NVDA是头号持仓/首选，估值合理，业务正从集中的数据中心销售转向更广泛的AI推理/边缘市场（通过Hugging Face的开发者渠道），支撑其增长延续至2026年。

It's not me and Kasey today, it's me and the dog talking about Nvidia.

What does that have to do with NVIDIA and its proposed acquisition of Hugging Face? As you probably have heard by now, NVIDIA said it plans on acquiring the company Hugging Face for 12.9 billion and some change.

Not sure what to make of that other than hopefully NVIDIA was planning on paying a similar price anyways, and so they just came up with a number that can just be a rounding error for software developer.

Anyways, NVIDIA, as you know, yes, it's a GPU company, it's an accelerated computing company, but they really control the entire supply chain behind the equipment that they develop.

They do co-design with their customers that do the other ancillary parts and pieces for the data centers before they sell that to hyperscalers, to enterprise customers, to NeoClouds.

NVIDIA said that there are more than 18 million developers, half a million datasets, a million applications, 200,000 companies that use the Hugging Face platform.

And so as I showed you on the last supply chain chart, what would Hugging Face bring to the mix for NVIDIA? It makes the hardware, it has the software that makes that hardware run and operate.

They've also been getting into developing AI models themselves. In fact, NVIDIA is the largest contributor.

They have the most number of repos on Hugging Face. They have a chart shared by Hugging Face illustrating this wide margin there between them and Alibaba Cloud, and Hugging Face itself in second and third place.

So what does NVIDIA really get here? They get access to developers, which is to say Hugging Face would be a distribution platform for NVIDIA hardware and an increasing number of NVIDIA services that run on that hardware.

Now, the idea is NVIDIA wants to keep Hugging Face open source so that developers can use whatever system they want when they're developing their AI and their machine learning.

But nevertheless, they would get that access to those developers, and that's probably going to introduce some serious regulatory risk for this deal. And I think that's pretty well illustrated in just taking a look at the acquisitions NVIDIA has done over the last seven years or so.

Of course, Mellanox, the famous one, probably go down in the history books as one of the greatest acquisitions ever, finished in April 2020. As small as that one was, $7 billion, at the time, pre-Open Ai moment, NVIDIA moment a couple of years later, that actually still went through quite a bit of scrutiny, specifically from regulators in China that took a really hard look at Mellanox and wanted to make sure that they were not going to get cut off from those networking chips.

Which we know today, Mellanox networking, especially InfiniBand, as well as the Ethernet networking capabilities they brought to the table, is an integral part of the Nvidia ecosystem.

It's quite literally what helped them do what they were able to do a couple of years later, networking these massive groups of GPUs together. So let's call that more of a vertical acquisition that Nvidia made.

The one that got the most scrutiny and led to Nvidia just completely dropping the acquisition altogether was Arm Holdings. That was dropped in 2022. It was proposed at $40 billion.

They were gonna acquire that from SoftBank. That was more of a horizontal acquisition, so adding CPU IP and design IP into the mix.

Regulators took a really hard look at this one, had too many requirements for Nvidia, and ultimately they just decided to pass. Run:ai was very small.

That one actually got cleared after going through quite a few hoops as well, only at $700 million.

I'm leaving out a couple smaller acquisitions, but since then, especially middle of last year with Enfabrica, that was actually an acquihire. So NVIDIA didn't actually acquire Enfabrica or Groq in December 2025, or most recently, Poolside, just last month in August 2026.

What they did was essentially licensed the technology and started moving over large parts of the technical team in each of these cases, especially the founders at Enfabrica and Groq also moved over with the licensing deal, thus the name acquihire.

This has come under some criticism as well as NVIDIA skirting that regulatory process that would normally take place if it's an outright acquisition. And so it's interesting to see this one now of Hugging Face for $12.9 billion, another large acquisition.

It would technically be the largest acquisition ever for NVIDIA if you exclude the Groq acquihire. We'll see how this works out, but there's more than a small likelihood this one gets shot down.

A lot of Hugging Face's team is in France, so there's going to be a lot of EU regulation besides regulators in China and the US and elsewhere taking a hard look at this because they want to make sure that AI does truly stay open source.

