Let's start with Nvidia.
We say it's a moment in market's history.
It broke its own record, but it's one of the bigger positions you hold.
Denny. So when you see the right headline on the Bloomberg terminal, how do you
react? How does that make you think?
Well, I, I, I think, you know, the video has been a remarkable stock for the last
year and a half. And that's primarily been because of the
financial performance. I mean, the ramp in revenues that we've
seen with its A.I. chips and GPUs is like nothing we've
ever seen before. And you know where the market's out
right now. It's really this push and pull.
And what we've seen, if you just look at the Hyperscalers earnings so far and
particularly the CapEx forecasts, the market is just really hinging on every
word that we're hearing. And you know, effectively what we've
heard is that nobody's blinking and you

have seen CapEx go up at Google. You've seen CapEx go up at Microsoft, and you're seeing CapEx go up at Meta. And all of those companies seem to be implying that there's no end in sight, at least, you know, through 2025. And that should be good for Nvidia. And, you know, it's, you know, if we actually look at the other data points in the industry, you know, AMD just reported, and the numbers were fine, but, you know, it's really just a really, really small fraction of the amount of AI-related chips that Nvidia's selling. And Nvidia is embarking on its probably its most significant product cycle ever with Blackwell as it goes to a full systems approach this fall. So I ask you that by that specific point. You know, I've discussed with Jensen Huang the annual cadence of new product refresh, but the bit that's kind of

concerning or hard to understand is what
about all those companies. The Hyperscalers, but also others that
invested at the H 100 generation, and they may have only just got that
capacity online. And then suddenly they have the option or choice of whether
they buy the next cutting edge generation of chip upgrade.
And I think that Jensen is probably talk to you as well about what the ability to
swap in in and out those those generation of CPU quite easily. But it's
hard to fathom really. Yeah, well, what's really interesting,
and I'll give you a couple of data points.
Last night, Mark Zuckerberg announced this publicly.
He talked about that llama for which is the next iteration of their open source,
large language model that they just, you know, have have been evolving is going
to require ten X to compute that llama three required.
And why that's important is, you know, if you talk to the leaders in industry,
you'll hear a comment.

And that comment is we want tomorrow's
GPUs. And the reason you hear that comment is
because one of the single biggest bottlenecks to actually advancing these
large language models, whether it's open air and throw up a Google with Gemini
Llama, whatever it might be, you need increased compute power.
And that is exactly what Blackwell brings.
And so the important part for these companies is making sure that they have
the ability to actually create fungibility with the GPUs so that when
you use each one hundreds and then you're going to move to Blackwell, you
can actually use those H1 hundreds for inference or you can use them for other
parts of your of your company. You know, whether, you know, for
Microsoft, that could be to support Office 365 and co-pilot or GitHub
co-pilot or in Metters case, to support the recommendation engine or content
creation or whatever it may be.

And so, having that fungibility of the GPUs, but then also having the ability to lean in hard to the latest generation, so that you can push the envelope with these frontier models. Suddenly, this is a good point to go to Apple because it's a different story, right? They are not a hyperscale. They don't make the same direct investment in infrastructure. The CapEx has always been low, but the AI story is the same. It's a smaller position for you, but your approach going into this evening. Well, it's always hard going. You just never know, right here on the day of earnings. But here's what I can say. We were really impressed with Apple intelligence and what we saw. You know, the world was waiting to hear what they were going to deliver. It's not ready for primetime yet. This is unlikely to be a massive cycle, you know, with the 16. But as we start looking into next year.

and if they get Siri right and they're
actually there, here's the beauty of the Apple model.
All this CapEx, this 200 billion plus dollars that the Hyperscalers are
deploying into CapEx. Apple gets to leverage that because they
are the gateway to the AI-enabled Internet, you know, through the iPhone
interface. And I think that's really
underappreciated as it's related to Apple.
It's a very capital-efficient model in. Sorry to jump around, but Amazon, I
mean, I think with the focus that it switches back to us and whether the
CapEx story and the investment cycle is still intact.
Yes, you know, Amazon's a little different.
And the reason it's a little different is, you know, as I mentioned, if you
think about Alphabet, you think about Microsoft, you think about Meta, and
that fungibility, the element of GPUs and capital expenditures.
Amazon doesn't quite have that.

You know, they can't really repurpose a bunch of GPUs into their core business because it's not a daily engagement app like Office 365 or Copilot or Metta or Search. And it really does hinge on their ability to provide the underlying compute for other companies. And so they're really in a push and pull of, you know, having been the dominant provider of cloud computing services for the last 15 years and now pivoting their business to try to compete as aggressively with Microsoft and Alphabet in this next generation of cloud compute, which is an AI enabled.

Nvidia Shares Add Record $329B

Welcome to the Bloomberg Technology YouTube channel. Subscribe for original series and mini-docs on game-changing technologies, profiles on innovators, live conversations with experts, and breaking technology news & analysis from Bloomberg TV.