March 22, 2025 • 39 mins
Karl Freund of Cambrian-AI Research, Jim McGregor and Francis Sideco of TIRIAS Research joined me on a neXt Curve reThink Podcast to recap NVIDIA GTC 2025 from the Signia Hilton Hotel in San Jose on the third day of the event.
This episode covers:
➡️ Key impressions from NVIDIA GTC 2025
➡️ NVIDIA's evolving identity
➡️ NVIDIA's moves widening the lead in AI supercomputing
➡️ NVIDIA's broadening AI computing portfolio and the software stack
➡️ The NVIDIA philosophy of "more is more"
➡️ Synthetic data, digital twins, and robotic training
➡️ Jensen's jacket and more!
Hit Leonard, Jim, Francis and Karl up on LinkedIn and take part in their industry and tech insights.
Check out Karl and his research at Cambrian AI Research LLC at www.cambrian-ai.com.
Check out Jim and Francis and their research at Tirias Research at www.tiriasresearch.com.
Please subscribe to our podcast which will be featured on the neXt Curve YouTube Channel. Check out the audio version on Buzzsprout or find us on your favorite Podcast platform.
Make sure to follow me here on LinkedIn and hit that 🔔 at the top of my profile for a constant diet of the tech and industry insights that matter.
⭐ Subscribe to the neXt Curve reThink Podcast on Buzzsprout here: https://bit.ly/43mr2Hm
⭐ Subscribe to the reThink YouTube channel here: www.youtube.com/@nextcurve
⭐ Follow neXt Curve at www.next-curve.com
⭐ Subscribe to the neXt Curve newsletter here: https://bit.ly/3LbXVgZ
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Leonard Lee (00:10):
Hey everybody, this is Leonard Lee, executive
Analyst at Ncur, and welcome tothis Rethink Podcast and it's a
really special edition thatwe're filming here live at G CT
GTC 2025.
It's been a long day.
It's been a long day.
It's been a long way, and thisis our Silicon Futures program.
(00:31):
And, of course you all knowwe're doing this in
collaboration with TEUs Researchand Cambrian hyphen AI research.
You don't have to say hyphen, Ilove saying hyphen ai.
Well, you know, it, it meansthat I care.
Oh, right.
It means you can spell Yeah,it's one extra, it's one extra
token.
Man.
It's like the brown m and ms.
(00:52):
Okay.
You just do not have brown m andms.
Okay.
In a hotel when Van Halen is intown.
So that, that, that is thehyphen.
But, I'm joined by, JimMcGregor, as well as Francis Ko,
as well as the illustrious CarlFre.
(01:13):
Right.
And I'm sure everyone out therein AI land is really glad to see
us here together in San Jose.
At the, San or we're at the,actually, the Hilton Insignia.
Hilton Insignia.
Signia.
Okay.
Well, we've been getting bouncedaround, back and forth, all over
the place.
So gentlemen, this is like thebig Nvidia, good event, AI
(01:34):
events, super Bowl of ai, right?
Yes.
In all of the universe.
And so, looking forward tohaving this chat and, it should
be a good one, right?
Mm-hmm.
So, hey, let's talk about.
Impressions.
What'd you guys think?
What did you, number one, what'dyou guys think about this year
and what did you think wasdifferent from last year?
Karl Freund (01:53):
I think everybody was expecting really fast
hardware and a roadmap ofreally, really fast hardware
uhhuh.
so that was not a surprise.
They met the bar.
I think what surprised us, Ithink, at least, certainly me,
was two things.
One was the optical.
Yes.
The co silicon
Jim McGregor (02:12):
Photonics.
Karl Freund (02:13):
Thank you.
Silicon photonics.
Yeah.
I can't even say it.
And the other is, is reallydynamo, which is the new
operating system as they callit, which effectively think of
it as, as, the hypervisor forhundreds of thousands of GPUs.
Yeah.
That optimizes across all theGPUs, depending on.
(02:35):
What you're trying to do.
I'm still trying to wrap my mindaround it or, yeah.
Jim McGregor (02:39):
Or let's just call it the operating system for the
AI factory.
Karl Freund (02:42):
Yeah, yeah.
The operating system for the AIfactory.
But it's kind of like ahypervisor, right?
Yeah.
Leonard Lee (02:46):
Kubernetes, it's been described as sort of
Kubernetes for AI inference, Ithink
Karl Freund (02:51):
across.
Hundreds of thousands of gps.
I think
Francis Sideco (02:54):
that was, obviously the networking bit,
which I think probably Jim'sgonna talk about.
But I think for me it, it, itwas that, Jensen definitely
tried to make a case.
I mean, he's been under pressurein terms of the industry going
from, obviously we're stillgonna do training workloads, but
we're gonna be doing inferenceworkloads a lot more and the
importance of that.
And I think just generallythey've gotten hit in terms of
(03:14):
them not being.
Potentially as optimized forinference or not.
Mm-hmm And I think Jensen duringthis show definitely made a case
for the fact that actually whenyou start looking at the
transition of inference, notjust from one shot inference to
reasoning inference, youactually do need scale up.
In inference, which before that,came out, the question had
(03:37):
always been, do you actuallyneed scale up for, for
inference?
and he made the case that youdo, whether or not that case is
gonna get bought.
I think the jury's still out,but it passes the smelt as
especially Yeah.
When you bring the dynamo intoplay.
Jim McGregor (03:51):
Yeah.
Francis Sideco (03:51):
And I think that was the key for me.
Uh, and
Jim McGregor (03:53):
Agentic
Francis Sideco (03:55):
ai and agentic ai, that was the key for me is
that the homogeneous.
Fungible resources that Dynamocan help manage, so, right.
Jim McGregor (04:03):
Okay.
I don't know.
For me, it's like drinking froma fire hose and the fire hose
gets bigger every year.
But the stuff from Cosmos to dosimulated, or synthetic data.
Mm-hmm.
And, and.
Scenarios, just thousands,millions of scenarios using
synthetic data, right?
To train robots that is gonna,and autonomous machines.
I think that's huge.
I think seeing some of the newplatforms, they have a complete
(04:26):
enterprise platform solution.
Now, from, PCIE cards to, whatused to be Project Digits, now
it's, DGX Sparks, sparks and theDGX station.
So you have desks.
Desktop and side solutions forAI developers.
You have the DGX Super pod, youhave, the next generation of
(04:48):
Blackwell Ultra coming out.
So it's just one thing afteranother, and obviously, I.
The, the networking, networkingstole the show, I think for the
second year in a row.
Yeah.
Last year it was the, envy linkswitch, which, allowed 72 GPUs
to act as a single one.
Right.
In a single rack.
But this year, once again,everyone's been saying, and
(05:08):
this, we've been developingsilicon photonics since about
2000 actually earlier than that,but.
Everyone keeps saying it's twoto three years away.
It's two to three years away,right?
They finally got an ecosystemtogether to actually produce
something that they're, gonna beoffering a quantum X infin bend
version of it and a specxethernet version of it.
(05:29):
So they're gonna have, Uh,silicon or they're gonna have,
photonic network switches.
That is significant in terms ofreducing components, power
consumption, increasingresiliency and reliability.
Reducing
Karl Freund (05:43):
it all ships this year.
Jim McGregor (05:45):
Oh yeah.
Well, the switch next year.
The infin, the Quantum X shipsthis year.
The max spec Max next year.
Spec max next, early next year.
Yeah.
Francis Sideco (05:52):
And it may, it makes sense.
I mean, they did scale up lastyear with NVLink Switch and
then, scale out this year, and
Jim McGregor (05:58):
I think he points to what we're gonna see.
I would expect futureenhancements to NVLink, so that
we see more than 72 supportedmore than 144.
2, 280.
Well, he already, he alreadykind.
Yeah.
We've already.
Seen that with previewed
Francis Sideco (06:11):
that with where they're going with Vera Ruben.
So
Jim McGregor (06:13):
yeah, with Reuben and Reuben Ultra, I think that,
obvious, I would expect, eventhough he said No, use copper in
the rack, by the time we get toRuben, I almost think that we're
gonna be looking ats atphotonics in the rack.
Well, he's
Karl Freund (06:25):
double down on it this afternoon, right?
Yeah, yeah.
I mean, he, this morning hesaid, no, we're gonna stay.
Copper in the rack, so I'm notconvinced
Francis Sideco (06:33):
of that one.
We'll see.
Well, he said use copper as muchas you can.
And I think that's the key.
As, as long as can, as long asyou can and as long as you can,
is the key factor.
That might not be that long.
So
Leonard Lee (06:44):
what about you?
Well, you guys have left me withnothing but scratch once we go
down the line.
Say, but I do have to saysomething.
We're, this is all likeunscripted of course.
Obviously we don't have anythingto refer referencing, so I need
to do the disclaimer.
Welcome.
And by the way, all the opinionsof these guys are entirely their
(07:05):
own, don't reflect next curve ormine, which is probably good and
completely accurate because theylike to be up on me.
Right.
I'm like, like the black sheepof the family.
Yeah.
Yeah.
Especially Carl.
But, I thought it was really.
Overwhelming, a lot of, detailson the roadmap where I think
like what Jensen has been tryingto emphasize, I think, over the
(07:28):
course of GTC is that, they're adifferent company.
Mm-hmm.
I think that's a, and we've beentalking about how they're
becoming more systems oriented.
And, Jim often services.
Right, right.
And then the software aspect andso.
All this stuff is comingtogether in a really compelling
way and I think Jensen does areally great job of, telling the
(07:50):
story.
Now, I do have to say there aresome gaps in what I've heard,
and I think these are areaswhere Nvidia really needs to,
like all the scaling stuff thatdidn't work for me.
It still doesn't work for meeven from last year.
Right.
And how does Huang's Law fromlast year now map into these
new, this new scaling thesisand, I'd like to see more
(08:13):
clarity on that as we goforward, but they are also
introducing a lot of stuff andthey're trying to figure out how
to frame this.
Whole scaling argument, againstthe backdrop of a lot of what's
happening in ai, right.
With moving toward reasoning.
Last year was about MOE, right?
Mm-hmm.
Yeah.
In terms of the ai, make suremakes sense now.
(08:34):
Yeah.
We have the time scaling, thislong thinking that Jensen talks
about where the industry istalking about in general.
And, these are all new conceptsand they have implications on
the requirements around compute.
And we hear Jensen talking aboutthe a hundred x, right?
With reasoning, we need, ahundred times more or more
(08:55):
tokens, compute, which isdifferent from tokens.
