Off-Grid Data Centers

In this episode of "Full Tech Ahead," host Amanda Razani interviews Reza Azizian, CEO of Ferveret. 

They discuss the company's innovative liquid cooling technology for data centers and AI factories, which is inspired by nuclear reactor cooling. 

Azizian highlights a recent benchmarking study conducted with UCLA, revealing that Ferveret's adaptive cooling solution delivers 35% more compute power from the same power envelope compared to traditional methods. Furthermore, the discussion emphasizes sustainability; while a typical 100MW data center consumes the equivalent of 4,500 Olympic swimming pools of water annually, Ferveret’s closed-loop system consumes essentially zero water. 

This technological breakthrough aims to make the "AI revolution" environmentally sustainable and enables the deployment of off-grid data centers in hot, arid regions.

Key Quotes
"We developed liquid cooling technologies for data centers and AI factories, which is inspired by nuclear reactor cooling."
"...in a 100 megawatt datacenter, that can translate into almost 230 million of more revenue per year, which is quite significant."
"A typical 100 megawatt data center can consume up to 4500 Olympic swimming pools of water per year for cooling... our solution basically works with a closed loop. So the water consumption is basically almost zero."
"We are here to help to make the AI revolution sustainable."

Takeaways

Massive Efficiency Gains: A recent UCLA study proved that Ferveret's cooling solution provides 15% more compute at the server level and a total of 35% more compute at the facility level using the exact same power contract.

Zero Water Waste: Traditional data centers rely on evaporative cooling, wasting massive amounts of water. Ferveret utilizes a sealed, closed-loop system that doesn't evaporate water, drastically reducing the environmental footprint.

Unlocking New Locations: Because the system operates efficiently without consuming water, it allows companies to build data centers in drought-prone, highly regulated areas (like Arizona or Nevada).

Off-Grid Potential: The efficiency of this cooling technology makes it financially and operationally viable to pair data centers entirely with solar power, creating truly sustainable, off-the-grid facilities.

Find Amanda Razani on LinkedIn.  https://www.linkedin.com/in/amanda-razani-990a7233/

Follow the FTA LinkedIn Page: https://www.linkedin.com/company/full-tech-ahead/

Visit the FTA website: https://fulltechahead.com/

Check out the Substack Channel: https://fulltechahead.substack.com/

Episode Transcript

SPEAKER_00 (00:19):
Hello and welcome to Full Tech Ahead.
This is season two, and I'm soexcited to be here today with
the CEO of Ferveret, RezaAzizian.
How are you doing today?

SPEAKER_01 (00:32):
Good, Amanda.
Nice to see you.

SPEAKER_00 (00:34):
Nice to see you as well.
Can you first off share with ouraudience what is fervirate and
what does it mean?

SPEAKER_01 (00:43):
Sure.
Ferverate is a Latin word forboiling.
And that's a kind of fundamentaluh process that we use in our
technology, and that's how wecame up with the name Ferverate.
Uh by the way, I would becurious to know what you would
have thought about the name andlike uh what you were thinking
the name means when you when youheard it first.

SPEAKER_00 (01:06):
I really wasn't sure.
That's why I asked.
What I can you share a littlebit about what your company does
then based on that name?

SPEAKER_01 (01:15):
Absolutely.
Um, so we developed liquidcooling technologies for data
centers and AR factories, whichis inspired by um nuclear
reactor cooling, interestinglyenough.
And I can kind of always giveyou the background, the story of
how we got to started and uh andhow we are here.
But in a nutshell, that's whatthat's what FairWet does.

SPEAKER_00 (01:37):
Okay, great.
Well, I would love to hear moreabout a recent study that y'all
did, I believe, with UCLA, Ithink.
Can you tell me a little bitmore about that study and what
it involved?

SPEAKER_01 (01:50):
Yeah, so um we have done a benchmark study in
collaboration with uh Departmentof Computer Science and
Engineering at UCLA.
Um, we have tested basicallyH200 GPUs, which is kind of one
of the latest NVIDIA GPUs.
It's like one generation beforethe Blackwell GPUs.

(02:12):
And the idea there was tobasically benchmark Ferret's
adaptive face cooling solutionin comparison to latest and
greatest directorship coolingsolution that that industry
currently uses.
And and then the and the resultswere um really interesting
because uh we find out that withFerret's adaptive face cooling

(02:36):
solution, you would be able toget 15% more teraflop per watt
out of each server.
That means you would be able toget 15% more compute for each
watt of power that goes to yourserver, which is quite
significant, right?
Because right now people arefighting to get more and more
compute, and accessing to morepower is not that easy.

