Marvell has announced the Structera S 30260, a 260-lane CXL switch device designed to enable rack-level memory pooling for AI infrastructure. The company says the switch is meant to give data center operators access to disaggregated memory resources outside the server, targeting what it calls the AI “memory wall” as clusters grow in scale and complexity.
Marvell positions the Structera S 30260 as part of a broader CXL fabric approach used “in concert” with Marvell Structera A near-memory accelerators, Structera X memory-expansion controllers, and Alaska P PCIe/CXL retimers. The company’s stated objective is to increase memory bandwidth and capacity by allowing pooled memory to be dynamically expanded and allocated across CPUs, GPUs, XPUs, and other accelerators, without replacing existing platforms or relying only on HBM stacking.
On the interface side, Marvell says the Structera S switch supports CXL 3.0 and provides aggregate bandwidth up to 4 TB/s. In the PR, Marvell also points to PCIe/CXL copper and optical cable solutions from ecosystem partners, describing them as a way to scale connections among GPUs, CPUs, XPUs, CXL memory, SSDs, and other PCIe components for low-power, high-speed, low-latency connectivity within servers and across clusters.
For engineers building large inference clusters, the key performance claim centers on latency and locality. “The CXL switch offers an ideal solution by providing a near‑local, shared memory pool with sub‑microsecond access, which eliminates multi-hop data movement and unlocks higher throughput, longer context and improved GPU utilization,” said Gerry Fan, senior vice president, engineering, Scale-up Switching at Marvell.
Marvell also frames the product as an architectural shift for AI platforms. “Breaking through the AI memory wall requires a fundamental architectural change,” said Rishi Chugh, vice president and general manager, Data Center Switch Business Unit at Marvell. “The Structera S CXL switch is the first true CXL switching solution purpose-built for AI.”
The PR ties the Structera S line to Marvell’s recent acquisition of XConn Technologies, saying it strengthens the company’s “industry-leading switch solutions” and contributes to an end-to-end CXL fabric architecture spanning expansion, acceleration, and pooling for both traditional compute and “next-generation scale-up AI environments.”
Availability-wise, Marvell says the Structera S 30260 is expected to begin sampling to customers in calendar Q3 2026. The company also notes that the Structera S 20256 CXL 2.0 switch is currently in production.
Source: Marvell
