As Nvidia approaches a fresh record high, Wedbush Managing Director Matt Bryson breaks down what it would take for the chipmaker to reach a $6 trillion market cap, why tight memory supply may actually help Nvidia, and why he doesn't think there will be enough memory next year. Bryson also weighs in on what Nvidia-backed Reflection's open-weight model could mean for AI demand, reports that Elon Musk could partner with Taiwan Semiconductor and what that means for Intel, and why reports of Microsoft and Meta pulling back on Anthropic's Claude may matter more for Anthropic than for Nvidia.
Market Domination, Market Domination Overtime and Asking for a Trend are live weekdays 3-5pm ET with Yahoo Finance Host Josh Lipton.
00:00 Can Nvidia hit a $6 trillion market cap? 00:18 Nvidia's growth vs. its valuation 00:44 Do memory costs weigh on Nvidia? 01:26 How long can Nvidia's runway last? 02:04 Memory pricing next year: Micron and SanDisk 02:59 What Reflection's open-weight model means for Nvidia 04:34 Elon Musk, TSMC and what it means for Intel 06:33 Microsoft and Meta reportedly pull back on Claude 07:58 Nvidia at the center of AI, whichever model wins
AI Generated·May contain errors
Video Summary
AI Generated
The discussion focuses on NVIDIA's market position, its ability to manage supply chain constraints, and the broader AI ecosystem. The conversation covers the impact of open-weight models on inference demand, memory pricing trends, and the competitive landscape involving TSMC and Intel.
Matt, thanks for joining us because we're waiting on that fresh record high, potentially from NVIDIA and some investors even saying that there's a 50 percent chance that NVIDIA could reach a six trillion dollar market cap by the end of this month.
What do we need to see in order to get there?
I mean, six trillion, I don't know.
I think the story with NVIDIA, though, has been that it's constantly been beating expectations, but the stock's been underperforming.
And in my mind, it doesn't make a whole lot of sense that a company that's going to grow in the triple digits this year, that's guided to 70 percent growth next year, is trading at a multiple that's under 20 times.
How much is higher memory costs still weigh on this company when you think about that outlook for revenue growth?
I mean, I think as much as higher memory costs are a problem, it's just availability of memory.
And I think one of the things NVIDIA has done and you've seen in their balance sheet is they've gone out and done the best job in terms of securing supply, not just for memory, but things like substrates and copper-clad laminate and logic.
And so in a way that allows them or sets them up to grow at least as fast as the market, if not faster than the market in an area of constraint or in a time of constraints.
And so I think they almost benefit from the tight supply.
Do you think that this benefit will take them through fiscal year 2028?
Because it seems like that's as far as we're seeing these targets kind of go to.
I think that's exactly where it gets them to.
And then you get to calendar year 28, fiscal year 29, and it's always been a bit of a black box what two years out looks like.
Having said that in the near term, again, everything looks great.
Foxconn gave us great numbers.
When you talk to Neo clouds, their payback times are improving.
So that means at the margin, you're seeing what people are willing to pay for GPUs go up, which means there's more demand for AI.
Like these are all things that are good for Nvidia.
When you think about those memory companies in particular, I think Micron, SanDisk.
I mean, what exactly are we seeing in terms of pricing heading into next year?
Because it seems like they kind of have control over what exactly it will look like.
But then where do we go from there?
Yeah.
So I think when you're thinking about next year, it's always hard to project what happens in memory pricing beyond the next quarter.
We know that pricing's up double digits again next quarter for both DRAM and NAND, or it looks to be up.
I think contract will solidify this month into next year.
My best guess is there's still going to be short supply, so pricing won't go down.
It's more a question of how much higher can it go?
Because we're at a point that's unprecedented in history.
These guys have close to 90% gross margins.
And so how much higher can it push?
I don't know, but I don't think that there's enough memory next year.
I want to ask you something in particular, if you can sort of play ball with me here.