(02:59):
So you need to basically squeezeas much power as you have to get
compute out of it.
And we have seen that there was15% improvement at the server
level, and on top of that, wehave seen significant
improvement on the PUE at thefacility level, which, if you
combine that improvement withthese 15% in total, in whole

(03:20):
kind of data center and AIfactory, we would be able to
deliver 35% more compute out ofthe same power envelope.
Which, if I give you an exampleof, you know, in in 100 megawatt
data center, that can translateinto almost 230 million of more
revenue per year, which is quitesignificant.

SPEAKER_00 (03:43):
Oh wow.
And how do you plan on applyingthis knowledge that you gained
moving forward?

SPEAKER_01 (03:52):
Yeah, so so we have different partners that that
that we are working with on theindustry, both from the OEM and
O and ODM side.
And we have some potentialcustomers that we are doing
pilot right now, and then we areplanning to scale our solution
with them.
The idea again is is is verysimple.
Um, you know, all these datacenter operators um have access

(04:16):
to uh constant amount of power,right?
Constant power envelope.
And the idea is basically to getas much compute as you can uh
from that constant powerenvelope.
And and we help them to do that.
Like you understand, 35% is isquite a bit of more compute that
you can get from the same powercontract that you have.

(04:36):
So we are working with differentpartners and with different
customers and try to bring yoursolution to the market.

SPEAKER_00 (04:42):
Yeah, absolutely.
Well, we keep hearing about moreand more data centers.
The need for data centers isgrowing exponentially.
Uh, we actually have some cominginto my city and uh neighboring
cities.
And of course, there areconcerns that come with data
centers moving into moving intocities.

(05:04):
Um, a lot of people are worriedabout the water usage.
Uh, from your experience, um,can you just share what
knowledge you have around that?
Can you compare that water usageto anything else that might be
in a city?

SPEAKER_01 (05:18):
Yeah, I mean, I I think a good example there would
be to compare the waterconsumption of a typical data
center to like kind ofvolume-wise to a swimming pool
that you guys have in the city,right?
Olympic-sized swimming pool.
And just to give you an idea, atypical 100 megawatt data center

(05:39):
can consume up to 4,500 Olympicswimming pool uh of water per
year for cooling, which is quitesignificant, right?
The example that I always giveto kind of people at my team is
that we already went through theindustrial revolution, you know,
and and and we did that kind ofunresponsible way, because we

(06:03):
went through the industrialrevolution, all the GDPs of
different countries went up, youknow, the life became more
prosperous, but no one wasthinking about the environmental
consequences that that we weredealing with.
And and now till today, we arestill dealing with those
consequences, right?
All the global warming issuesthat we have is still kind of

(06:24):
the tale of that industrialrevolution that we went through.
And I truly believe that now weare going through AI revolution
or or digital revolution, if ifyou wish.
And we need to do it in asustainable way, because if we
don't do it, we similar to theindustrial revolution, we would
need need to deal with all theseconsequences for the years to

(06:47):
come.
So we truly believe that, andthat's the mission of Ferret, is
to enable AI revolution in asustainable way.
Uh, and as a matter of fact, oursolution basically works with
the closed loop.
So it the water consumption isbasically almost zero with with
our solution.
So we kind of do the AIrevolution in a very, very

(07:11):
sustainable way.

SPEAKER_00 (07:13):
Yeah, that was my next question, as I know that
was for a regular data centerwith the original cooling
methods.
So now uh we are hearing aboutclosed loop systems.
We're hearing more about that.
So tell me more about the closedloop system and your solution
and how it works better.

SPEAKER_01 (07:30):
Yeah, absolutely.
So a typical um data center,right?
Uh use evaporative cooling uhboth for air cooling and also
for the director chip.
The evaporative cooling is asthe name implies, they basically
evaporate huge amounts of waterto take advantage of evaporative
cooling and and cool down theair that they are using in their

(07:54):
data center.
The way that we do it, I mean,our system doesn't have any
airflow, right?
The whole server is kind ofcontained in a in a in a sealed
chassis, and and our systemworks with the closed loop,
which is just the dry cooleroutside.
Our system can work at um liquidinlet temperature of up to 45
degrees C, which kind of make itsuitable for most environments

(08:18):
without using any evaporativecooling.
So the water consumption withour solution is is basically
zero.

SPEAKER_00 (08:24):
So that so once you've filled up with the amount
of water needed, you that staysin there and doesn't evaporate.

SPEAKER_01 (08:31):
Correct.

SPEAKER_00 (08:32):
Correct.
Okay.
Okay.
Does it ever have to be flushedout for any reasons?
Um, or is it pretty much umthat's the water always used
from then on out?