That was one of the things thatI took away.
As we talk about tokens, we needto think about it not just in
general terms, but very specificterms in
Francis Sideco (09:05):
well terms
Leonard Lee (09:06):
of the, inputs and then what.
It gets generated in terms of atoken and how that translates
into how much more compute youneed.
Mm-hmm.
And so, there, I mean there'sall these layers that are being
placed on the conversation.
Yeah.
And that makes it really complexand I think, difficult for,
(09:26):
folks to.
Keep up with, I think.
Francis Sideco (09:29):
I think that's one thing that we do all need to
keep in mind, right?
Because obviously Jenssen'sgonna keep pushing, Jenssen's
gonna keep pushing theboundaries of what's needed.
But that doesn't mean that's theonly way.
Enterprise AI or AI in generalis going to create value for the
economy at large, right?
Yeah.
There's going, there's a ton ofbifurcations that are happening
(09:52):
in terms of use, in terms of,how you architect these data
centers for what differententerprises need and so forth.
So not everything is going to bethe top of the line, Uber scale
stuff that, that jenssen'stalking about.
There's going to be room forother architectures and and
other use cases that don't needthat.
(10:13):
That, screaming high endcapability.
Jim McGregor (10:15):
Well, and especially with the
infrastructure requirements thatcome with'em in terms of the
cooling and power and everythingelse.
Everything, yeah.
Yeah.
But one thing that really amazed me.
And, they made this change acouple years ago to combine, uh,
their automotive stuff and theirother robotics.
Robotics and other stuff, andinto autonomous machines.
And hearing from some of theother groups learning that even
(10:38):
though avs have kind of beenpushed out and fallen by the
wayside mm-hmm.
That a lot of these other groupsfocused on robotic machines for
healthcare and other types ofapplications, how they've taken
that learning that's alreadygone into automotive, even if
it's not being deployed, howthey've taken that learning and
applied it to other forms ofautonomous machines,
particularly robotics.
(10:59):
Yeah.
Yeah.
Leonard Lee (10:59):
Oh, really quickly.
Yeah.
I wanted to bring this upbecause I thought it, we've been
talking about digital twins fora long time.
Mm-hmm.
And quite honestly, a lot ofconversation, especially.
Like two, I think about twoyears ago it was nonsensical.
This year actually, the stuffwith Cosmos, right?
Mm-hmm.
Yeah.
And synthetic data.
What I think we all need torecognize is we're not talking
about synthetic data in terms ofa SOA generation.
(11:22):
This is like using some of thephysical models that they used
to have.
Mm-hmm.
And all the things that theyhave in gaming, it's like has
gaming roots that are used tomodel like certain environments
and then they augment that withAI to, synthetically generate.
Scenarios New data.
Data, yeah.
Right.
Scenarios.
(11:42):
And I think, that's where, somefolks might have been confused
in the past and what is all thesynthetic mm-hmm.
Data stuff.
But, one of the things that Iwas really impressed with this
year is, how it will essentialin supporting a lot of the.
Test acceleration, right?
Mm-hmm.
That are gonna provide thelearning environments and
learning experience for thesemodels that are gonna be
(12:04):
trained.
And this has roots, like whatyou're saying in automotive and
just there's this, all thisheritage that's coming together
mm-hmm.
In some of these concepts that,I think are, actually pretty
powerful.
Karl Freund (12:16):
So I think one of the things that, I'll never
forget the quote from thekeynote, which is Jensen calling
himself the Chief RevenueDestruction Officer.
I.
That's, that kind of rings truebecause when they announced they
were going to an annual productcamps, a lot of us who've been
in the computer industry a longtime said, that's gonna be
really hard to do.
Customers don't like having whatthey just bought be obsoleted so
(12:38):
quickly.
But he really embraced it.
He really embraced it.
And he said, he said, yes.
Broad, Blackwell, is shippedthree times more than, already
this year mm-hmm.
Than Hopper did all of lastyear.
And then he said that, revenueramp is gonna continue to drive
forward.
Jim McGregor (12:58):
And with Dynamo, they're getting a 40 x.
40 times improvement overHopper.
Karl Freund (13:05):
Yeah.
Jim McGregor (13:05):
That's just phenomenal.
Karl Freund (13:07):
It's wild.
Yeah.
You, that's an inferencestatement.
Yeah.
Right.
This hard inference is really atthe heart of everything they're
doing this year at GTC.
Yeah.
Whether it's for robotics or forthe data center, or even for the
little, miniature.
DGX.
The, yeah,
Jim McGregor (13:23):
The Spark.
Karl Freund (13:23):
Yeah.
Jim McGregor (13:24):
Love that.
Karl Freund (13:24):
Wanted to call it digit.
I want one.
It's called Spark.
I want one.
I one so bad.
It's super cool.
We all want one.
Leonard Lee (13:31):
Yeah.
So Jensen send us, we all wantedDX Spark.
Yeah.
And a workstation and station,right?
That would be great.
Like, and then sign up please.
Okay.
I'm not going that far.
Oh no.
His jacket by the way, everyone.
Yes, he was wearing his originaljacket once again.
So that jacket that he wore atCES must have been a CES
(13:54):
special.
Yeah.
So, yeah.
Well, he wore two differentjackets
Jim McGregor (13:57):
today.
Yesterday.
Did he?
Yeah.
Yeah.
Oh,
Leonard Lee (14:01):
he's got, he got a lot diversity in his jacket
collection.
He does.
I wonder if
Karl Freund (14:05):
he wore that jacket or a suit when he went to the
White House.
I don't know.
I didn't see any videos of hisarrival.
Maybe it's a simulation.
Francis Sideco (14:12):
Yeah.
But let's not underestimate whatyou said, Leonard, about the
simulation and like with cosmosof being able to create
synthetic mm-hmm.
Training information.
If you really take it intolayman's terms, that's like
being able to, force feedtraining Yeah.
To a human being.
Like the way they did it inMatrix, frankly, it's just like
(14:34):
shoving it all in, in, it's likedriving a billion miles.
It's like driving a billionmiles in a very short amount of
time.
And there's the sheer amount ofexperience that you're able to
gather one of and
Jim McGregor (14:44):
apply one of, one of the stats was 27 years of
simul or of what would be realtime, physical training.
In, what was it?
Francis Sideco (14:54):
In one day?
One day on one GPU.
and that's assuming in realworld you actually encounter
those situations.
Yeah.
Dropping Godzilla in the middleof the freeway happen.
Yeah.
You might not do that.
But the fact that you're able todo that kind of concentrated
learning, you can, IM, you canimmediately see how some of
(15:16):
these automated systems.
Are becoming more highly trainedthan a human because you can't
do that.
Well, it's
Karl Freund (15:22):
given me renewed confidence that we will solve
the.
Automated driving.
Well, you problem.
Leonard Lee (15:27):
And and it's not new.
'cause they were, they weretalking about this stuff like
four years ago.
Yeah.
You know, it just is at the sortof like the height of like the
av, av yeah.
Mm-hmm.
Pipe.
Mm-hmm.
But its
Francis Sideco (15:38):
application wasn't as.
As clear Well, how you can seehow well divers diversified
Leonard Lee (15:44):
its application.
I think, I think it was stillclear and necessary back then,
right?
Yeah.
In accelerating the, thetraining of, AV systems.
Now, I, I mean they're broadenit, especially with their focus
on, um, robotics, right?
Yeah.
With Isaac group, right, right.
Yeah.
Being
Jim McGregor (16:01):
able to take, human actually doing something,
combining it with videos ofhumans doing something.
Yeah.
And then using AI to createsimulations and solutions.
Yeah.
And just phenomenal.
I know that even a lot ofsurgery being done today, like,
knee replacements and shoulderreplacements and stuff like
that, they're actually usingrobotic arms to do all the
cutting.
(16:22):
Yeah.
And everything else, it'samazing that just blows me away.
Karl Freund (16:24):
One of the things that, that I thought was really
interesting was the discussionabout the automotive industry
and the, and NVIDIA's approachto the automotive industry,
which I always thought it wasprimarily in vehicle.
And now I'm starting to realizethat's nice.
If it, if they have an Nvidiachip in the vehicle,
Francis Sideco (16:40):
that's a what they're really after is the
Karl Freund (16:42):
cloud for simulation, for the vehicle
operation, for the training.
Yeah.
And the command and control.
And an important point was thatonly 5% of the vehicles in the
world today are level two andabove.
Jim McGregor (16:53):
Yeah.
Karl Freund (16:54):
So there's a huge opportunity there.
Yeah.
And the opportunity is not justin the vehicle, it's all a
simulation.
All those vehicles, all thoseOEMs will require before they
put those cars in the road.
And,
Jim McGregor (17:05):
And while they did back out of the infotainment
solutions.
Mm-hmm.
Now they've got the wholepartnership with media tech.
Using their technology, theirIEP to develop an I develop by.
I'm surprised.
Karl Freund (17:16):
Hear an update about that.
Jim McGregor (17:17):
Yeah, I was kinda surprised too, but I would
expect that we will.
Funny.
Francis Sideco (17:22):
Well, on the automo, on the automotive
vehicle bit, right?
I mean put it in in perspectiveright now, like in Nevada, you
train your 16-year-old on 50hours of drive time and they
give him a license.
Jim McGregor (17:34):
Oh my god.
Francis Sideco (17:34):
Okay.
And like 30 hours of classroomwork, 50 hours of drive time,
they gave him a license.
Mm-hmm.
Okay.
Compare that with training a veuh, an automated vehicle with 27
years worth.
Of basically drive time.
Yeah and it just, it puts it inperspective.
Well, as long as it's good
Jim McGregor (17:51):
driving.
I was just gonna say, it couldbe really bad driving actually,
but hey, no.
Well, that's the thing about
Francis Sideco (17:56):
The synthetic data is because you can control
the, you can control thesynthetic data.
It's not garbage and garbage outbecause you can control the
quality of the synthetic.
Well, but,
Leonard Lee (18:05):
but I'll argue this, you need to have a crap
driving algorithm to create.
Stupid situations in your Eurosynthetic data.
'cause that's actually whatYeah.
A vehicle needs to respond to isto bad drugs because, there's
only finite numbers of way todrive properly at any given
time.
There's infinite ways ofdriving.
Really crappy crap.
(18:25):
Yeah, yeah, yeah.
Exactly.
And they can get 90%
Francis Sideco (18:27):
of that in Vegas.
'cause the drivers thereshocked.
I was trying to get this guy to
Jim McGregor (18:31):
it, it, it strangle me.
