Recently, the market's attention to AI infrastructure has focused on the network communication end. The impressive Q2 performance has driven the leading communication chip, whose stock price has reached a new all-time high and the latest market cap is as high as 778.1 billion U.S. dollars. In the same industry, Arista also achieved a new high on the same day, with a market cap exceeding 100 billion US dollars.$Broadcom (AVGO.US)$reached a new all-time high with a market cap of 778.1 billion US dollars due to the impressive Q2 performance. The market's attention to AI infrastructure has focused on the network communication end. $Arista Networks (ANET.US)$Arista also achieved a new high on the same day with a market cap exceeding 100 billion US dollars.

Both companies are excellent network infrastructure providers committed to Ethernet solutions. Broadcom's main products include Ethernet switching chips, packet processors, ASCI, etc. Arista is a leader in high-end Ethernet switches, and its switches are mainly used in super large-scale data centers such as Microsoft and Facebook.

Following a similar trend as Nvidia, the share prices of Broadcom and Arista began to rise rapidly around 2022 to 2023, coinciding with the release and popularity of ChatGPT.

After the stock price reaches new highs, will it continue to rise? Wall Street analysts are very bullish on this track.

Bank of America stated that Broadcom, like Nvidia, is the best choice in AI field. After the impressive performance report, Broadcom still has greater potential. Broadcom is very likely to become the next member of the trillion-dollar club in the US stock market, and Ethernet is seen as an important driving force for its growth.

What is the relationship between Ethernet and AI?

Switches as the representative of networking equipment are an important part of hardware for AI computing power. At present, Nvidia's IB (Infini Band) plan dominates in the AI connection field, and Ethernet plan has its own advantages.

IB has better performance and has the advantages of high bandwidth, low latency, high reliability, and scalability. However, it has a higher premium, and users often need to pay more invisible networking costs. Under the same budget, the number of purchased graphics cards, optical modules, and switch hardware will be less.

Ethernet, as the most widely used communication protocol in the past few decades, has a large audience, a high acceptance rate among users, adopts universal network equipment, has lower cost and strong compatibility. However, because it was used less in high-performance AI computing scenarios in the past, and the suppliers were scattered, the overall effect of Ethernet was somewhat inferior to the closed Nvidia solution in the early stage of AI development.

The cost advantages of Ethernet are highlighted. The switch/switch chip and optical module industries are expected to benefit.

The high cost of computing power infrastructure is a bottleneck for the development of AI. With the development process of AIGC accelerating, the demand for cluster capabilities and computing power is becoming more and more huge, and the cost advantage of Ethernet is gradually becoming apparent. According to the research report of Guosheng Securities, in Meta's test, the Ethernet networking price is less than half of IB, and the performance is 10% higher, which proves that Ethernet is "buy more, save more."

Nvidia has already chosen to "do both", and at the Computex conference on June 2nd, Jensen Huang officially showed Nvidia's layout in the Ethernet switching system. Guosheng Securities believes that Nvidia's entry into Ethernet is a reaffirmation of the "vast ocean" of AI network side demand, and it will also drive the Ethernet into a new speed update cycle.

In July 2023, the Ultra Ethernet Consortium (UEC) was established, with members including AMD, Arista, Broadcom, Cisco, Meta, Microsoft, etc., aiming to solve many shortcomings in the actual application process of Ethernet.

Zhongtai Securities stated that with the establishment of UEC, several issues of Ethernet running AI workloads may be effectively resolved, but it does not believe that IB and Ethernet are completely interchangeable or that Ethernet will replace IB networks. For a considerable period of time, due to the high iterative efficiency of Nvidia's GPU products, IB networks will continue to occupy a higher market share. Ethernet will benefit more from market, cost, commercial competition and other factors in the short term. However, the final results of the UEC and Ethernet's products/technologies need to be verified. But regardless of which protocol, the underlying chips and switch products of switches will benefit.

Guosheng Securities also stated that Nvidia's bet on Ethernet is not a lack of confidence in its own IB, but an early grab for a broader AI reasoning market. Whichever solution (IB VS Ethernet) dominates, it will promote the expansion of the switch market space and accelerate the overall development of the industry. It is recommended to pay attention to the switch industry, which is expected to benefit from the gradual penetration rate of Ethernet solutions in AI application scenarios from chips, manufacturers, and brands.

In addition, the institution stated that one of the biggest advantages of the AI Ethernet solution of Nvidia/UEC over the IB solution is low cost. In order to continue implementing the characteristics of low cost and low power consumption, the institution also tends to choose the low-cost LPO (linear-driven pluggable optical module) scheme for the selection of optical modules. Therefore, the penetration rate of LPO is expected to accelerate in the future, but since this scheme needs to cooperate with switches, it requires higher cooperation and synergy among optical module manufacturers in the upstream and downstream of the industry. Leading companies such as Zhongji Innolight and Eoptolink Technology Inc. will have more advantages.

Gtja believes that we should pay attention to Nvidia and the investment opportunities of the whole industry in the trend of Ethernet solution networking and the investment opportunities of optical module suppliers and switches.

Editor/new