SPEAKER_01 (08:42):
No, that's the water always used, unless you have
some sort of failure or or leakin the system.
But the typically the water thatyou fill in your closed loop,
that's the one that stays in thesystem.

SPEAKER_00 (08:54):
So that's a huge deal then compared to you said,
all, you know, the Olympic-sizedpools example.
That's a huge savings in water.
Um, big picture.
How important is this to theenvironment?
With the need of all these datacenters, how big of an
environmental impact do youthink that this would make?
And are you working with othercompanies that are also sort of

(09:15):
on the same page of reducing thewater in the data centers?

SPEAKER_01 (09:19):
Yeah, so so majority of the water consumption comes
from the cooling, right?
So data centers typicallyconsume water in two ways.
One is direct, that's basicallydirectly on the cooling tower
that they have at the side oftheir data center.
Um, and the other way that theyconsume water is basically
indirect, and that's basicallythe cooling tower and the power

(09:40):
plant that generates power forthe data center.
Which in most cases we help themsave that water, right?
So on the cooling side, wecompletely eliminate the cooling
tower on the data center side.
And uh we also help them saveenergy.
So, in a way, we also help themto save water from the power
plant size as well.

(10:01):
And the concern is here,especially for the areas that
you are dealing with the dryout, you are dealing with the
regulatory environments, right?
For example, Arizona, Nevada,right?
The areas that's like extremelyhot, you would need even more
water for cooling.
And and our solution is kind ofrevolutionary because it it
enables you to build datacenters in those regions without

(10:22):
any water consumption, which isquite significant.
And if you think of it in a in akind of like a big picture,
right, it would also enable youin a way to combine your data
center with solar and deploy itoff-the-grid.
Because right now, whenever youhave solar energy, it's too hot,
and the data center is notefficient, and the ratio doesn't

(10:44):
make it financially viable tohave completely off-the-grid
data center.
But with our solution, uh, yourdata center is efficient, it
doesn't need any water, so youcan easily deploy kind of an
off-the-grid data center in theareas that it was not possible
in the past, which is like ahuge advantage and consequence

(11:05):
of the solution like FevRet thatenable you to build data centers
in those regions, which was notpossible before.

SPEAKER_00 (11:11):
So, are how quickly do you think that this new
technology can be deployed andused throughout all the upcoming
data centers?

SPEAKER_01 (11:22):
Yeah, well, we have our solution.
We are piloting our solution.
Um, if anyone is listening andthey are interested, we would be
more than happy to pilot thesolution with them and with some
of our customers we are planningto scale.
Uh, one thing is very clear theproblem is becoming bigger and
bigger as we move things forwardbecause you know people have a

(11:43):
problem of access to power, aproblem of getting permits for
water consumption, right?
And and the example that Ialways give is that you cannot
have a leaky bucket and pullmore water to it.
So you cannot just keep gettingmore and more permits for power
if you don't use your power inan efficient way.
So the problem is just gettingbigger and bigger, and we are

(12:04):
here to help.
So we're here to help anycustomer or any potential
partners that's out there andthey are building data centers
today.
We are ready to work with themto bring this solution to the
market.

SPEAKER_00 (12:16):
Are you hitting any walls when you go to talk to
data center builders orproviders?
Are you hitting any walls whenit comes to changing their
current cooling methods?

SPEAKER_01 (12:26):
Yeah, well, I mean, I think most of the data
centers, right, currently usethis kind of a liquid cooling
solution they call direct tochip.
And that's what the majority ofthem use.
But even that one is stilldealing with the water
consumption uh at the datacenter level and doesn't provide
the performances that oursolution does provide.

(12:47):
So again, as as the problembecomes bigger and bigger, and
and these data centers arefeeling the pain more and more,
I think then there's a chancefor the companies like ours to
you know to grow and come to themarket and offer the solution
that they don't have.

SPEAKER_00 (13:03):
Absolutely.
Well, if there was one keytakeaway you could leave our
audience with today, what wouldthat be?

SPEAKER_01 (13:09):
Yeah, well, we are here to help to make the AI
revolution sustainable and uh weenable you to be data centers in
the areas that you were not, youknow, able to build data centers
in the past.
Just come and talk to us, and weare here to help.

SPEAKER_00 (13:26):
Wonderful.
Thank you so much.
And um, where where can theyfind more information about this
study if they wanted to look atthe full study?

SPEAKER_01 (13:35):
Absolutely.
So we have the press release andwe have the white paper in our
website, which is onwww.febread.com.

SPEAKER_00 (13:44):
All right.
Thank you so much.
And thank you to our audience.
If you have any questions aboutthis, leave them in the comments
and we'll try to get themanswered for you.
Have a great day.