It, it, it really needs toaccount for me on the road.
Yeah.
Oh yeah.
Terrible.
The car that's going three timesfaster, my god.
Crazy.
Geez, that's really horrible.
I actually used to terrorizethe, Waymo's in Phoenix just to
see how they would react.
Oh.
So there was one Jim sittingthere poking the beer.
(18:54):
I would, I'd actually, there'sone place on Mill Avenue where
it goes down to one lane and Iwould actually pull it next to
the Waymo's and test them andsee how they react.
Hopefully
Francis Sideco (19:02):
without a passenger in there.
But no,
Leonard Lee (19:05):
you, I wanted to take things back to the new AI
supercomputing lineup.
Right.
And last year they introduced,what the transformer,
transformer engine, right.
Is that what that was?
No, that was, no, no, no.
That was.
Karl Freund (19:18):
Now transformer engine was,
Leonard Lee (19:19):
two years ago.
Yeah, two years ago.
Right, right.
But how there's a lot more talkabout, so last year they
introduced transformer engine totwo, right?
Yeah.
For, yeah.
And so with the support for, FPfour, right, exactly.
But how now that's becoming evenmore important.
There's a lot more talk about,uh, mixed precision.
(19:42):
And the role that, it plays indriving the, the scaling of
these systems.
Right.
And so when we look at, one ofthe things I noticed, if you
look at the roadmap, the chip,it's, the individual GPUs
themselves are not scaling asfast as everything else around
it.
We talked about networkingalready, but it is interesting
(20:03):
to see how, you have the memory,the HBM, right to, the upgrades
there that are happening overthe course of the next, two
iterations of the system as wellas the networking.
Really playing a outsized role,in, in scaling the systems.
Yeah.
And I, one of the things Ithought was really interesting
was Jensen, trying to clarifywhat a GPU is.
(20:26):
Mm-hmm.
Mm-hmm.
Right?
Because everyone thinks it'sjust the chip, but in actually
their parlance it, it issomething much bigger.
Right.
And, we see this, I guess it's aconfusion.
E, especially amongst the mediaabout what a GP is.
Just turn on a TV and have, Idon't even to call'em GPUs.
Yeah, it's not really a GPU,right?
(20:47):
I just called an
Jim McGregor (20:48):
accelerator, or I call it a GPU accelerator.
'cause it's really anaccelerator.
Yeah, but it's a accelerator.
It's it literally is a freaking,you look
Francis Sideco (20:56):
at the GPUs that are in like the G-G-D-G-X sparks
and all that stuff, they don'tdo graphics.
Let's ask this question.
Okay, let's go down
Leonard Lee (21:04):
the line here.
What the hell is a GPU?
Now is
Karl Freund (21:08):
Nvidia Accelerator.
Well.
Jim McGregor (21:10):
It's a graphics the best you can do.
It's a graphics processing unit,but these things aren't GPUs.
Francis Sideco (21:16):
Yeah,
Jim McGregor (21:16):
that's the problem.
Francis Sideco (21:17):
Yeah.
It's K FFC is no longer justKentucky Fried Chicken.
Right.
It's just KFC and it's the samething.
It's just another, it's justanother, it's another, where
it's not an acronym are bringing
Leonard Lee (21:26):
out Kentucky Fried Chicken as.
Source of an analogy.
It's been a long day.
He's been king a long day.
Oh my gosh.
The king food analogy, bringingup Pizza Hut.
Talking about splitting up pizzapies.
Yeah.
Well, no, I mean, I think theimportant thing to understand is
(21:46):
like what you alluded to or youpointed out before, is that it's
a logical it.
It's a logical gpu.
Massive, right?
Mm-hmm.
And I don't know if that's quiteregistered in everyone's mind,
but it's an important point for,everyone who's observing the AI
industry and AI supercomputingto clearly understand.
Mm-hmm.
Right.
(22:06):
I think
Karl Freund (22:07):
one of the, interesting conversations with
Jensen was about what do youwant NVIDIA to be known as?
What do you want the Nvidia tobe known for in three to five
years?
Mm-hmm.
And the answer was not.
The data center, the, which iswhat many of us were expecting.
What he said was that Nvidia isfoundational mm-hmm.
(22:30):
To the entire world forartificial intelligence.
Yeah.
And I thought back to theslides, he flashed up really
fast, too fast for any of us toread, during his keynote, which
was of all of his major partners
Jim McGregor (22:45):
Yeah.
Karl Freund (22:45):
And how they're adopting.
Ai.
And if you look at each one ofthose, they had these little
green squares on them.
Those green squares were allnims.
Okay?
Nvidia Inference, inferenceservices.
And that to me is foundational.
That is more of a competitivemoat than Cuda.
Jim McGregor (23:08):
Well, it's an entire stack.
It's an
Karl Freund (23:09):
entire stack, and you have the entire industry
adopting that entire stack,which is gonna make it really
hard for a MD or Intel.
Well, and
Jim McGregor (23:19):
I don't think it's just, uh, I don't think it's
just that, when you look atDynamo mm-hmm.
Being an operating system for anAI factory or even a data
center, when you look atOmniverse, which is, you know,
depending on how you look at it,you could actually classify it
as an operating system for thecloud.
Yeah.
So there are, not to mentionthese endless, almost endless
(23:40):
foundation models are creatingfor each individual vertical
segment.
Francis Sideco (23:44):
Yeah.
So just going back to my earlierstatement though, I agree
completely that Nvidia is.
Is continuing to push theenvelope.
They're gonna get a lot of valueout of this and a lot of return
on their investment.
But that does not mean thatthere's no, no place and no role
for other competitors becausenot everybody needs like the
Uber scale.
Mm-hmm.
(24:05):
Training or inference tofrankly,
Jim McGregor (24:08):
there, there's still a lot of traditional AI
solutions out there, and AI atthe edge
Francis Sideco (24:13):
there, there's a lot of a AI at the edge.
And even when you look at thecompetitor, the more direct
competitors to, to Nvidia, I'msure they're not.
Standing still, right?
Yeah.
and they're going, there is anappetite, for not just one, one.
We don't want to end up withanother kind of Wintel
situation, from a, from.
20, 30 years ago.
(24:33):
And I think the industry thereis an appetite for other
options.
And especially at the edge.
Especially at the edge, there'sopportunity
Jim McGregor (24:39):
that, and for specific applications cases,
that's why specific we still seehyperscalers doing their own
silicon.
Yeah.
Because they have use caseswhere they know they can
optimize for.
Leonard Lee (24:49):
Yes.
And those are, so one of thethings that I think it still
has, it's still not beenclarified.
Or is well understood is thatthere is a difference between
hyperscale and cloud serviceprovider.
Oh yeah.
Right?
Oh yeah.
And that, for the hyperscalerslike meta who have singular
workloads that are massive, itmight be a recomme scale to
(25:12):
optimize.
Yeah.
I mean that, that's gonna be aproduction application that req
requires that level of massivescale where yeah, you're going
to be going custom.
You're going to, tune the entireinfrastructure singularly to
run, that set of workloadsoptimally, right?
(25:33):
It's an operational system and,for the cloud and for the cloud
service providers and the modelbuilders, that's where I think.
Nvidia has its sweet spotbecause it's programmable, it's
more, quote unquote generalpurpose, right?
It has the flexibility to thesoftware where the, the hardware
(25:54):
now can be used in a widedifferent range of, purposes,
right?
And projects.
And that's a different.
Value proposition almost.
But I, I do feel that Nvidia istrying to, break that, notion,
right?
That no, actually we're good foreverything, right?
(26:14):
And that mm-hmm.
It's an argument obviously,that, they'll want to make and,
and but it's important tounderstand that dynamic.
Karl Freund (26:22):
Yeah.
Jim McGregor (26:25):
Yeah.
And we should also note that.
This is just mayhem.
They have outgrown San Jose,period.
they shut down most the roads.
Yeah.
they have to do it somewheredowntown San.
This is ridiculous.
Yeah.
Took over
Karl Freund (26:38):
the entire park.
Yeah.
For, for food.
Well,
Francis Sideco (26:40):
hopefully they'll be out in Vegas at some
point, but
Leonard Lee (26:43):
Yeah.
Yeah.
So, let's talk about enterprisegeneral ai Understood Uhhuh.
Did you?
Well, I talked about thesystems.
No, you saw, talked about thesystem, but enterprise, because
one of the things that they'rereally trying to push for this
year, because you're seeingthings like, I don't know if you
noticed, there's like a whole AIdata, data, and storage movement
(27:06):
going on, right?
People want to get a handle onthe.
Enterprise data and figure outthat out.
Right.
Which is different from whatwe're seeing with what the model
Yeah.
Builders are producing orpursuing, which is, hey, how do
we get as much, publiclyavailable or maybe in
copywritten data as possible forus to train these foundation
models.
(27:26):
But for the enterprise, it'sdifferent, right?
Yes.
And you're now, this is reallyhow do we make data available
for compound AI applicationsthat are gonna enable us to
take, advantage of this new AIcap generative AI capability.
I wanna make that distinction.
'cause everyone, talksgenerically about ai, the thing
(27:48):
that's really caused the stir inthe last two years is generative
ai.
Right.
The big question.
And now we just
Jim McGregor (27:53):
AI And those are two different things.
Leonard Lee (27:55):
Yeah, they are.
I think they're closely boundtogether in, in a certain way.
Well, in, in terms of function
Jim McGregor (28:05):
or, or, yeah.
Well, generative AI is thought,agent AI is reasonable.
So, um, yeah, you know what?
I'm glad you brought that up,okay.
Keep going.
One is going to give you afeedback on everything it's been
trained on.
The other one is actually gonnasit there, think about it, and
use a lot of differentinformation that's available to
(28:26):
it to actually make a judgmentcall.
And it may be significantlydifferent than Gen ai.
Francis Sideco (28:31):
And you add onto that, it's abil, it's agency,
it's ability to take thatjudgment call and then act upon
it.
Take action, take action.
I think that's'cause a lot ofpeople are using agent ai.
To mean basically copilotassistance or that's completely
different.
Yeah.
Leonard Lee (28:46):
And you know what, so the reason why I'm glad you
brought that up is there is asemantic disconnect out there
right now.
Yes.
The way that people talk aboutAG agentic ai, it's as certain
circles, well, there'sassistance, but then.
Now, like when you, what you seea lot of ISVs talking about is
AgTech AI in the context ofautomation, process automation.
(29:10):
Right?
Right.
And so there is really thisYeah.
