribution, And how can
we how can we kind of unblock that so that
we can get these these projects on track and delivering
more quickly.

Speaker 3 (30:58):
We can?

Speaker 1 (30:59):
We can we you ride on that policy question a
little more than I mean the complexity of the power system.
Part of it is that there isn't one entity that
we go to for all of these permits or approvals.

(31:21):
There are state regulatory commissions, there's the Federal Energy Regulatory
Commission that's overseeing interstate transmission.

Speaker 3 (31:31):
Then when you start crossing multiple states, then you're dealing
with the states, And how do we navigate a space
where there are so many.

Speaker 1 (31:43):
Players and complicated issues to result. I mean for us
to say that, well, okay, so permitting is a problem,
but the reality is we have a whole bunch of
different players. There's not a singular entity in charge? How
do how do we deal with all this?

Speaker 2 (32:04):
Well, I'll take a very quick stab at it, you know,
Sean Hyatt tells me. And it may not be true
because we don't ask everybody everything about their history. But
I believe I'm the only current or former elected official
on any of the panels today, and I was in
a highly contentious seat, a very swing seat, and energy

(32:26):
and environmental policies were right in the middle of it.
I spent a ton of time in the legislature on
these issues, and I.

Speaker 3 (32:34):
Have bad news.

Speaker 6 (32:35):
I have no good answer.

Speaker 2 (32:38):
That the complexity is overwhelming. I would say a few
very simple things. The grid itself is vastly more complicated
today than it was in the early nineties. And when
Mark was referring to the two percent growth rates that
we sustained in the nineties without this, you know, but

(32:59):
I wouldn't say that there wasn't trauma at the time.
Gray Davis was recalled as governor of California in two
thousand and three for multiple reasons, but one of them
was a botched electricity deregulation that led to rapidly rising
electricity prices for California rate payers. So this is you know,

(33:20):
this has happened. We may not remember it, but some
of this has happened. I also, I can tell you
that at a very big picture level, the US has
spent forty years offshoring industrial capacity generally, and we are

(33:42):
just now in the last five ish years beginning to
re shore things. And we're taking, I believe, a more
strategic approach in some ways to what we need as
a matter of national security and control over our own destiny.
You know, the mining discussion this morning was a great
example of a little bit of that.

Speaker 4 (34:02):
Uh.

Speaker 3 (34:02):
But we outsourced uh forty years.

Speaker 2 (34:06):
Of talent of capital, and and bringing that back to
the US in a way that we can meet those
capacity needs like we maybe could have in the early
nineties is going to be a very formidable task.

Speaker 4 (34:19):
Well.

Speaker 1 (34:20):
So, so then are we talking about Nate's uh basically
the rise of the private grid, the virtual power plants,
the micro grids. Uh, it is this moment sort of
thrusting us into the private grid, the smaller grids.

Speaker 7 (34:41):
Make sure think, yeah, I would just say one more
thing to add to like the public policy, like you know,
this is you know, this is beyond just like business
right or like wanting to you know, us have more
you know, better answers to our chat questions.

Speaker 6 (34:58):
You know, Judge, this is like.

Speaker 7 (35:01):
This is you know, really like a national security issue,
is what you know, the administration of the country is
looking at it like, you know, this is like I
keep going back to an AI race and I think,
you know, the last thing we need is, you know,
we're building these data centers and all of a sudden,
everyone's power prices are going up. Because you know, we
can talk about all the nuances of public policy, but

(35:23):
at the end of the day, when demands a stripping supply,
I mean it's econ one to one prices are going
to go up, right, And so you know, the last
thing we need is like a public backlash against you know,
data center buildout or AI and then and then you know,
then it just makes it that much harder for you know,
for the US to to excel in this race. So

(35:45):
you know, going back to our solution, you know, private
grids that alleviates that because essentially the users paying you know,
directly for the power that they're are you know, paying
directly for the power they're using.

Speaker 1 (35:57):
And Mark, how how does the industry feel about that
we wrote about how the tech industry is increasingly.

Speaker 4 (36:07):
Involved in every aspect.

Speaker 1 (36:09):
Of the electricity business, and so it seems grid connection
is actually part of the vision as opposed to.

Speaker 3 (36:21):
More private grids. But how is the industry, how is
Google seeing this picture?

