The ‘most undervalued’ company of Nvidia starts as a ‘rocket ship’

When NVIDIA (NVDA) reports his second quarter of income on 27 August, investors will focus directly on the results of the company. After all, that is where the chip giant realizes income when selling its powerful AI processors.

But the data engine segment contains more than just chip sales. It is also responsible for some of the most important, although often overlooked, offers from Nvidia: the network technologies.

Composed from his NVLink, Infini Band and Ethernet solutions, NVIDIA’s network products are possible that his chips can communicate with each other, can talk servers in solid data centers and ensure that end users can connect to everything to perform AI applications to perform AI -applications.

“The most important part when building a supercomputer is the infrastructure. The most important part is how you connect those calculation motors to form that larger unity of computing,” Gilad Shainer, Senior Vice President, explained Nvidia.

Paris, France - 11 June: Co -founder and Chief Executive Officer of Nvidia Corp., Jensen Huang attends the 9th edition of the Vivatech Trade Show at the Parc des La Porte de Versailles on 11 June 2025, in Paris. Vivatech, the largest technical exhibition in Europe, offers a unique digital format for four days of re -connection and recovery through innovation. The event brings startups, CEOs, investors, technology leaders and all digital transformation players together that shape the future of the internet. Founded in 2016 by Publicis Groupe and Groupe Les Echos, this annual technology conference, also known as Vivatech, is devoted to promoting innovation and startups .. (photo by Chesnot/Getty Images) — Nvidia CEO Jensen Huang attends the 9th edition of the Vivatech Trade Show at the Parc des Expositions de la Porte de Versailles on 11 June 2025, in Paris. (Photo by Chesnot/Getty images) · Chesnot Via Getty Images

That also translates into a large sale. Nvidia’s network sales accounted for $ 12.9 billion of his $ 115.1 billion in data center income in his earlier financial year. That may not seem impressive when you consider that chip sales yielded $ 102.1 billion, but it has overshadowed the $ 11.3 billion that it has dealt with the second largest segment of Nvidia, Gaming, for the year.

In the first quarter, networks $ 4.9 billion in the income of $ 39.1 billion from NVIDIA $ 39.1 billion. And it will continue to grow while customers continue to build up their AI capacity, whether it is for research universities or mass data centers.

“It is the most undervalued part of the Nvidia company, by order of size,” Deepwater Asset Management Managing Partner Gene Munster told Yahoo Finance. “In short, networking does not get attention because it is 11% of sales. But it grows like a rocket ship.”

When it comes to the AI explosion, says Nvidia Senior Vice President of NetWorking Kevin Deierling that the company must work in three different types of networks. The first is the NVLink technology, which connects GPUs within a server or several servers in a long, cupboard-like server rack, allowing them to communicate and stimulate overall performance.

Then there is Infini Band, which connects several server buns on data centers to form what is essentially a huge AI computer. Then there is the front-end network for storage and system management, which uses Ethernet connectivity.

NVIDIA CEO Jensen Huang presenteert een Grace Blackwell NVLink72 terwijl hij een keynote -adres geeft op de Consumer Electronics Show (CES) in Las Vegas, Nevada op 6 januari 2025. Gadgets, robots en voertuigen die zijn ingezaaid met kunstmatige vie op zoek zijn naar de consumenten die zijn op zoek naar de consumenten die zijn op zoek naar de consumenten die zijn op zoek naar de Tiniffs looking for tiniffs behind the tiniffen looking for tiniffs threatened by consumers to threaten consumers to show consumers, the attention for consumers behind the tiniffs. US President-Elect Donald Trump. The annual Consumer Electronics Show (CES) formally opens in Las Vegas on January 7, 2025, but previous days are packed with product announcements. (Photo by Patrick T. Fallon / AFP) (Photo by Patrick T. Fallon / AFP via Getty Images) — Nvidia CEO Jensen Huang presents a Grace Blackwell Nvlink72 while he delivers a keynote address at the Consumer Electronics Show (CES) in Las Vegas, Nevada on January 6, 2025. (Photo by Patrick T. Fallon/AFP via Getty images) · Patrick T. Fallon via Getty Images

“Those three networks are all needed to build a gigantic AI scale, or even a moderate business scale, AI computer,” Deierling explained.

However, the purpose of all these different connections is not only to help communicate chips and servers. They are also meant to let them do this as quickly as possible. If you try to perform a series of servers as a single computer unit, they have to talk to each other in an instant.

A lack of data that goes to GPUs slows down the entire operation, slows down other processes and influences the overall efficiency of a full data center.

‘[Nvidia is a] Very different company without networks, “Munster explained.” The output that the people who buy all Nvidia chips [are] Desire would not happen if it was not for their networks. ‘

And while companies continue to develop larger AI models and autonomous and semi-autonomous agent AI possibilities that can perform tasks for users, it is becoming increasingly important with each other in Lockstep.

This is particularly the case, because leading AI models – requires more powerful data center systems.

The AI industry is located in the midst of a broad rehearsing around the idea of inference. At the start of the AI explosion, it was thinking that training AI models would require extremely powerful AI computers and that the actually running of them would be a little less power-intensive.

Earlier this year, this led to some fear on Wall Street, when Deepseek claimed that it trained its AI models on the upper Nvidia chips under the top-of-line-line. The thinking was at the time that if companies could train their AI models and run their AI models on under-covered chips, no expensive powerful systems from Nvidia were needed.

But that story turned quickly when chip companies pointed that those same AI models benefit from performing powerful AI computers, so that they can reason more information faster than they would do while they are performed on less advanced systems.

“I think there is still a misconception that inference is trivial and easy,” Deierling said.

“It turns out that it is starting to look more and more on training when we come [an] Agentic workflow. So all these networks are important. When they have them together, tightly linked to the CPU, the GPU and the DPU [data processing unit]That is all of vital importance to make education a good experience. “

However, the rivals of Nvidia are circling. AMD wants to take more market share of the company and cloud giants such as Amazon, Google and Microsoft continue to develop their own AI chips.

Industrial groups also have their own competitive network technologies, including Ualink, which is intended to hold it against each other with NVLink, Forrester analyst Alvin Nguyen explained.

But for now, Nvidia continues to lead the peloton. And while technical giants, researchers and companies continue to fight for Nvidia’s chips, the network company of the company is almost guaranteed to continue to grow.