We're hearing reports that a new model from an Nvidia-backed startup called Reflection is coming out.
They're sort of leaning into this idea of an AI factory concept.
This open weight model compared to these closed models like Anthropic, as well as OpenAI.
So what would that mean with Nvidia's backing?
What would that mean for the company?
Can you walk investors through that?
Yeah, so from an Nvidia perspective, you've got puts and takes.
So if you look at OpenAI and Anthropic, those kind of models have been driving the industry,
in particular, driving training demand.
And so that's good for Nvidia.
And so if open weight models take off, to some extent, maybe you lose some of that training
demand.
Having said that, we've seen a big shift towards inference.
So use cases for AI instead of training models.
If open weight models are successful, all of a sudden inference becomes cheaper.
And I think what's been driving that NeoCloud demand, what's driving Amazon or allowing
Amazon to raise prices for its GPU instances, that's all this demand.
And if all of a sudden you make it cheaper to use inference resources, I think that's
a great thing for Nvidia.
It's a great thing for the AI complex.
And it benefits everyone, because if it's cheaper to buy a resource, then you end up
using more of it.
I want to get your take on something else that we're watching today, and that's Intel,
as well as Taiwan Semiconductor.
Now, we heard reports that Elon Musk potentially is looking to partner with Taiwan Semiconductor
for that TerraFab factory.
What's the impact there?
And what's the likelihood that we could end up seeing this partnership that, of course,
would be a hit a bit to rival Intel?
Yeah, it's hard for me to determine what the probability is that TSM would work with SpaceX.
So historically, they haven't been willing to share technology, if you will.
Now, certainly, they've done partnerships before, particularly with Japanese suppliers,
where they will build the fab and they will produce certain parts for those customers
or dedicate capacity for those customers.
So that's certainly something that we could see.
But it seemed like Elon Musk wanted a closer partnership where he had more say over what
was happening in the facilities.
I don't know that that's something TSM would do.
In terms of what it tells you about Intel, look, the Intel trades on an NVIDIA sales
basis at a relatively similar rate to NVIDIA, to Taiwan Semiconductor, to AMD.
Yet their margins aren't as good, their growth isn't as strong.
I mean, what that tells me is that the market's already anticipating that Intel's caught up.
And we don't have a whole lot to tell us where Intel is.
This Elon Musk partnership was one thing that suggested that Intel was catching up on the fab
side.
So I guess if that goes away, it's what would appear to be a positive sign for Intel that
isn't there.
Having said that, I don't have a strong stance on where Intel is.
I'm still waiting to see if they've got things figured out or if there's still a whole lot
that they need to do.
Yeah, Matt, I should point out here that Intel stock year to date up about 195 percent.
We also Taiwan Semiconductor up about 42 percent.
Matt, while I have you, I want to get your take on another report that we're seeing out
there.
There's a report that said that Microsoft and Meta are looking to wean employees from
Claude.
Of course, that's Anthropx enterprise model there.
For Microsoft, they're saying that potentially it could be up to by a third.
So are there downstream effects for a company like NVIDIA if these companies pull back a
bit from having their employees use Claude?
So I've seen that article.
I haven't fully read it yet.
What I'd say is it depends what else they're using.
Right.
So if that Microsoft employee is using an open way model, if they're using Microsoft
resources versus Anthropx resources, they still are using compute.
They're still using accelerators.
So they're still they're still using NVIDIA hardware.
NVIDIA or NVIDIA matters less what the model is and rather is is AI gaining traction?
And I think from that standpoint, it doesn't matter so much for NVIDIA.
From an Anthropx perspective, it matters more.
Right.
And then it all comes down to can Anthropx differentiate so much that Microsoft and Meta
can't make this choice because they're just giving you a better model.
So NVIDIA is still kind of at the heart of all these different AI models, no matter what.
I think that's exactly it.
And I think that's why the stock's doing what's doing today.
Matt Bryson, thank you so much for joining us.
I appreciate it.