Speaker 5 (36:27):
It's a really interesting moment and I wish I had
a crystal ball because I could see this going in
a number of different directions and I think the jury
is still out on where it's going to land. And
I see looking across the industry, different players are adopting
different strategies. At Google, our preference has generally been to
connect to the grid and to source our power from

(36:50):
the grid and through power purchase agreements that are delivered
to us through the utility. And I expect that that
will continue to be a very significant share of how
we go about meeting our demand growth. We have been
exploring other models. We announced a partnership with Intersect Power
to look at how to sort of code design data

(37:15):
centers with large sort of direct generation solar, mainly with
storage for the firming, and to think about, you know,
is there a model where you can develop a kind
of industrial park that brings its own power and you know,
we anticipate that those parks would still be grid connected,

(37:38):
but in terms of how much they're relying on the
grid for their prime power versus those behind the meter
resources or front of the meter resources, it would look
different than a you know, a traditional data center that
gets all of the power from the grid. So you know,
that's been our approach is, let's try to do this
through the grid. Let's work really closely with our utility
partners and with project developers to do this in a

(38:00):
way that really takes into account the full complexity of
the system and doesn't kind of optimize locally in ways
that ultimately kind of undercut from the broader system. But
there are a lot of different models that are being
tried out there, and you know, Texas is really the
kind of epicenter of this right now.

Speaker 4 (38:18):
There there are a number of different.

Speaker 5 (38:20):
Developers, both hyperscalers but also smaller data center developer companies
who are, yeah, you know, approaching it just as let's
find a place that has fiber and land and you know,
we'll build our own power supply, we'll build the data center,
and you know, I expect we'll see a lot more
of that too.

Speaker 1 (38:41):
What if in the immediate as the competition over data centers,
over artificial intelligence gets tighter and tighter, and especially as
the US is competing with China and you just need
more power right now, what do you do?

Speaker 5 (39:00):
It's a great question, and it's you know, timely and relevant,
right I think that you know, depending on who you
ask in the industry, they might.

Speaker 4 (39:08):
Say that we're already there.

Speaker 5 (39:10):
You know, for us, we have been able to grow
in a way that it has has has not put
us in that sort of emergency state where you know,
we just have to you know, have our hair on fire,
so to speak, and you know, make really rash decisions.
You know, we're being very thoughtful about where we're siting
new data centers, looking at opportunities to expand existing ones

(39:33):
where there's you know, slack capacity on the system, where
the grid can actually meet that that growing demand. But
it is I think, you know, not just for us,
I mean for everyone. You know, it is a really
pertinent question of what happens if if if the grid,
the system, the investments just can't keep up. And you know,

(39:54):
to the point that was raised earlier, I think that
that is a situation that we should all really want
to void from a you know, national security point of view,
from an economic competitiveness point of view, from a you know,
residential power supply and and and affordability point of view,
which I think, you know, just takes us back to this,

(40:15):
this this set of questions around policy and investment and
what do we need to be doing now to make
sure that we don't end up in that situation and
and and you know, whether that's at the state level
at the federal level, how can we how can we
sort of look a little bit down the line and
start making decisions now that that get us onto a

(40:36):
path where we don't end up having to make really
hard decisions or have those decisions made for us. You know,
it was mentioned that you know, there there have been
bills in state legislatures around you know, kill switches for
large industrial customers, and and you know, that's a really
bad state of affairs, not just for us, but for
for anyone. You know, as I mentioned earlier, electricity is

(40:57):
is really the key input into a lot of economic
activity that supports a lot of jobs and investment. And
we really don't want to be in a situation as
a country and as a society where you know, we're
having to throttle that because we haven't planned, we haven't prepared,
we haven't invested.

Speaker 3 (41:14):
So they've a singular policy that you think would help.

Speaker 2 (41:18):
Well, you know, the state of New Hampshire just an
active legislation just signed by the governor that if electricity
capacity is built but does not connect to the grid,
it's not subject to Public Utility Commission regulation. And I

(41:39):
think that's an important safety valve in this conversation for
two reasons. One is that I appreciate what Mark and
Google are doing to support the grid and not and
be efficient in that way. But I also do not
want the United States falling behind or outsourcing AI because
Google can't meet their needs in the US. And then secondly,

(42:01):
there's a huge aspect of risk here. I mean, for
those of you who are able to be here for lunch.
You know, Robert Friedlan basically said AI is a dead
man walking. We don't have the power to support it.
It's not going to be economic, et cetera. As a
public policy matter, we cannot allow rate payers to be
on the hook for tens of billions, hundreds of billions

(42:22):
of stranded electrical generation assets that are built to support
AI that doesn't fully materialize over the coming decades. So
I think supporting the grid is great, but having an
off ramp to protect both rate payers and AI companies
is an important public policy element to this.

Speaker 3 (42:42):
Well and.