Like disconnect that's happeningand that's what they call, they
call it semantic Dspecification.
That's actually a, actually,it's, it goes the other way
around.
I can, I
Francis Sideco (29:24):
can't even get my semantics in my head straight
on that, but I mean, that isreally important if you're
talking about agent ai and thefirst thing outta your thought
process is, a work, a processworkflow.
Mm-hmm.
That's not a agentic ai, that's a, that's an
automated process workflow.
That.
That's different than actualagency.
Don't say that.
(29:44):
Yeah.
you're gonna
Leonard Lee (29:45):
make some ISVs very unhappy by saying what you just,
Francis Sideco (29:49):
well, here's the, here's the thing.
We have to make sure that weunder, I mean you can, from a
marketing standpoint.
Okay, fine.
Use that.
But if you're really trying tounderstand the implications to
the technology and what you canor can't do about it and what
kind of infrastructure you needto support it, you have to be
clear about those nuances.
Leonard Lee (30:09):
Well, yeah.
And then we also have toreconcile that against what
we're hearing, from the, a lotof when we were in the A IPC
world and the on.
Device AI world.
Mm-hmm.
How they talk about ag agenticai which is personalization,
it's doing things for you.
Yes.
These are automation concepts.
(30:29):
Action.
Well, there's
Francis Sideco (30:30):
there's context.
Yeah.
There's the context element toit.
There's the judgment and thenthere's the action, which is
exactly agentic ai.
Leonard Lee (30:39):
Okay.
Karl Freund (30:39):
Yeah.
Without action is not agent
Jim McGregor (30:41):
And enterprise is, really trying to figure out how
do they, they use all theinformation in free and open
source models that are outthere, combine it with their
information and then whatresources do they need on prem
and the cloud.
Everyone's looking at hybridsolutions, but it is.
And it's a, it's, a colluschallenge right now for them to
figure out.
(31:02):
But I think over the next yearwe're gonna see a little bit
more clarity and ability to beable to do that.
Francis Sideco (31:09):
And I think where enterprises need to start,
and this is where they kind, getlost a little bit, is there's so
much possibility out there.
They're like, they don't evenknow where to start.
And I think really, if you'regoing to be serious about
enterprise ai, you really needto start and.
This is from every singleconversation I've had is start
with specific problems thatyou're trying to solve for your
(31:30):
business, and then let thatdrive what model you're using,
what architecture, hybridon-prem, whatever you're using,
what you know, data you need tobe bringing in and so forth,
governance, all of that stuff.
But you have to start with a,with specific use cases and
problems that you're solving.
Absolutely.
So.
Leonard Lee (31:50):
So I did want to give a shout out to IBM.
I know you guys like doing that,right?
So I'm gonna, I haven't done it.
You guys have, but I'm gonnagive a shout out to IBM because
they're the only company I'veencountered in two years that
are, it's, that's actuallyaddressing, enterprise AI
security for ai.
Yeah.
They have a thing called,context aware storage Folks.
(32:14):
Keep an eye out on that.
And also.
Here's a name that you need toknow.
His name is Vincent Su, HSU.
He's an IBM fellow, and he isworking on this stuff.
And, they were collaboratingwith, Nvidia to integrate what's
called, context to Wear Storage.
Into the NIMS framework, right?
(32:35):
Mm-hmm.
And this, I think this is reallyimportant work.
Yeah.
Jim McGregor (32:37):
And
Karl Freund (32:38):
I think we'll hear a lot more about next week in
New York.
Yeah.
Jim McGregor (32:40):
That and then combined with their granite
models, which are really theonly lab enterprise ready.
AI solutions.
Yeah.
Well it
Francis Sideco (32:50):
From a security, and guardrail standpoint, but
also couple that with instructlab, where there is, where it
makes it really easy to thenbring your enterprise data
context in, into the, to retrainfor, specialized
Jim McGregor (33:04):
model.
Yeah.
But
Francis Sideco (33:05):
again, even then.
Even in my conversations withIBM, they say it's very
important for you to know as anenterprise exactly what problems
you're trying to solve.
Yeah.
'cause then that's when thetools really sink.
Leonard Lee (33:16):
Yeah.
And you know why I'm bringingthis up is that, I've identified
this gap in rag architectures.
This is gonna resolve the, I'm,that might be premature to say
it will resolve, but it startsto address the issue of being
able to apply fine-grainedsecurity controls.
Yes.
On top of a rag.
Mm-hmm.
And so whether it's, graphdatabase.
(33:37):
That you're dealing with, orVector database.
A lot of folks don't know thatyou can apply, fine grain
controls on top of a vectordatabase.
They're, they were neverdesigned to do this kind of
stuff.
Yeah.
And so they're doing somecreative, abstractions of
controls on top of, and this issomething I need to find out
more about, but I'm veryexcited, at least IBM recognizes
(33:59):
this problem because for thiswhole time people have just been
focused on.
Hardware level security, networklevel security, but not really
looking at the application andthe data.
Francis Sideco (34:10):
And you can see the convergence there.
Aside from that, that,collaboration, they also
announced a collaboration withNVIDIA around their IBM
consulting, engagements, becauseit is right now requiring.
A lot of consultation forenterprises to figure out, okay,
what is their use case?
What are they trying to solve?
Leonard Lee (34:29):
Yeah.
Francis Sideco (34:30):
Um, and then go from there
Karl Freund (34:31):
and what's the state of the data
Francis Sideco (34:32):
Yeah.
And what's the state of thedata.
Absolutely.
That's a key one.
Leonard Lee (34:36):
Yeah.
So, yeah.
Cool.
So, yeah.
And so what do you guys expectnext year?
I, I guess you, you already knowwhat to expect.
Already know that's a
Jim McGregor (34:45):
bigger fire hose.
Leonard Lee (34:46):
Yeah.
Bigger, bigger fire hose.
Jim McGregor (34:48):
Come on, let, let's do this.
We gotta do this.
Well, we already know.
We already know.
Next year, do we?
Next year it's,
Francis Sideco (34:52):
do we, it's Ruben, right?
Yes.
Next year will be Reuben.
It's very Ruben.
Yeah.
Karl Freund (34:57):
Next year it'll be Ruben.
Francis Sideco (34:58):
And then on the networking side, they went scale
up last year.
They did scale out this year.
So we'll see what's next.
Next up for that,
Jim McGregor (35:05):
They seem to be strategically targeting.
Each bottleneck in the datacenter, especially in the data
center.
Something around ai.
So I, definitely I think we'llsee enhancements in memory and
everything else.
And networking.
Yeah.
I'm hoping that they startaddressing the bottleneck that's
emerging out of theinfrastructure for the data
center.
(35:25):
Because right now all thecooling and, power solutions are
really customized to eachimplementation.
Yeah.
Yeah.
And that's gotta change.
When you're spending hundreds ofmillions of dollars on a data
center, yeah.
You have to, you don't want todo it each time you're employ,
you're, putting new GPUs intothe data center.
(35:46):
So it's gonna be interesting.
Do you realize what you'resaying?
Yes.
Oh my gosh.
It has to be fucking play.
I, on that note, I think isprobably, I worked for the
Motorola computer.
Matter of fact, I'm surprisedthat they're not looking at
alternative structures likeblade servers.
Leonard Lee (36:01):
Ah, here we go.
I'm serious.
He's back on the soapbox.
Blades are good.
Blades are good.
All right.
On that note, on that note, no,there's nothing wrong with that.
No, I think it's a greatcomment.
Yeah.
So seriously, you guys, there'sno expectations.
You know everything for nextyear.
No, no,
Karl Freund (36:21):
no.
Leonard Lee (36:21):
Oh, absolutely.
Just
Karl Freund (36:22):
like we didn't expect optics.
Leonard Lee (36:23):
Okay, so he got one.
What about you?
Francis Sideco (36:26):
Yeah, I think I said me, I think, don't say
Leonard Lee (36:28):
that you, you're gonna just repeat.
No, no, no.
I think
Francis Sideco (36:30):
I said earlier when you first asked it was, I
think the memory subsystem okay.
Is going to become extremely, italready is extremely critical.
It's, it's going to become evenmore critical and I think, I
think we'll see probably someinnovations around that.
Karl Freund (36:46):
Yeah.
Carl.
You've stumped the band.
I, I, I really don't know whatto expect next year.
I thought I knew what I, what toexpect this year.
Mm-hmm.
I was wrong.
Yeah.
Right.
Was not expecting storage.
We got storage was not expectingoptical.
We got optical networking.
I don't know, may, maybe they'll do optical in the
(37:06):
rack.
Maybe that'll be the surprise.
I don't know.
But, it's gonna be a fun spaceto watch this space.
Leonard Lee (37:11):
Yeah.
Like for me, I mentioned beforesecurity.
Hopefully we'll see traction.
I don't know if we will.
It might be a pretty tough nutto crack.
And, this whole, KV cash thing.
Yeah.
With Dynamo.
Yes.
Well, that's, it's sointeresting.
That's the memory.
That's the memory that, becausehere's the thing, I wanna know
whether or not, because thisyear it's all about a hundred x.
(37:33):
How about the other way around?
How about compression and howabout reduction of the amount of
compute required because, it'sone, it, I understand why Nvidia
wants
Francis Sideco (37:44):
more.
Leonard Lee (37:44):
More.
Yeah.
But what about less?
Because someone's gonna takethat opportunity.
Right.
And that's, that's, and I think
Jim McGregor (37:52):
we're seeing that with their Jetson, even though
it wasn't really highlightedthis year.
Yeah.
Their Jetson platform forautonomous machines, whether
it's robots or cars or whatever,they're going there even though
that wasn't the focus of thisgtc.
Mm-hmm.
I still think that you're gonnasee more.
Leonard Lee (38:08):
And so that's what we might be able to expect.
Maybe Yes, at GTC 2020 6 26.
Yeah.
And with that, I guess we, wewrap it up.
We
Francis Sideco (38:17):
wrap it up, yeah.
Sounds good.
Well,
Leonard Lee (38:18):
gentlemen, that was fun.
It was great hanging out withall of you here at G.
It's been a fun week, GTC 2026and, to our audience, thanks for
listening in.
Remember to like, share, commenton this episode.
And remember to subscribe to thenext curve, rethink a podcast,
as well as our research portal,www.next curve.com.
(38:40):
And also, these guys arepublished
Jim McGregor (38:43):
everywhere.
Find us on Forbes.
Find us on E Times, eur,YouTube.
