The acquisition of xConn will complete the core puzzle of memory pooling, Wells Fargo maintains MRVL.US (MRVL.US) “gain” rating

The Zhitong Finance App learned that Wells Fargo said that the transaction of Maywell Technology (MRVL.US) to acquire xConn for US$540 million is considered critical to memory pooling (memory pooling) and is expected to boost the company's earnings in the near future. The bank maintains its “overweight” rating with a target price of $135.

Analyst Aaron Rex wrote in a note to customers: “We believe this acquisition further demonstrates the importance of memory pooling technology in high performance/competitive hardware solutions, particularly its ability to support larger models, larger context windows, and improved inference performance (faster decoding speed). We would like to highlight Maywell Technology's comments about the company's ability to combine its CXL memory expansion controller with xConn's CXL switches.”

Rex said the proposed deal, which consists of 60% cash payments and 40% stock payments, will also help boost earnings for the 2027 fiscal year. The revenue from this business is expected to contribute to the company's overall revenue starting in the second half of this fiscal year, and may contribute as much as $100 million by fiscal year 2028.

The path to AI 2.0

In the AI 2.0 era, the core contradiction in the development of computing power has evolved from simply “not being able to calculate fast enough” to “not keeping up with data handling.” The advent of CXL (Compute Express Link) technology is essentially a “big operation” on traditional computing models on the underlying physical architecture. The logic for improving AI computing power can be improved in the three directions of memory decoupling, capacity expansion, and communication collaboration.

Rather than simply increasing bandwidth, CXL technology restructures an otherwise fragmented data center into a collaborative whole through resource pooling, capacity decoupling, and consistent communication. It is a key technical foundation for AI computing power to shift from a “single performance competition” to a “cluster performance game,” and it is also the only way to achieve large-scale, high-performance AI infrastructure in the future.