Speaker 1 (42:44):
The natural gas and along the lines of those stranded assets.
Is there something being done so that we can bridge
the gap between now and a depreciation schedule of forty
years for for for a gas asset to transition, so

(43:09):
we don't get saddled with those costs in the out years.

Speaker 7 (43:16):
Yeah, I mean I think you know, the way we're
modeling the projects that we're developing, I mean really we're
looking at you know, ten or fifteen year PPAs, so
essentially after that, you know, the investors are paid back,
the debts paid back at the project could essentially be
you know, mopball at that. But like you we don't
need like the full fury years based on I think

(43:37):
where the price of energy would be. So yeah, you know,
I really see it as like this is like a
bridge solution to get to kind of where some of
these other solutions that everyone's been talking about today. You
know that mark has been mentioned, like I think you know,
of course nuclear like then where that can take over
and then you do have like this, you know, the

(43:57):
clean you know, non emission.

Speaker 6 (44:00):
Super efficient.

Speaker 7 (44:01):
But I think, you know, we need a bridge to
get there, and you know, we need to build the
data centers and you know, build have the power to
build the smarter models. That's going to solve a lot
of these issues for us.

Speaker 1 (44:13):
Well, thank thank you, gentlemen, and we're going to actually
we're going to bring you all into this discussion. If
there are some questions from the audience, we are happy
to entertain them. Yes, and we have a student, Okay, yes, jeh.

Speaker 9 (44:32):
I question from Mark just kind of curious how much
of the cost of new data centers does energy, the
cost of energy make up as a general sense.

Speaker 5 (44:48):
It's a great question, and I don't have you know,
precise figures at my fingertips, but what I understand to
be the case is that in terms of the capital investment,
which is you know, the first hurdle in building a
new data center, you know, that's mostly land and building
the large shell of the building, networking it a lot

(45:11):
of fiber and cabling, heating and cooling, and then filling
it with racks and chips and so the sort of
total cost of just getting it commissioned and up and
running is actually really is really about just the sort
of capital equipment that that goes in there, and then

(45:32):
the energy really kicks in once it's commissioned and it
becomes really part of the op X for running the
data center long term. And so you know that that
would probably be the way that I would think about
it is you know, in terms of the commissioning and
the and the capital stack to build a data center,
energy doesn't you know, figure hugely into that picture. But

(45:53):
then as the data center becomes operational, then the energy
cost to just keep it running, you know, is a
portion of the just overall operating cost over the years
that it's in operation.

Speaker 3 (46:09):
Yes, gentlemen at the table.

Speaker 10 (46:15):
Hey, quick question, and this is also for Mark. So
given in the capital expenditures that's going to go into
creating these data centers and not just one, but you know,
tens of thousands of them across all the tech companies,
what are the motivations that the tech companies have to
willingly give up the twenty five to one hundred hours
a year in computing at those specific specific data centers, Like.

Speaker 4 (46:40):
It's a huge capbex.

Speaker 10 (46:42):
So, like, what motivation do the tech companies have to
just willingly give that up?

Speaker 4 (46:48):
It's a great question.

Speaker 5 (46:48):
So you're asking about the demand response piece that I
mentioned earlier. Yeah, so I think the most pertinent one
is that if we're able to, through that that flexibility,
avoid the need for new generating capacity for which we
would be financially responsible, then the numbers start to look compelling.

(47:11):
So you know, it's it's of course a trade off.
You know, you have to you know, ramp down an
economically productive asset for a period of time. But because
that amount of time is relatively small, and because it
may and in some of the systems, reduce the need
for you know, hundreds of megawatts of new generation, which
is very expensive and also very time consuming to build,

(47:35):
it starts to be a compelling case. And I think
the time piece is also important. In addition to the cost.
These are solutions that we can you know, start to
implement very quickly, whereas you know, if we need to
bring on that new capacity to be able to grow,
we may be you know, waiting five years to to to.

Speaker 4 (47:53):
Be able to do that.

Speaker 5 (47:55):
And so this is why also it becomes a really
important kind of partnership with the utility to make sure
that we're really factoring this into the broader system planning
that they're doing and can kind of kind of co optimize,
let's say, for how we're interfacing with the grid, both
from an operational point of view, but also from a
kind of long term resource planning and investment point of view.

Speaker 8 (48:20):
A gentlemen in the back, First of all, thank you
all for taking the time today discussed this very important
topic that's probably going to be a part of our
future for a long time.

Speaker 4 (48:31):
One of the big themes, at least in my mind
that I.