YouTube, a number of otherpublications and outlets and
definitely, look for us at, theother trade shows that are
coming
Leonar (38:55):
up@www.teusresearch.com.
Come on man.
You gotta front your companyproperly and of course, and
Cameron
Karl Freund (39:04):
hyphen hyphen hyphens a i.com.
Leonard Lee (39:06):
Yes.
And I do that because I don'twant people to like, to not have
that.
Yeah.
Put in the website.
Thank you very much.
Well, it's been fun.
Yeah.
And that's Carl Fre of, CambrianAI Research.
Karl Freund (39:18):
Well said.
Leonard Lee (39:18):
Yeah.
And so, until next time,remember to always tune into the
Rethink Podcast and our SiliconFutures, program here for the
tech and industry insights fromthe world of AI and like GPU
stuff.
We'll see you next time.
Bye-bye.
Thanks.
(39:38):
Bye-bye.
Hey everybody, this is Leonard Lee, executive
Analyst at Ncur, and welcome tothis Rethink Podcast and it's a
really special edition thatwe're filming here live at G CT
GTC 2025.
It's been a long day.
It's been a long day.
It's been a long way, and thisis our Silicon Futures program.
(00:31):
And, of course you all knowwe're doing this in
collaboration with TEUs Researchand Cambrian hyphen AI research.
You don't have to say hyphen, Ilove saying hyphen ai.
Well, you know, it, it meansthat I care.
Oh, right.
It means you can spell Yeah,it's one extra, it's one extra
token.
Man.
It's like the brown m and ms.
(00:52):
Okay.
You just do not have brown m andms.
Okay.
In a hotel when Van Halen is intown.
So that, that, that is thehyphen.
But, I'm joined by, JimMcGregor, as well as Francis Ko,
as well as the illustrious CarlFre.
(01:13):
Right.
And I'm sure everyone out therein AI land is really glad to see
us here together in San Jose.
At the, San or we're at the,actually, the Hilton Insignia.
Hilton Insignia.
Signia.
Okay.
Well, we've been getting bouncedaround, back and forth, all over
the place.
So gentlemen, this is like thebig Nvidia, good event, AI
(01:34):
events, super Bowl of ai, right?
Yes.
In all of the universe.
And so, looking forward tohaving this chat and, it should
be a good one, right?
Mm-hmm.
So, hey, let's talk about.
Impressions.
What'd you guys think?
What did you, number one, what'dyou guys think about this year
and what did you think wasdifferent from last year?
Karl Freund (01:53):
I think everybody was expecting really fast
hardware and a roadmap ofreally, really fast hardware
uhhuh.
so that was not a surprise.
They met the bar.
I think what surprised us, Ithink, at least, certainly me,
was two things.
One was the optical.
Yes.
The co silicon
Jim McGregor (02:12):
Photonics.
Karl Freund (02:13):
Thank you.
Silicon photonics.
Yeah.
I can't even say it.
And the other is, is reallydynamo, which is the new
operating system as they callit, which effectively think of
it as, as, the hypervisor forhundreds of thousands of GPUs.
Yeah.
That optimizes across all theGPUs, depending on.
(02:35):
What you're trying to do.
I'm still trying to wrap my mindaround it or, yeah.
Jim McGregor (02:39):
Or let's just call it the operating system for the
AI factory.
Karl Freund (02:42):
Yeah, yeah.
The operating system for the AIfactory.
But it's kind of like ahypervisor, right?
Yeah.
Leonard Lee (02:46):
Kubernetes, it's been described as sort of
Kubernetes for AI inference, Ithink
Karl Freund (02:51):
across.
Hundreds of thousands of gps.
I think
Francis Sideco (02:54):
that was, obviously the networking bit,
which I think probably Jim'sgonna talk about.
But I think for me it, it, itwas that, Jensen definitely
tried to make a case.
I mean, he's been under pressurein terms of the industry going
from, obviously we're stillgonna do training workloads, but
we're gonna be doing inferenceworkloads a lot more and the
importance of that.
And I think just generallythey've gotten hit in terms of
(03:14):
them not being.
Potentially as optimized forinference or not.
Mm-hmm And I think Jensen duringthis show definitely made a case
for the fact that actually whenyou start looking at the
transition of inference, notjust from one shot inference to
reasoning inference, youactually do need scale up.
In inference, which before that,came out, the question had
(03:37):
always been, do you actuallyneed scale up for, for
inference?
and he made the case that youdo, whether or not that case is
gonna get bought.
I think the jury's still out,but it passes the smelt as
especially Yeah.
When you bring the dynamo intoplay.
Jim McGregor (03:51):
Yeah.
Francis Sideco (03:51):
And I think that was the key for me.
Uh, and
Jim McGregor (03:53):
Agentic
Francis Sideco (03:55):
ai and agentic ai, that was the key for me is
that the homogeneous.
Fungible resources that Dynamocan help manage, so, right.
Jim McGregor (04:03):
Okay.
I don't know.
For me, it's like drinking froma fire hose and the fire hose
gets bigger every year.
But the stuff from Cosmos to dosimulated, or synthetic data.
Mm-hmm.
And, and.
Scenarios, just thousands,millions of scenarios using
synthetic data, right?
To train robots that is gonna,and autonomous machines.
I think that's huge.
I think seeing some of the newplatforms, they have a complete
(04:26):
enterprise platform solution.
Now, from, PCIE cards to, whatused to be Project Digits, now
it's, DGX Sparks, sparks and theDGX station.
So you have desks.
Desktop and side solutions forAI developers.
You have the DGX Super pod, youhave, the next generation of
(04:48):
Blackwell Ultra coming out.
So it's just one thing afteranother, and obviously, I.
The, the networking, networkingstole the show, I think for the
second year in a row.
Yeah.
Last year it was the, envy linkswitch, which, allowed 72 GPUs
to act as a single one.
Right.
In a single rack.
But this year, once again,everyone's been saying, and
(05:08):
this, we've been developingsilicon photonics since about
2000 actually earlier than that,but.
Everyone keeps saying it's twoto three years away.
It's two to three years away,right?
They finally got an ecosystemtogether to actually produce
something that they're, gonna beoffering a quantum X infin bend
version of it and a specxethernet version of it.
(05:29):
So they're gonna have, Uh,silicon or they're gonna have,
photonic network switches.
That is significant in terms ofreducing components, power
consumption, increasingresiliency and reliability.
Reducing
Karl Freund (05:43):
it all ships this year.
Jim McGregor (05:45):
Oh yeah.
Well, the switch next year.
The infin, the Quantum X shipsthis year.
The max spec Max next year.
Spec max next, early next year.
Yeah.
Francis Sideco (05:52):
And it may, it makes sense.
I mean, they did scale up lastyear with NVLink Switch and
then, scale out this year, and
Jim McGregor (05:58):
I think he points to what we're gonna see.
I would expect futureenhancements to NVLink, so that
we see more than 72 supportedmore than 144.
2, 280.
Well, he already, he alreadykind.
Yeah.
We've already.
Seen that with previewed
Francis Sideco (06:11):
that with where they're going with Vera Ruben.
So
Jim McGregor (06:13):
yeah, with Reuben and Reuben Ultra, I think that,
obvious, I would expect, eventhough he said No, use copper in
the rack, by the time we get toRuben, I almost think that we're
gonna be looking ats atphotonics in the rack.
Well, he's
Karl Freund (06:25):
double down on it this afternoon, right?
Yeah, yeah.
I mean, he, this morning hesaid, no, we're gonna stay.
Copper in the rack, so I'm notconvinced
Francis Sideco (06:33):
of that one.
We'll see.
Well, he said use copper as muchas you can.
And I think that's the key.
As, as long as can, as long asyou can and as long as you can,
is the key factor.
That might not be that long.
So
Leonard Lee (06:44):
what about you?
Well, you guys have left me withnothing but scratch once we go
down the line.
Say, but I do have to saysomething.
We're, this is all likeunscripted of course.
Obviously we don't have anythingto refer referencing, so I need
to do the disclaimer.
Welcome.
And by the way, all the opinionsof these guys are entirely their
(07:05):
own, don't reflect next curve ormine, which is probably good and
completely accurate because theylike to be up on me.
Right.
I'm like, like the black sheepof the family.
Yeah.
Yeah.
Especially Carl.
But, I thought it was really.
Overwhelming, a lot of, detailson the roadmap where I think
like what Jensen has been tryingto emphasize, I think, over the
(07:28):
course of GTC is that, they're adifferent company.
Mm-hmm.
I think that's a, and we've beentalking about how they're
becoming more systems oriented.
And, Jim often services.
Right, right.
And then the software aspect andso.
All this stuff is comingtogether in a really compelling
way and I think Jensen does areally great job of, telling the
(07:50):
story.
Now, I do have to say there aresome gaps in what I've heard,
and I think these are areaswhere Nvidia really needs to,
like all the scaling stuff thatdidn't work for me.
It still doesn't work for meeven from last year.
Right.
And how does Huang's Law fromlast year now map into these
new, this new scaling thesisand, I'd like to see more
(08:13):
clarity on that as we goforward, but they are also
introducing a lot of stuff andthey're trying to figure out how
to frame this.
Whole scaling argument, againstthe backdrop of a lot of what's
happening in ai, right.
With moving toward reasoning.
Last year was about MOE, right?
Mm-hmm.
Yeah.
In terms of the ai, make suremakes sense now.
(08:34):
Yeah.
We have the time scaling, thislong thinking that Jensen talks
about where the industry istalking about in general.
And, these are all new conceptsand they have implications on
the requirements around compute.
And we hear Jensen talking aboutthe a hundred x, right?
With reasoning, we need, ahundred times more or more
(08:55):
tokens, compute, which isdifferent from tokens.
That was one of the things thatI took away.
As we talk about tokens, we needto think about it not just in
general terms, but very specificterms in
Francis Sideco (09:05):
well terms
Leonard Lee (09:06):
of the, inputs and then what.
It gets generated in terms of atoken and how that translates
into how much more compute youneed.
Mm-hmm.
And so, there, I mean there'sall these layers that are being
placed on the conversation.
Yeah.
And that makes it really complexand I think, difficult for,
(09:26):
folks to.
Keep up with, I think.
Francis Sideco (09:29):
I think that's one thing that we do all need to
keep in mind, right?
Because obviously Jenssen'sgonna keep pushing, Jenssen's
gonna keep pushing theboundaries of what's needed.
But that doesn't mean that's theonly way.
Enterprise AI or AI in generalis going to create value for the
economy at large, right?
Yeah.
