In the “gold rush” set off by AI models, Nvidia has always been a “shovel seller” favored by capital. So who's next at Nvidia?

Just as AMD and Intel were scrambling and eager to try it out, they held their hands in front of their chest$Broadcom (AVGO.US)$, an undetectable smile drifted through the corners of his mouth:

You are an AI shovel seller, I'll be your shovel seller!

Last Wednesday, Broadcom held an investor conference on the theme “Enabling Artificial Intelligence Infrastructure”. At the conference, Broadcom not only unveiled a variety of AI infrastructure innovations such as switches, PCIe, and copper connections, but also announced its third-largest customized chip customer to a crowd of analysts — this news sparked a capital frenzy on Wall Street.

As a result, Broadcom's stock price continued to soar. In the past week, Broadcom's stock price surged nearly 8%; in the past year, Broadcom's stock price surged 115.92%.

Overall, last Wednesday's Broadcom investment conference was full of information, and there was a lot of content worth digging into. Here are a few noteworthy points we've compiled for you:

You are an AI shovel seller, I'll be your shovel seller!

Training large AI models requires the use of a large number of processors such as GPUs, NPUs, and LPUs made by companies such as Nvidia and AMD — or XPUS, and connecting to these high-performance chips and transmitting data between them requires network chips such as Ethernet switches and routing — and this is an area where Broadcom excels.

The following PPT is easier to understand:

This is the “XPU” infrastructure necessary to support AI workloads. Broadcom's technology is shown on the right side of the image, and the corresponding numbers on the left reflect their position in the system package.

Next, we see a diagram of a configuration that scales up and out to many clusters. Through optical devices marked red and yellow, Broadcom's technology enables the connection of all parts of the system, from backbone networks and top of the rack switching to NiCS and PCIe innovations, covering a huge market far beyond GPUs and CPUs.

In this conference, all of Broadcom's PPTs revolved around one of the key ideas the company has always emphasized — openness, scalability, and energy efficiency. Charlie Kawwas, president of Broadcom Semiconductor Solutions Group, said:

For providers responding to the growing demand for generative AI clusters, the key to success will be a network-centric platform based on open solutions that scale with minimal power consumption.

Our innovations extend our leading position in custom AI accelerators, Ethernet, PCI Express, and optical interconnect product portfolios.

They are built on world-class foundational technologies such as SerDes and DSP to support optimal custom XPU and commercial networking solutions for AI infrastructure.

In addition, Kawwas also emphasized that Broadcom has established good partnerships with a wide range of customer groups, including many consumer internet companies, device manufacturers, and hyperscale computing companies, including Apple, Dell, and HP, including Nvidia, currently the most critical company in the AI field. Kawwas stated:

Broadcom has a good relationship with Nvidia, and Nvidia is its fastest growing customer.

Notably, Jas Tremlay, general manager of Broadcom's Data Center Solutions Group, also shared an interesting case about Nvidia Blackwell: although Blackwell's southbound connection uses proprietary NVLink, its northbound connection is an open PCIe. This shows that even Nvidia, which relies heavily on proprietary technology, had to consider the importance of openness when building its architecture. Tremblay said:

“Blackwell's southbound connection is proprietary NVLink, but its northbound connection uses PCIe.”

After the meeting, TD Cowen's analyst Matthew Ramsay upgraded Broadcom's stock rating to “outperform the market” and raised its target price from $1,400 to $1,500.

Admittedly, our ratings have been on the wrong side for a while, as the stock has outperformed the broader SOXX index over the past two years, with the latter rising by around 50%.

Despite this recent situation, we are confident in the sustainability and diversity of growth and the potential upside of consensus forward-looking predictions driven by AI computing/networking and software growth/profits.

Research firm Baird's Tristan Gerra said that Broadcom is a “world-class” competitor in AI networks, but the current valuation does not reflect this. He also gave Broadcom shares an “outperforming market” rating and set a target price of $1,500.

The Mysterious “Third Biggest Buyer”: ByteDance?

In particular, Broadcom announced its third-largest custom-chip customer, and the news attracted a great deal of attention. The management company did not clearly reveal who it was, but many analysts believe that this “mystery buyer” is probably ByteDance.

Citigroup's Christopher Danely analyst team believes that the third most mysterious customer is probably ByteDance. Broadcom has yet to respond to this. On Thursday, Citi reaffirmed its buying rating and target price of $1,560 for Broadcom.

Broadcom expects 35% of its chip revenue to come from AI by 2024, an increase of 10% over previous predictions, and far more than 5% in 2021:

By 2024, 35% of Broadcom's chip revenue will come from AI.

Undoubtedly, after last week's investor conference, Broadcom, which stepped on the AI cusp, seems to have become Wall Street's “new capital favorite.” In this context, excited investors and analysts seem to “not care” about the worrying aspects of Broadcom's earnings report, such as the poor revenue of traditional networking and storage businesses.

After all, there is a consensus on Wall Street this year: if you have AI, everything else, don't worry about it...

Editor/Somer