Speaker 8 (48:34):
Was taking away from what we were discussing was there's
a lot of demand, not of supply, and there are
different steps being taken, and it's going to be difficult
to incorporate in technology for a long time, and there's
optimization on the actual kind of end user tech stack.
But I was wondering, is there a current standard for
a power supplier actually collaborating with someone who's operating a

(48:57):
data center to make sure that the dator center is
running as optimally as possible from an energy perspective, and
if not, and if this is feasible, should this be
something that's actually incorporated into public policy and a requirement
moving forward until we have the necessary technology to supply
this point demand.

Speaker 6 (49:19):
Should I can take that one?

Speaker 7 (49:20):
Yeah, So, I mean I think we're starting to see that,
and you know, I mentioned the Senate the new.

Speaker 6 (49:25):
Law in Texas at SB six. Essentially that's it's requiring.

Speaker 7 (49:30):
What you're you're mentioning as far as any new industrial
load that gets built, they have to have their own
generation to be built along with it, and then they
also have to be willing to be curtailed at certain times.
So that's kind of the demand response, you know, in
other word for that kind of the demand response where
they can they can uh, you know, ramp down their

(49:52):
power need. So you're seeing you're starting to see laws
that are like implementing that. The issue is like, you know,
a lot of these users of power, the data centers,
like the GPUs that they're putting in the they're like
thirty million dollars a mega lot, and you know, building
the actual data center is like ten million may just
kind of rough numbers, so extremely expensive equipment. You know,

(50:16):
this is why in videos worth like five trillion, right,
it's because of these GPUs. I mean, they're really expensive
and they'd appreciate like twenty five percent a year. Right,
So you have this extremely expensive semiconductor that's they're appreciating
like very fast. So like how much do you really
want to turn off in that? And I know Google
has a view on that, and then working on what
they can do to to to ram down the power

(50:39):
when they when they need to. But yeah, I think
you're starting to see that also PGM I mentioned that
they had they had their board had proposed that all
new load becomes curtailable. I think what they're latest that
that we've read on that market is that they're instituting
some new regulations laws where they're like pricing demand response,

(51:02):
where you know, at certain hours you get paid a
lot more to not use the energy. So yeah, there's
definitely mechanisms that are that are coming.

Speaker 11 (51:15):
Oh yes, So question for David so I work. I
work at Southern California Edison and Transmission Planning, and I
noticed that every year building transmission lines gets incrementally harder.

(51:36):
So my question is is there any specific policies or
maybe states where that it's not getting more challenging to
build out new transmission but maybe getting a little bit
easier things that could work, like here in California or
anywhere else in the United States.

Speaker 2 (51:55):
It doesn't feel like it's getting easier to build anything anywhere.
There are there are degrees of difficulty. Just ask one
follow up question, are you speaking primarily of intrustate or
interstate transmission? Because if you're talking about interstate transmission, I

(52:16):
do think that there is some appetite in h the
administration and in Congress.

Speaker 9 (52:23):
Uh.

Speaker 6 (52:24):
You know, I don't want to.

Speaker 2 (52:25):
Be too optimistic about it, but I I do think
that there are there's enough recognition of the scale of
the challenges that there is some bipartisan energy around reforming
interstate transmission. But as in most things in Washington, I'll
believe it when I see it.

Speaker 12 (52:50):
Thank you, yes, yeah, Thank you for talk us AI
system and that the centers shop or shape or society
and technology.

Speaker 13 (53:10):
How can policy makers and reviewers and workers work together
to ensure that revelation and education evolve fast enough to
manage both innovation and impact.

Speaker 2 (53:29):
Uh, I'll take I'll take the first dab. I.

Speaker 6 (53:34):
It is.

Speaker 2 (53:37):
Probably my biggest single lesson coming out of my time
in the legislature is UH is humility I think that
you know, AI and electricity, uh and the cost of
that and the availability is one of you know, hundreds

(53:57):
of challenging issues facing our society, and I think in
many ways some of our political polarization it comes from
the fact that we are asking too much of our
political institutions, particularly in a world where we're talking here
primarily around a b to be phenomenon. We're talking about,

(54:20):
you know, some of the largest, most sophisticated companies in
the world, like Google, and we're talking about very capable, sophisticated,
well financed power providers like Nate's business. And so I
think that the one of the things that the public

(54:40):
policy can best do is focus on protecting the public
and so think about these issues primarily from a ratepayer standpoint,
primarily from from a pollution environmental carbon policy standpoint, and
allow the Nates in the marks of the world to

(55:01):
do business that you know, I think if we ask
our government institutions to solve all these challenges, we're not
going to be happy with the results.

Speaker 5 (55:13):
Only thing I'd add to is just to say that
it's a great question. It's a big question.

Speaker 4 (55:20):
There are so.