There's going, there's a ton ofbifurcations that are happening
(09:52):
in terms of use, in terms of,how you architect these data
centers for what differententerprises need and so forth.
So not everything is going to bethe top of the line, Uber scale
stuff that, that jenssen'stalking about.
There's going to be room forother architectures and and
other use cases that don't needthat.
(10:13):
That, screaming high endcapability.
Jim McGregor (10:15):
Well, and especially with the
infrastructure requirements thatcome with'em in terms of the
cooling and power and everythingelse.
Everything, yeah.
Yeah.
But one thing that really amazed me.
And, they made this change acouple years ago to combine, uh,
their automotive stuff and theirother robotics.
Robotics and other stuff, andinto autonomous machines.
And hearing from some of theother groups learning that even
(10:38):
though avs have kind of beenpushed out and fallen by the
wayside mm-hmm.
That a lot of these other groupsfocused on robotic machines for
healthcare and other types ofapplications, how they've taken
that learning that's alreadygone into automotive, even if
it's not being deployed, howthey've taken that learning and
applied it to other forms ofautonomous machines,
particularly robotics.
(10:59):
Yeah.
Yeah.
Leonard Lee (10:59):
Oh, really quickly.
Yeah.
I wanted to bring this upbecause I thought it, we've been
talking about digital twins fora long time.
Mm-hmm.
And quite honestly, a lot ofconversation, especially.
Like two, I think about twoyears ago it was nonsensical.
This year actually, the stuffwith Cosmos, right?
Mm-hmm.
Yeah.
And synthetic data.
What I think we all need torecognize is we're not talking
about synthetic data in terms ofa SOA generation.
(11:22):
This is like using some of thephysical models that they used
to have.
Mm-hmm.
And all the things that theyhave in gaming, it's like has
gaming roots that are used tomodel like certain environments
and then they augment that withAI to, synthetically generate.
Scenarios New data.
Data, yeah.
Right.
Scenarios.
(11:42):
And I think, that's where, somefolks might have been confused
in the past and what is all thesynthetic mm-hmm.
Data stuff.
But, one of the things that Iwas really impressed with this
year is, how it will essentialin supporting a lot of the.
Test acceleration, right?
Mm-hmm.
That are gonna provide thelearning environments and
learning experience for thesemodels that are gonna be
(12:04):
trained.
And this has roots, like whatyou're saying in automotive and
just there's this, all thisheritage that's coming together
mm-hmm.
In some of these concepts that,I think are, actually pretty
powerful.
Karl Freund (12:16):
So I think one of the things that, I'll never
forget the quote from thekeynote, which is Jensen calling
himself the Chief RevenueDestruction Officer.
I.
That's, that kind of rings truebecause when they announced they
were going to an annual productcamps, a lot of us who've been
in the computer industry a longtime said, that's gonna be
really hard to do.
Customers don't like having whatthey just bought be obsoleted so
(12:38):
quickly.
But he really embraced it.
He really embraced it.
And he said, he said, yes.
Broad, Blackwell, is shippedthree times more than, already
this year mm-hmm.
Than Hopper did all of lastyear.
And then he said that, revenueramp is gonna continue to drive
forward.
Jim McGregor (12:58):
And with Dynamo, they're getting a 40 x.
40 times improvement overHopper.
Karl Freund (13:05):
Yeah.
Jim McGregor (13:05):
That's just phenomenal.
Karl Freund (13:07):
It's wild.
Yeah.
You, that's an inferencestatement.
Yeah.
Right.
This hard inference is really atthe heart of everything they're
doing this year at GTC.
Yeah.
Whether it's for robotics or forthe data center, or even for the
little, miniature.
DGX.
The, yeah,
Jim McGregor (13:23):
The Spark.
Karl Freund (13:23):
Yeah.
Jim McGregor (13:24):
Love that.
Karl Freund (13:24):
Wanted to call it digit.
I want one.
It's called Spark.
I want one.
I one so bad.
It's super cool.
We all want one.
Leonard Lee (13:31):
Yeah.
So Jensen send us, we all wantedDX Spark.
Yeah.
And a workstation and station,right?
That would be great.
Like, and then sign up please.
Okay.
I'm not going that far.
Oh no.
His jacket by the way, everyone.
Yes, he was wearing his originaljacket once again.
So that jacket that he wore atCES must have been a CES
(13:54):
special.
Yeah.
So, yeah.
Well, he wore two differentjackets
Jim McGregor (13:57):
today.
Yesterday.
Did he?
Yeah.
Yeah.
Oh,
Leonard Lee (14:01):
he's got, he got a lot diversity in his jacket
collection.
He does.
I wonder if
Karl Freund (14:05):
he wore that jacket or a suit when he went to the
White House.
I don't know.
I didn't see any videos of hisarrival.
Maybe it's a simulation.
Francis Sideco (14:12):
Yeah.
But let's not underestimate whatyou said, Leonard, about the
simulation and like with cosmosof being able to create
synthetic mm-hmm.
Training information.
If you really take it intolayman's terms, that's like
being able to, force feedtraining Yeah.
To a human being.
Like the way they did it inMatrix, frankly, it's just like
(14:34):
shoving it all in, in, it's likedriving a billion miles.
It's like driving a billionmiles in a very short amount of
time.
And there's the sheer amount ofexperience that you're able to
gather one of and
Jim McGregor (14:44):
apply one of, one of the stats was 27 years of
simul or of what would be realtime, physical training.
In, what was it?
Francis Sideco (14:54):
In one day?
One day on one GPU.
and that's assuming in realworld you actually encounter
those situations.
Yeah.
Dropping Godzilla in the middleof the freeway happen.
Yeah.
You might not do that.
But the fact that you're able todo that kind of concentrated
learning, you can, IM, you canimmediately see how some of
(15:16):
these automated systems.
Are becoming more highly trainedthan a human because you can't
do that.
Well, it's
Karl Freund (15:22):
given me renewed confidence that we will solve
the.
Automated driving.
Well, you problem.
Leonard Lee (15:27):
And and it's not new.
'cause they were, they weretalking about this stuff like
four years ago.
Yeah.
You know, it just is at the sortof like the height of like the
av, av yeah.
Mm-hmm.
Pipe.
Mm-hmm.
But its
Francis Sideco (15:38):
application wasn't as.
As clear Well, how you can seehow well divers diversified
Leonard Lee (15:44):
its application.
I think, I think it was stillclear and necessary back then,
right?
Yeah.
In accelerating the, thetraining of, AV systems.
Now, I, I mean they're broadenit, especially with their focus
on, um, robotics, right?
Yeah.
With Isaac group, right, right.
Yeah.
Being
Jim McGregor (16:01):
able to take, human actually doing something,
combining it with videos ofhumans doing something.
Yeah.
And then using AI to createsimulations and solutions.
Yeah.
And just phenomenal.
I know that even a lot ofsurgery being done today, like,
knee replacements and shoulderreplacements and stuff like
that, they're actually usingrobotic arms to do all the
cutting.
(16:22):
Yeah.
And everything else, it'samazing that just blows me away.
Karl Freund (16:24):
One of the things that, that I thought was really
interesting was the discussionabout the automotive industry
and the, and NVIDIA's approachto the automotive industry,
which I always thought it wasprimarily in vehicle.
And now I'm starting to realizethat's nice.
If it, if they have an Nvidiachip in the vehicle,
Francis Sideco (16:40):
that's a what they're really after is the
Karl Freund (16:42):
cloud for simulation, for the vehicle
operation, for the training.
Yeah.
And the command and control.
And an important point was thatonly 5% of the vehicles in the
world today are level two andabove.
Jim McGregor (16:53):
Yeah.
Karl Freund (16:54):
So there's a huge opportunity there.
Yeah.
And the opportunity is not justin the vehicle, it's all a
simulation.
All those vehicles, all thoseOEMs will require before they
put those cars in the road.
And,
Jim McGregor (17:05):
And while they did back out of the infotainment
solutions.
Mm-hmm.
Now they've got the wholepartnership with media tech.
Using their technology, theirIEP to develop an I develop by.
I'm surprised.
Karl Freund (17:16):
Hear an update about that.
Jim McGregor (17:17):
Yeah, I was kinda surprised too, but I would
expect that we will.
Funny.
Francis Sideco (17:22):
Well, on the automo, on the automotive
vehicle bit, right?
I mean put it in in perspectiveright now, like in Nevada, you
train your 16-year-old on 50hours of drive time and they
give him a license.
Jim McGregor (17:34):
Oh my god.
Francis Sideco (17:34):
Okay.
And like 30 hours of classroomwork, 50 hours of drive time,
they gave him a license.
Mm-hmm.
Okay.
Compare that with training a veuh, an automated vehicle with 27
years worth.
Of basically drive time.
Yeah and it just, it puts it inperspective.
Well, as long as it's good
Jim McGregor (17:51):
driving.
I was just gonna say, it couldbe really bad driving actually,
but hey, no.
Well, that's the thing about
Francis Sideco (17:56):
The synthetic data is because you can control
the, you can control thesynthetic data.
It's not garbage and garbage outbecause you can control the
quality of the synthetic.
Well, but,
Leonard Lee (18:05):
but I'll argue this, you need to have a crap
driving algorithm to create.
Stupid situations in your Eurosynthetic data.
'cause that's actually whatYeah.
A vehicle needs to respond to isto bad drugs because, there's
only finite numbers of way todrive properly at any given
time.
There's infinite ways ofdriving.
Really crappy crap.
(18:25):
Yeah, yeah, yeah.
Exactly.
And they can get 90%
Francis Sideco (18:27):
of that in Vegas.
'cause the drivers thereshocked.
I was trying to get this guy to
Jim McGregor (18:31):
it, it, it strangle me.
It, it, it really needs toaccount for me on the road.
Yeah.
Oh yeah.
Terrible.
The car that's going three timesfaster, my god.
Crazy.
Geez, that's really horrible.
I actually used to terrorizethe, Waymo's in Phoenix just to
see how they would react.
Oh.
So there was one Jim sittingthere poking the beer.
(18:54):
I would, I'd actually, there'sone place on Mill Avenue where
it goes down to one lane and Iwould actually pull it next to
the Waymo's and test them andsee how they react.
Hopefully
Francis Sideco (19:02):
without a passenger in there.
But no,
Leonard Lee (19:05):
you, I wanted to take things back to the new AI
supercomputing lineup.
Right.
And last year they introduced,what the transformer,
transformer engine, right.
Is that what that was?
No, that was, no, no, no.
That was.
