Who makes PCIe/CXL switches & retimers for AI data centers
Fabric silicon — switches that fan a single PCIe or CXL root out to many downstream devices, and retimers that clean up the signal over longer runs — connecting accelerators, SSDs, and pooled memory back to the host.
This part is the physical and link layer underneath the CXL memory expanders in 2.05, and it also carries general PCIe traffic to SSDs and accelerators elsewhere in a system. AI servers pack far more PCIe and CXL endpoints per node than general-purpose servers — accelerators, NICs, SSDs, and memory expanders all compete for lanes — and a single host cannot drive that many endpoints directly, which is what makes a switch necessary. Longer high-speed board traces at newer PCIe generations also need retiming just to close timing margin, so retimer content per system tends to rise alongside endpoint count and generation. For an investor, the supplier base spans both merchant interconnect chip makers with broad portfolios and specialists built specifically around switching and retiming silicon, and each new PCIe generation is a fresh design-win cycle rather than a carryover of the prior one.
AI delta: Packing many more PCIe and CXL endpoints per AI server than a general-purpose server requires drives up switch and retimer content per system.
Companies on this part (5)
| Company | Role | Ownership | Ticker | Price | 1d | Confidence | Note |
|---|---|---|---|---|---|---|---|
| Astera Labs | manufacturer | public | ALAB | 257 USD | +0.8% | inferred | Aries; Scorpio |
| Broadcom | manufacturer | public | AVGO | 322 USD | -0.3% | inferred | PLX |
| Marvell | manufacturer | public | MRVL | 229 USD | +3.7% | inferred | |
| Microchip | manufacturer | public | MCHP | 74.05 USD | -1.1% | inferred | |
| Xconn | manufacturer | private | inferred |
Who competes here
The same table read the other way: every company above competes for PCIe/CXL switches & retimers sockets. Confidence tags matter — an inferred placement is a lead, not a confirmed rivalry.
Nearby parts
- Previous in layer: CXL controllers & memory expanders
- Next in layer: NAND flash
- Layer: Layer 2 — Memory & storage
From the blog
Latest feed
- [AI Ecosystem] The semiconductor paradigm shifts toward memory — newsroom:sk-hynix, 2026-10-20The shift in AI workloads towards inference and agentic AI means that system performance is increasingly dependent on data storage, data movement speed, and processing efficiency, rather than just raw compute power.
- “Memory-Centric Server Systems Are the Future”: SK hynix Presents Its Next-Generation AI Infrastructure Vision at OCP Global Summit 2026 — newsroom:sk-hynix, 2026-10-16SK hynix showcased its vision for future AI infrastructure centered around memory-centric server systems at the OCP Global Summit 2026.
- “Memory-Centric Server Systems Are the Future”: SK hynix Presents Its Next-Generation AI Infrastructure Vision at OCP Global Summit 2026 — newsroom:sk-hynix, 2026-10-16SK hynix unveiled its future vision for AI infrastructure, emphasizing memory-centric server systems.
- “Memory-Centric Server Systems Are the Future”: SK hynix Presents Its Next-Generation AI Infrastructure Vision at OCP Global Summit 2026 — newsroom:sk-hynix, 2026-10-16SK hynix unveiled its vision for next-generation AI infrastructure, emphasizing memory-centric server systems.
- Samsung, AMD validate CXL memory for AI servers — DigiTimes, 2026-10-14Samsung and AMD have validated the use of CXL (Compute Express Link) memory for AI server applications.
- Marvell Unveils 2nd-Gen Structera X with Six-channel CXL 3.2 Memory Expanders for AI Inference — HPCwire, 2026-10-12Marvell has announced the second generation of its Structura X family, featuring six-channel CXL 3.2 memory expanders designed to support AI inference workloads.
- Marvell Structera X 3406 and 3506 CXL 3.2 Memory Expanders — ServeTheHome, 2026-10-12Marvell has launched its Structura X 3406 and 3506 memory expanders, which support CXL 3.2 for high-speed memory expansion and pooling.
- Marvell's custom AI chip upgrade points to a broader fight over peripherals — DigiTimes, 2026-10-08Marvell is highlighting the importance of custom silicon and high-speed interconnect peripherals in the broader AI infrastructure landscape.
- SK hynix Hosts ‘2026 Global Forum,’ Sharing Its Technology Vision and Future Strategy for the AI Era — newsroom:sk-hynix, 2026-09-22SK hynix presented its vision for the AI era at a global forum, detailing future strategies for its memory and storage components.
- SK hynix Hosts ‘2026 Global Forum,’ Sharing Its Technology Vision and Future Strategy for the AI Era — newsroom:sk-hynix, 2026-09-22SK hynix hosted a global forum to discuss its technology vision and future strategies relevant to the AI era.
Questions about PCIe/CXL switches & retimers
- What are PCIe/CXL switches & retimers?
- Fabric silicon — switches that fan a single PCIe or CXL root out to many downstream devices, and retimers that clean up the signal over longer runs — connecting accelerators, SSDs, and pooled memory back to the host.
- Which companies make PCIe/CXL switches & retimers for AI data centers?
- This catalog lists 5 companies on PCIe/CXL switches & retimers, 4 of them with listed equity: Astera Labs, Broadcom, Marvell, Microchip, Xconn.
- How does AI change demand for PCIe/CXL switches & retimers?
- Packing many more PCIe and CXL endpoints per AI server than a general-purpose server requires drives up switch and retimer content per system.
- How reliable are these PCIe/CXL switches & retimers supplier placements?
- The 5 placements on this part break down as 5 inferred. "disclosed" means the company or a filing says so; "reported" comes from trade press or a reference-design list; "inferred" is a taxonomy lead with no citation yet.
Sources: placements without a source URL come from the taxonomy seed and are tagged accordingly. How this data is built.· Markdown · JSON