Source: Caixin News

"Sell NVIDIA, Buy Broadcom!" In the past few days, $Broadcom (AVGO.US)$ with $NVIDIA (NVDA.US)$ the performance can be described as "two extremes".

Mizuho Analyst Jordan Klein stated that Wall Street is paying attention to the demand for ASICs from large Technology companies, which could be one of the reasons for the decline in NVIDIA Stocks; some investors may choose to Sell NVIDIA Stocks in the short term to strengthen their Hold Positions in Broadcom.

In fact, if the timeline is extended to nearly a month, it is clear that in stark contrast to the fluctuating and somewhat bleak stock price trend of NVIDIA, a new batch of AI Infrastructure individual stocks in the USA is rising:

Broadcom, which has been gaining momentum recently, has increased by over 50% in the past month, while its "peers". $Marvell Technology (MRVL.US)$ Increased by over 42%. In addition, the high-speed connectivity solutions provider. $Credo Technology (CRDO.US)$ Increased by over 90%, while "Little NVIDIA". $Astera Labs (ALAB.US)$ The increase has also exceeded 50%, and the Data Storage company. $Pure Storage (PSTG.US)$ Fiber optic communications equipment manufacturers $Ciena (CIEN.US)$ Increased by over 35%, switch suppliers $Arista Networks (ANET.US)$ Increased by 24.5%……

Moreover, most of these companies reached their highest stock prices in recent years on the latest trading day in the USA.

Both Broadcom and Marvell primarily focus on ASICs and have both delivered outstanding Earnings Reports.

Credo is a provider of high-speed connectivity and high-speed serial link communication solutions, with its core product being AEC (Active Cable). In the second quarter Earnings Reports for fiscal year 2025 released at the beginning of the month, its revenue reached a new high at $72.034 million, a 63.6% year-on-year increase and a 20.6% quarter-on-quarter increase. The company indicated, 'The revenue inflection point has begun.' The revenue driver mainly comes from the downstream increase of AEC, and the vast majority of AEC shipments are based on AI applications, 'It is expected that all American hyperscale customers will use AEC in some form in the long term.'

Astera Labs primarily provides semiconductor-based connectivity solutions for cloud and AI infrastructure. The company's third quarter earnings report released in early November shows that Q3 revenue reached $0.113 billion, a year-on-year increase of 206% and a quarter-on-quarter increase of 47%, setting a new quarterly record; it is expected that Q4 revenue will continue to set new highs, with growth driven mainly by the PCIe Retimer and AEC product lines. The management stated that the visibility of performance is very strong, whether from the perspective of backlogged orders or design-in.

Pure Storage mainly provides enterprise data storage platform. The company reported third-quarter results that exceeded Wall Street expectations earlier this month and provided strong guidance for the fourth quarter, upgrading the full-year performance forecast. At the same time, the company announced a contract with one of the "Big Four" AI massive suppliers, with CEO Charles Giancarlo mentioning that it is the first time large-scale enterprises have provided customer-facing standard storage through system suppliers, "The solutions we provide can replace 90% of their storage."

Ciena mainly focuses on optical communications. The company announced its latest earnings report a few days ago. Although profits have declined, unexpectedly strong order flows have driven the company's stock price to new highs. Ciena revealed that as of the end of the 2024 fiscal year, its backlog reached $2.1 billion, with orders in this quarter starting "very strongly." The company sees opportunities in AI datacenters, primarily in the metro datacenter campus network sector.

Arista mainly supplies switches, routers, and other networking devices used in datacenters. The company recently stated that it expects AI-related revenue to reach $0.75 billion in fiscal year 2025, and Arista is upgrading from its current 800G products to 1.6Tb and even 3.2Tb technologies to support larger-scale AI cluster deployments.

▌What changes are happening in AI Infrastructure trends?

These companies cover aspects including chips, high-speed interconnects, storage, optical communication, switches, and so on—in short, it is still about computing power, transportation capacity, and storage capacity. From their respective earnings reports and statements, combined with analyses from various institutions and brokerages, we may not find it difficult to glimpse the latest trends in AI infrastructure investment:

GTJA's research report today pointed out that Broadcom's conference call indicates that not only will cloud vendors invest more in self-developed AI ASIC chips over the next three years, but there is also potential for constructing over one million XPU clusters, which will drive the sustained prosperity of upstream industries such as optical modules, switches, PCBs, and high-speed cables.

In terms of DCI (Datacenter Interconnect), Meta, Google, Microsoft, and OpenAI have already begun multi-DC distributed training. Ciena has also stated that the major opportunities for AI datacenters lie in the metro datacenter campus network area. Guolian Securities believes that the demand for network due to AI computing power deployments is spreading towards DCI scenarios, which is expected to drive rapid growth in the DCI market.

In terms of AEC, both Credo and Astera Lab have mentioned strong downstream demand. Previously, Amazon AWS released the latest Trainium2 and Ultra versions of their cabinet solutions, which utilized AEC. Brokerages indicated that the trend of interconnecting Neuronlinkv3 through AEC has become increasingly common.

It is worth mentioning that a recent report from Bloomberg Intelligence pointed out that enterprise clients may engage in larger-scale AI investments by 2025, with AI spending growth focusing more on the inference side to realize investment monetization or enhance productivity.

Want to enter the market but timing is difficult. "Monthly Payment Zone"The feature can help you, open Futubull > Market > US Stocks > Click on 'Monthly Payment Zone' > 'Create Monthly Payment' > Set your monthly payment plan to invest regularly and seize investment opportunities!

编辑/jayden