Karl Freund (19:18):
Now transformer engine was,
Leonard Lee (19:19):
two years ago.
Yeah, two years ago.
Right, right.
But how there's a lot more talkabout, so last year they
introduced transformer engine totwo, right?
Yeah.
For, yeah.
And so with the support for, FPfour, right, exactly.
But how now that's becoming evenmore important.
There's a lot more talk about,uh, mixed precision.
(19:42):
And the role that, it plays indriving the, the scaling of
these systems.
Right.
And so when we look at, one ofthe things I noticed, if you
look at the roadmap, the chip,it's, the individual GPUs
themselves are not scaling asfast as everything else around
it.
We talked about networkingalready, but it is interesting
(20:03):
to see how, you have the memory,the HBM, right to, the upgrades
there that are happening overthe course of the next, two
iterations of the system as wellas the networking.
Really playing a outsized role,in, in scaling the systems.
Yeah.
And I, one of the things Ithought was really interesting
was Jensen, trying to clarifywhat a GPU is.
(20:26):
Mm-hmm.
Mm-hmm.
Right?
Because everyone thinks it'sjust the chip, but in actually
their parlance it, it issomething much bigger.
Right.
And, we see this, I guess it's aconfusion.
E, especially amongst the mediaabout what a GP is.
Just turn on a TV and have, Idon't even to call'em GPUs.
Yeah, it's not really a GPU,right?
(20:47):
I just called an
Jim McGregor (20:48):
accelerator, or I call it a GPU accelerator.
'cause it's really anaccelerator.
Yeah, but it's a accelerator.
It's it literally is a freaking,you look
Francis Sideco (20:56):
at the GPUs that are in like the G-G-D-G-X sparks
and all that stuff, they don'tdo graphics.
Let's ask this question.
Okay, let's go down
Leonard Lee (21:04):
the line here.
What the hell is a GPU?
Now is
Karl Freund (21:08):
Nvidia Accelerator.
Well.
Jim McGregor (21:10):
It's a graphics the best you can do.
It's a graphics processing unit,but these things aren't GPUs.
Francis Sideco (21:16):
Yeah,
Jim McGregor (21:16):
that's the problem.
Francis Sideco (21:17):
Yeah.
It's K FFC is no longer justKentucky Fried Chicken.
Right.
It's just KFC and it's the samething.
It's just another, it's justanother, it's another, where
it's not an acronym are bringing
Leonard Lee (21:26):
out Kentucky Fried Chicken as.
Source of an analogy.
It's been a long day.
He's been king a long day.
Oh my gosh.
The king food analogy, bringingup Pizza Hut.
Talking about splitting up pizzapies.
Yeah.
Well, no, I mean, I think theimportant thing to understand is
(21:46):
like what you alluded to or youpointed out before, is that it's
a logical it.
It's a logical gpu.
Massive, right?
Mm-hmm.
And I don't know if that's quiteregistered in everyone's mind,
but it's an important point for,everyone who's observing the AI
industry and AI supercomputingto clearly understand.
Mm-hmm.
Right.
(22:06):
I think
Karl Freund (22:07):
one of the, interesting conversations with
Jensen was about what do youwant NVIDIA to be known as?
What do you want the Nvidia tobe known for in three to five
years?
Mm-hmm.
And the answer was not.
The data center, the, which iswhat many of us were expecting.
What he said was that Nvidia isfoundational mm-hmm.
(22:30):
To the entire world forartificial intelligence.
Yeah.
And I thought back to theslides, he flashed up really
fast, too fast for any of us toread, during his keynote, which
was of all of his major partners
Jim McGregor (22:45):
Yeah.
Karl Freund (22:45):
And how they're adopting.
Ai.
And if you look at each one ofthose, they had these little
green squares on them.
Those green squares were allnims.
Okay?
Nvidia Inference, inferenceservices.
And that to me is foundational.
That is more of a competitivemoat than Cuda.
Jim McGregor (23:08):
Well, it's an entire stack.
It's an
Karl Freund (23:09):
entire stack, and you have the entire industry
adopting that entire stack,which is gonna make it really
hard for a MD or Intel.
Well, and
Jim McGregor (23:19):
I don't think it's just, uh, I don't think it's
just that, when you look atDynamo mm-hmm.
Being an operating system for anAI factory or even a data
center, when you look atOmniverse, which is, you know,
depending on how you look at it,you could actually classify it
as an operating system for thecloud.
Yeah.
So there are, not to mentionthese endless, almost endless
(23:40):
foundation models are creatingfor each individual vertical
segment.
Francis Sideco (23:44):
Yeah.
So just going back to my earlierstatement though, I agree
completely that Nvidia is.
Is continuing to push theenvelope.
They're gonna get a lot of valueout of this and a lot of return
on their investment.
But that does not mean thatthere's no, no place and no role
for other competitors becausenot everybody needs like the
Uber scale.
Mm-hmm.
(24:05):
Training or inference tofrankly,
Jim McGregor (24:08):
there, there's still a lot of traditional AI
solutions out there, and AI atthe edge
Francis Sideco (24:13):
there, there's a lot of a AI at the edge.
And even when you look at thecompetitor, the more direct
competitors to, to Nvidia, I'msure they're not.
Standing still, right?
Yeah.
and they're going, there is anappetite, for not just one, one.
We don't want to end up withanother kind of Wintel
situation, from a, from.
20, 30 years ago.
(24:33):
And I think the industry thereis an appetite for other
options.
And especially at the edge.
Especially at the edge, there'sopportunity
Jim McGregor (24:39):
that, and for specific applications cases,
that's why specific we still seehyperscalers doing their own
silicon.
Yeah.
Because they have use caseswhere they know they can
optimize for.
Leonard Lee (24:49):
Yes.
And those are, so one of thethings that I think it still
has, it's still not beenclarified.
Or is well understood is thatthere is a difference between
hyperscale and cloud serviceprovider.
Oh yeah.
Right?
Oh yeah.
And that, for the hyperscalerslike meta who have singular
workloads that are massive, itmight be a recomme scale to
(25:12):
optimize.
Yeah.
I mean that, that's gonna be aproduction application that req
requires that level of massivescale where yeah, you're going
to be going custom.
You're going to, tune the entireinfrastructure singularly to
run, that set of workloadsoptimally, right?
(25:33):
It's an operational system and,for the cloud and for the cloud
service providers and the modelbuilders, that's where I think.
Nvidia has its sweet spotbecause it's programmable, it's
more, quote unquote generalpurpose, right?
It has the flexibility to thesoftware where the, the hardware
(25:54):
now can be used in a widedifferent range of, purposes,
right?
And projects.
And that's a different.
Value proposition almost.
But I, I do feel that Nvidia istrying to, break that, notion,
right?
That no, actually we're good foreverything, right?
(26:14):
And that mm-hmm.
It's an argument obviously,that, they'll want to make and,
and but it's important tounderstand that dynamic.
Karl Freund (26:22):
Yeah.
Jim McGregor (26:25):
Yeah.
And we should also note that.
This is just mayhem.
They have outgrown San Jose,period.
they shut down most the roads.
Yeah.
they have to do it somewheredowntown San.
This is ridiculous.
Yeah.
Took over
Karl Freund (26:38):
the entire park.
Yeah.
For, for food.
Well,
Francis Sideco (26:40):
hopefully they'll be out in Vegas at some
point, but
Leonard Lee (26:43):
Yeah.
Yeah.
So, let's talk about enterprisegeneral ai Understood Uhhuh.
Did you?
Well, I talked about thesystems.
No, you saw, talked about thesystem, but enterprise, because
one of the things that they'rereally trying to push for this
year, because you're seeingthings like, I don't know if you
noticed, there's like a whole AIdata, data, and storage movement
(27:06):
going on, right?
People want to get a handle onthe.
Enterprise data and figure outthat out.
Right.
Which is different from whatwe're seeing with what the model
Yeah.
Builders are producing orpursuing, which is, hey, how do
we get as much, publiclyavailable or maybe in
copywritten data as possible forus to train these foundation
models.
(27:26):
But for the enterprise, it'sdifferent, right?
Yes.
And you're now, this is reallyhow do we make data available
for compound AI applicationsthat are gonna enable us to
take, advantage of this new AIcap generative AI capability.
I wanna make that distinction.
'cause everyone, talksgenerically about ai, the thing
(27:48):
that's really caused the stir inthe last two years is generative
ai.
Right.
The big question.
And now we just
Jim McGregor (27:53):
AI And those are two different things.
Leonard Lee (27:55):
Yeah, they are.
I think they're closely boundtogether in, in a certain way.
Well, in, in terms of function
Jim McGregor (28:05):
or, or, yeah.
Well, generative AI is thought,agent AI is reasonable.
So, um, yeah, you know what?
I'm glad you brought that up,okay.
Keep going.
One is going to give you afeedback on everything it's been
trained on.
The other one is actually gonnasit there, think about it, and
use a lot of differentinformation that's available to
(28:26):
it to actually make a judgmentcall.
And it may be significantlydifferent than Gen ai.
Francis Sideco (28:31):
And you add onto that, it's abil, it's agency,
it's ability to take thatjudgment call and then act upon
it.
Take action, take action.
I think that's'cause a lot ofpeople are using agent ai.
To mean basically copilotassistance or that's completely
different.
Yeah.
Leonard Lee (28:46):
And you know what, so the reason why I'm glad you
brought that up is there is asemantic disconnect out there
right now.
Yes.
The way that people talk aboutAG agentic ai, it's as certain
circles, well, there'sassistance, but then.
Now, like when you, what you seea lot of ISVs talking about is
AgTech AI in the context ofautomation, process automation.
(29:10):
Right?
Right.
And so there is really thisYeah.
Like disconnect that's happeningand that's what they call, they
call it semantic Dspecification.
That's actually a, actually,it's, it goes the other way
around.
I can, I
Francis Sideco (29:24):
can't even get my semantics in my head straight
on that, but I mean, that isreally important if you're
talking about agent ai and thefirst thing outta your thought
process is, a work, a processworkflow.
Mm-hmm.
That's not a agentic ai, that's a, that's an
automated process workflow.
That.
That's different than actualagency.
Don't say that.
(29:44):
Yeah.
you're gonna
Leonard Lee (29:45):
make some ISVs very unhappy by saying what you just,
Francis Sideco (29:49):
well, here's the, here's the thing.
We have to make sure that weunder, I mean you can, from a
marketing standpoint.
Okay, fine.
Use that.
But if you're really trying tounderstand the implications to
the technology and what you canor can't do about it and what
kind of infrastructure you needto support it, you have to be
clear about those nuances.