Speaker 5 (55:20):
Many different aspects of the economy and society that are
implicated in these trends that we've been talking about on
the panel and really all day. And there are a
number of areas that I think there are huge opportunities
that are not being seized and invested in as much
as they can. You know, for example, we've heard a
bit about workforce and about the opportunities that all of

(55:43):
this infrastructure built out might create. And you know, you
see statistics about the amount of electricians that were forecasted
to need in this country in the next five ten
years and how many we currently have and how many
were capable of training.

Speaker 4 (55:57):
And it's these kinds of issues that seem to me.

Speaker 5 (55:59):
To be really important public policy questions that no individual
actor in the economy is gonna be able to solve
on their own, but that you know, working together with
the right kind of support from government, whether that's federal
or state, with companies like Google who are willing to
invest in you know, education programs and have a long
history of doing that, creating you know, apprentice and other

(56:21):
training opportunities at developers and other companies who are who
are really building the infrastructure to.

Speaker 4 (56:27):
Meet this moment.

Speaker 5 (56:29):
It seems to me that there's there's a lot of
opportunity there, and so I don't know that I have
a great answer to the question other than I validate
the premise of it and I encourage all of us
to be thinking more about about how to answer it.

Speaker 3 (56:44):
Yeah, I think we probably have time for one more question.
Gentlemen in the back. Oh oh, I'm sorry, I'm sorry.

Speaker 14 (56:54):
Thank you so much for sharing your insights with us.
Greatly appreciated as a student. Learn a lot just from
the panel discussion, I wanted to specifically ask about hyperscaler
strategy for data center operations currently with the given demand
and the demand to you know, win the AI race
with geminized growth. What does sort of the fuel composition

(57:17):
like look like and what rule does like natural gas
play as a part of like data center operations, And
what I wanted to ask was a big part of
data center operations is like consistency that firms require that
they operate you know, twenty four seven and you can
have foresight into the reliability of operations. And are like

(57:41):
hyper skillers willing to pay like paymiums for that or
do they prefer to pay like spot market rates going forward?

Speaker 4 (57:50):
Thank you sure.

Speaker 5 (57:51):
It's a great question, and I can speak for Google,
maybe less so for for for others in the industry.
But but from our point of view, you know, we
we have a pretty kind of inclusive approach to technology
and the company set some very ambitious carbon free energy
goals that we we are are you know, really investing in,

(58:12):
and so there are certain technologies that are off the
table for us. But within the broader category of carbon
free energy technologies, we've we've you know, done, as I mentioned,
quite a lot in solar wind storage, both luthium ion
and also kind of newer, longer duration storage technologies. We've

(58:33):
invested in in small modular nuclear reactors and enhanced you
thermal We you know, really are are open to two
different technologies. Historically, we have been willing to pay a
little bit of a premium for those technologies, recognizing that
they are you know, early in their their maturity. And
we really see our role in our ambition with respect

(58:55):
to these newer technologies as helping bring them down the
cost curve that they can be more adoptable and scalable
across the broader system. And so, you know, part of
our program is really identifying what are the really promising
technologies that are maybe five ten years out, and how
can we use the demand signals that we can send
to the market to prove out these technologies to get

(59:17):
the first projects built. We try to structure these these
these deals for these emerging technologies as an order book
so that it's not just one project, but the counter
party you know, has a few of them, and so
they know that if they invest in the production capacity
or the workforce to build that first project, there there's
going to be a second project and on from there.

(59:38):
And so yeah, I don't know exactly what the what
the final mix will be in the end, but I
suspect for us and for others you'll see quite a
lot of of all of the different energview technologies and
generating technologies, but also hopefully a lot on the demand side.
And then and then some other things that we haven't
talked about. Back to the transmission question. You know, there's

(59:58):
a lot of really exciting technol that we're seeing come
online in terms of uprating existing transmission lines, reconductoring, dynamic
line rating, that sort of thing, and so we need
it all and we need it all fast.

Speaker 1 (01:00:13):
Well, thank you all. I think our time has ended.
We appreciate you spending time with us this afternoon.

Speaker 10 (01:00:21):
Thank you,

All Episodes

Episode Transcript

Popular Podcasts

Las Culturistas with Matt Rogers and Bowen Yang

The Joe Rogan Experience

Stuff You Should Know

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}How Green Innovation Is Turning Startups Into the New Climate Economy Powerhouses

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}Las Culturistas with Matt Rogers and Bowen Yang

The Joe Rogan Experience

Stuff You Should Know

All Episodes

How Green Innovation Is Turning Startups Into the New Climate Economy Powerhouses

Las Culturistas with Matt Rogers and Bowen Yang