Now, this is where the market gets really fuzzy because this battle between NVIDIA and the open AI model ecosystem and the closed model ecosystem from the AI labs kind of gets siloed by a lot of investors. This is one or the other.

And in all reality, the lines aren't so clear-cut between the two. The reason why this kind of gets compared as a this or that discussion among investors is because the AI models have massive funding.

All three now have massive amounts of revenue and are doing a lot of custom chip design or ASICs, application-specific integrated circuits. Maybe they do that themselves, or they're using an ASIC, for example, Anthropic's Claude using a lot of Amazon's custom silicon. So it's not really an this or that type of scenario.

NVIDIA does a lot of custom chip design for its customers as well, custom piecing together of networking systems. But, this is the argument, though, that NVIDIA is going to make.

We'll see what kind of regulatory scrutiny it gets and what kind of things NVIDIA will have to do to appease regulators and convince them that Hugging Face will indeed be left open and NVIDIA doesn't use it to somehow allocate more of its systems, more of its ecosystem towards the future development of AI, be that in the data center or physical AI operating actually out in the field.

So that's the comparison that's being made, and that'll be a big source of debate as NVIDIA tries to make the case over the course of the next year to try to get the Hugging Face deal done with regulators.

Either way, whether this goes through or not, Nvidia is most certainly at some sort of inflection point here as the initial wave of growth from selling its systems to the hyperscalers and the early, Neocloud adopters of GPUs has worn off.

We had that bit of a slump in revenue growth as they started to scale up Blackwell last year. But now Blackwell in production, Vera Rubin is ramping up. Revenue has re-accelerated.

And so what is Nvidia doing here? Are they trying to stuff the sales channel and distribution so that they can continue selling to the hyperscalers and the Neoclouds?

We're of the opinion that's not really the case. What they're looking at is AI inference, and not just AI inference in the data center, but once AI begins to proliferate throughout enterprises, including actually in the real world, where AI can be applied more broadly across the economy, that is really what they are probably looking at, making sure they have a funnel of lots of developers that they can at least get in front of and help develop AI.

The claim is they can use Nvidia or they can use some other system, but Hugging Face would essentially act as this future funnel of AI development well beyond the centralized compute in data centers.

And that makes sense. NVIDIA also shuffled up its revenue segmentation just this year. This KPI chart only goes back to the quarter that finished up in January 2025.

But in orange and purple, hyperscalers and the AI clouds and other enterprises make up the vast bulk of NVIDIA's revenue. Edge computing, which includes PCs, gaming, is still a very small portion of the overall revenue segmentation.

Automotive would also fall into here. And so if NVIDIA wants to continue to diversify its business and make sure it has a very, very long runway of growth, not just this five, seven years period of hypergrowth driven by AI data centers, but many decades of growth, sort of like Intel was able to enjoy as it diversified its product offerings over the years.

NVIDIA would really want to foster AI beyond just these relatively concentrated customers that are depicted here in orange and purple, and foster more customers that would fall into the green revenue segment.

And just another quick comparison to another company, Nvidia and Broadcom with their trailing 12-month revenue. So lots of discussion there about NVIDIA being able to hold onto its market share leadership.

But if you compare the companies side by side, yes, Broadcom has lots of software revenue in the mix as well that dilutes the growth. But as it stands at the moment, NVIDIA doing more than its fair share holding on to its revenue growth.

As the AI labs design more custom silicon , fostering more open AI would, in our opinion, probably help this trend continue for many years to come. All of that is to say, whether Hugging Face goes through or not, we actually do like Nvidia, and we like Broadcom as well, as well as the other hyperscaler stocks, Microsoft and the others.

Nvidia and Broadcom, this is probably not, you know, one wins and the other doesn't.

They're going to be various shades of both companies picking up lots of sales and lots of market share as things progress. Valuation looks agreeable, especially on Nvidia. This has remained our top position and our top pick throughout 2026, and we see no reason to change that now as the stock has ran back up close to or briefly touched all-time highs again.

And it would appear based on the trends, that could continue through the rest of 2026.