Leonard Lee (30:09):
Well, yeah.
And then we also have toreconcile that against what
we're hearing, from the, a lotof when we were in the A IPC
world and the on.
Device AI world.
Mm-hmm.
How they talk about ag agenticai which is personalization,
it's doing things for you.
Yes.
These are automation concepts.
(30:29):
Action.
Well, there's
Francis Sideco (30:30):
there's context.
Yeah.
There's the context element toit.
There's the judgment and thenthere's the action, which is
exactly agentic ai.
Leonard Lee (30:39):
Okay.
Karl Freund (30:39):
Yeah.
Without action is not agent
Jim McGregor (30:41):
And enterprise is, really trying to figure out how
do they, they use all theinformation in free and open
source models that are outthere, combine it with their
information and then whatresources do they need on prem
and the cloud.
Everyone's looking at hybridsolutions, but it is.
And it's a, it's, a colluschallenge right now for them to
figure out.
(31:02):
But I think over the next yearwe're gonna see a little bit
more clarity and ability to beable to do that.
Francis Sideco (31:09):
And I think where enterprises need to start,
and this is where they kind, getlost a little bit, is there's so
much possibility out there.
They're like, they don't evenknow where to start.
And I think really, if you'regoing to be serious about
enterprise ai, you really needto start and.
This is from every singleconversation I've had is start
with specific problems thatyou're trying to solve for your
(31:30):
business, and then let thatdrive what model you're using,
what architecture, hybridon-prem, whatever you're using,
what you know, data you need tobe bringing in and so forth,
governance, all of that stuff.
But you have to start with a,with specific use cases and
problems that you're solving.
Absolutely.
So.
Leonard Lee (31:50):
So I did want to give a shout out to IBM.
I know you guys like doing that,right?
So I'm gonna, I haven't done it.
You guys have, but I'm gonnagive a shout out to IBM because
they're the only company I'veencountered in two years that
are, it's, that's actuallyaddressing, enterprise AI
security for ai.
Yeah.
They have a thing called,context aware storage Folks.
(32:14):
Keep an eye out on that.
And also.
Here's a name that you need toknow.
His name is Vincent Su, HSU.
He's an IBM fellow, and he isworking on this stuff.
And, they were collaboratingwith, Nvidia to integrate what's
called, context to Wear Storage.
Into the NIMS framework, right?
(32:35):
Mm-hmm.
And this, I think this is reallyimportant work.
Yeah.
Jim McGregor (32:37):
And
Karl Freund (32:38):
I think we'll hear a lot more about next week in
New York.
Yeah.
Jim McGregor (32:40):
That and then combined with their granite
models, which are really theonly lab enterprise ready.
AI solutions.
Yeah.
Well it
Francis Sideco (32:50):
From a security, and guardrail standpoint, but
also couple that with instructlab, where there is, where it
makes it really easy to thenbring your enterprise data
context in, into the, to retrainfor, specialized
Jim McGregor (33:04):
model.
Yeah.
But
Francis Sideco (33:05):
again, even then.
Even in my conversations withIBM, they say it's very
important for you to know as anenterprise exactly what problems
you're trying to solve.
Yeah.
'cause then that's when thetools really sink.
Leonard Lee (33:16):
Yeah.
And you know why I'm bringingthis up is that, I've identified
this gap in rag architectures.
This is gonna resolve the, I'm,that might be premature to say
it will resolve, but it startsto address the issue of being
able to apply fine-grainedsecurity controls.
Yes.
On top of a rag.
Mm-hmm.
And so whether it's, graphdatabase.
(33:37):
That you're dealing with, orVector database.
A lot of folks don't know thatyou can apply, fine grain
controls on top of a vectordatabase.
They're, they were neverdesigned to do this kind of
stuff.
Yeah.
And so they're doing somecreative, abstractions of
controls on top of, and this issomething I need to find out
more about, but I'm veryexcited, at least IBM recognizes
(33:59):
this problem because for thiswhole time people have just been
focused on.
Hardware level security, networklevel security, but not really
looking at the application andthe data.
Francis Sideco (34:10):
And you can see the convergence there.
Aside from that, that,collaboration, they also
announced a collaboration withNVIDIA around their IBM
consulting, engagements, becauseit is right now requiring.
A lot of consultation forenterprises to figure out, okay,
what is their use case?
What are they trying to solve?
Leonard Lee (34:29):
Yeah.
Francis Sideco (34:30):
Um, and then go from there
Karl Freund (34:31):
and what's the state of the data
Francis Sideco (34:32):
Yeah.
And what's the state of thedata.
Absolutely.
That's a key one.
Leonard Lee (34:36):
Yeah.
So, yeah.
Cool.
So, yeah.
And so what do you guys expectnext year?
I, I guess you, you already knowwhat to expect.
Already know that's a
Jim McGregor (34:45):
bigger fire hose.
Leonard Lee (34:46):
Yeah.
Bigger, bigger fire hose.
Jim McGregor (34:48):
Come on, let, let's do this.
We gotta do this.
Well, we already know.
We already know.
Next year, do we?
Next year it's,
Francis Sideco (34:52):
do we, it's Ruben, right?
Yes.
Next year will be Reuben.
It's very Ruben.
Yeah.
Karl Freund (34:57):
Next year it'll be Ruben.
Francis Sideco (34:58):
And then on the networking side, they went scale
up last year.
They did scale out this year.
So we'll see what's next.
Next up for that,
Jim McGregor (35:05):
They seem to be strategically targeting.
Each bottleneck in the datacenter, especially in the data
center.
Something around ai.
So I, definitely I think we'llsee enhancements in memory and
everything else.
And networking.
Yeah.
I'm hoping that they startaddressing the bottleneck that's
emerging out of theinfrastructure for the data
center.
(35:25):
Because right now all thecooling and, power solutions are
really customized to eachimplementation.
Yeah.
Yeah.
And that's gotta change.
When you're spending hundreds ofmillions of dollars on a data
center, yeah.
You have to, you don't want todo it each time you're employ,
you're, putting new GPUs intothe data center.
(35:46):
So it's gonna be interesting.
Do you realize what you'resaying?
Yes.
Oh my gosh.
It has to be fucking play.
I, on that note, I think isprobably, I worked for the
Motorola computer.
Matter of fact, I'm surprisedthat they're not looking at
alternative structures likeblade servers.
Leonard Lee (36:01):
Ah, here we go.
I'm serious.
He's back on the soapbox.
Blades are good.
Blades are good.
All right.
On that note, on that note, no,there's nothing wrong with that.
No, I think it's a greatcomment.
Yeah.
So seriously, you guys, there'sno expectations.
You know everything for nextyear.
No, no,
Karl Freund (36:21):
no.
Leonard Lee (36:21):
Oh, absolutely.
Just
Karl Freund (36:22):
like we didn't expect optics.
Leonard Lee (36:23):
Okay, so he got one.
What about you?
Francis Sideco (36:26):
Yeah, I think I said me, I think, don't say
Leonard Lee (36:28):
that you, you're gonna just repeat.
No, no, no.
I think
Francis Sideco (36:30):
I said earlier when you first asked it was, I
think the memory subsystem okay.
Is going to become extremely, italready is extremely critical.
It's, it's going to become evenmore critical and I think, I
think we'll see probably someinnovations around that.
Karl Freund (36:46):
Yeah.
Carl.
You've stumped the band.
I, I, I really don't know whatto expect next year.
I thought I knew what I, what toexpect this year.
Mm-hmm.
I was wrong.
Yeah.
Right.
Was not expecting storage.
We got storage was not expectingoptical.
We got optical networking.
I don't know, may, maybe they'll do optical in the
(37:06):
rack.
Maybe that'll be the surprise.
I don't know.
But, it's gonna be a fun spaceto watch this space.
Leonard Lee (37:11):
Yeah.
Like for me, I mentioned beforesecurity.
Hopefully we'll see traction.
I don't know if we will.
It might be a pretty tough nutto crack.
And, this whole, KV cash thing.
Yeah.
With Dynamo.
Yes.
Well, that's, it's sointeresting.
That's the memory.
That's the memory that, becausehere's the thing, I wanna know
whether or not, because thisyear it's all about a hundred x.
(37:33):
How about the other way around?
How about compression and howabout reduction of the amount of
compute required because, it'sone, it, I understand why Nvidia
wants
Francis Sideco (37:44):
more.
Leonard Lee (37:44):
More.
Yeah.
But what about less?
Because someone's gonna takethat opportunity.
Right.
And that's, that's, and I think
Jim McGregor (37:52):
we're seeing that with their Jetson, even though
it wasn't really highlightedthis year.
Yeah.
Their Jetson platform forautonomous machines, whether
it's robots or cars or whatever,they're going there even though
that wasn't the focus of thisgtc.
Mm-hmm.
I still think that you're gonnasee more.
Leonard Lee (38:08):
And so that's what we might be able to expect.
Maybe Yes, at GTC 2020 6 26.
Yeah.
And with that, I guess we, wewrap it up.
We
Francis Sideco (38:17):
wrap it up, yeah.
Sounds good.
Well,
Leonard Lee (38:18):
gentlemen, that was fun.
It was great hanging out withall of you here at G.
It's been a fun week, GTC 2026and, to our audience, thanks for
listening in.
Remember to like, share, commenton this episode.
And remember to subscribe to thenext curve, rethink a podcast,
as well as our research portal,www.next curve.com.
(38:40):
And also, these guys arepublished
Jim McGregor (38:43):
everywhere.
Find us on Forbes.
Find us on E Times, eur,YouTube.
YouTube, a number of otherpublications and outlets and
definitely, look for us at, theother trade shows that are
coming
Leonar (38:55):
up@www.teusresearch.com.
Come on man.
You gotta front your companyproperly and of course, and
Cameron
Karl Freund (39:04):
hyphen hyphen hyphens a i.com.
Leonard Lee (39:06):
Yes.
And I do that because I don'twant people to like, to not have
that.
Yeah.
Put in the website.
Thank you very much.
Well, it's been fun.
Yeah.
And that's Carl Fre of, CambrianAI Research.
Karl Freund (39:18):
Well said.
Leonard Lee (39:18):
Yeah.
And so, until next time,remember to always tune into the
Rethink Podcast and our SiliconFutures, program here for the
tech and industry insights fromthe world of AI and like GPU
stuff.
We'll see you next time.
Bye-bye.
Thanks.
(39:38):
Bye-bye.
Popular Podcasts
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
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