WEBVTT

NOTE language en-US

00:00:00.240 --> 00:00:01.000
You're not a breakdown.

00:00:01.000 --> 00:00:04.363
What am I mean for this NVIDIA and everything else

00:00:04.430 --> 00:00:08.331
CNBC tech check Co anchor Deirdre Bosa and Deirdre, here's

00:00:08.398 --> 00:00:13.040
what's interesting yesterday Howard Lutnick, I pressed him on NVIDIA.

00:00:13.039 --> 00:00:16.136
I think you heard this and he said, well, we're,

00:00:16.200 --> 00:00:19.554
we're allowing them to sell the 4th best NVIDIA chip

00:00:19.619 --> 00:00:20.200
to China.

00:00:20.480 --> 00:00:23.480
Is that how others would characterize the what is it

00:00:23.538 --> 00:00:24.000
the H20?

00:00:24.239 --> 00:00:27.074
Yeah, the H20, I mean, it is not even close

00:00:27.140 --> 00:00:30.304
to any of Nvidia's best chips, but you know, the

00:00:30.370 --> 00:00:34.325
Chinese have done a really good job around optimizing around

00:00:34.390 --> 00:00:37.818
older hardware and that's why the H20 is valuable in

00:00:37.884 --> 00:00:38.280
China.

00:00:38.280 --> 00:00:40.319
But what I would say when we talk about this

00:00:40.365 --> 00:00:42.775
sort of NVIDIA picture and what makes Wall Street so

00:00:42.822 --> 00:00:44.119
bullish, talking about 5:00.

00:00:44.120 --> 00:00:46.635
And you're not crazy to say $6 trillion, Brian, at

00:00:46.685 --> 00:00:49.352
this point, it's not just about chips, it's about the

00:00:49.402 --> 00:00:50.559
whole NVIDIA ecosystem.

00:00:50.560 --> 00:00:53.706
We talked about this yesterday, but the CUDA software has

00:00:53.761 --> 00:00:57.680
been so critical to Nvidia's continued success and continued dominance.

00:00:57.680 --> 00:01:00.633
And that's also why the Chinese are prepared to take

00:01:00.690 --> 00:01:03.871
this less performant H20 chip because it operates on the

00:01:03.928 --> 00:01:04.439
software.

00:01:04.440 --> 00:01:06.600
And that's what everyone is building on, right?

00:01:06.720 --> 00:01:08.720
So that has been extremely positive for NVIDIA.

00:01:09.000 --> 00:01:11.828
And even without China, I mean, the China headlines when

00:01:11.879 --> 00:01:14.202
they couldn't sell, I mean, it was a huge hit,

00:01:14.253 --> 00:01:15.920
billions and billions of dollars.

00:01:15.920 --> 00:01:19.253
Yet NVIDIA still continue to do well because it's only

00:01:19.315 --> 00:01:20.920
a small part of the story.

00:01:20.920 --> 00:01:21.880
There's inference, right?

00:01:21.880 --> 00:01:23.240
That comes after training.

00:01:23.400 --> 00:01:25.695
Once you have the models, it's about how much the

00:01:25.741 --> 00:01:28.645
models are being used by enterprise, by consumers every single

00:01:28.692 --> 00:01:28.880
day.

00:01:28.880 --> 00:01:31.440
That is just like hockey stick going up and up.

00:01:31.840 --> 00:01:34.160
So it has inference in its favor.

00:01:34.160 --> 00:01:36.480
It has the huge CapEx.

00:01:36.480 --> 00:01:38.320
You mentioned, Brian, all these companies.

00:01:38.319 --> 00:01:40.520
You have Zuckerberg this week saying he wants to spend

00:01:40.561 --> 00:01:42.679
hundreds of billions of dollars in compute capacity.

00:01:42.680 --> 00:01:44.040
You know, Google's doing the same thing.

00:01:44.040 --> 00:01:47.480
You've got Project Stargate with Open AI and Oracle.

00:01:48.320 --> 00:01:50.200
You just got a lot going in its favor.

00:01:50.200 --> 00:01:53.594
And even though there's more competition in this phase of

00:01:53.653 --> 00:01:57.465
the AI race, inference versus training, it's in videos ecosystem

00:01:57.524 --> 00:02:00.442
that keeps it so dominant, not just in China, but

00:02:00.502 --> 00:02:03.479
everywhere in the world and it's rain to progress.