Compute Express Link (CXL) 메모리 시장 규모, 점유율,

Dublin, Nov. 06, 2026 (GLOBE NEWSWIRE) -- "Compute Express Link (CXL) Memory Market Size, Share, and Industry Analysis - Global Forecast to 2036" has been added to ResearchAndMarkets.com's offering.

The global CXL memory market was valued at USD 0.85 billion in 2025 and is estimated to reach USD 1.53 billion in 2026. Revenue is projected to rise to USD 16.70 billion by 2036, representing a CAGR of 27.0% during the 2026-2036 forecast period. The report assesses the technology landscape, CXL specification development, processor and platform support, memory supply and pricing, use cases, total cost of ownership, competitive positioning, and long-term growth opportunities.

The market scope includes CXL memory expansion modules based on DDR5 or DDR4 DRAM, EDSFF E3.S devices, add-in cards, memory controllers, switches & pooling systems, and software for tiering, pooling, and memory management. Standard DIMMs, high-bandwidth memory (HBM), and interconnect technologies that do not use CXL are excluded.

Market Development and Growth Drivers

CXL memory entered commercial deployment during 2025. SK hynix completed customer validation of a 96-gigabyte CXL 2.0 module based on DDR5, reporting 50% more server capacity and 30% more bandwidth. The CXL Consortium subsequently published CXL 4.0, using PCIe 7.0 to provide bandwidth of 128 gigatransfers per second. Microsoft Azure also opened a preview of CXL-attached memory using Astera Labs' Leo controllers, marking the first public cloud CXL memory offering.

Market expansion is being driven by AI inference, large language models, recommendation engines, analytics, in-memory databases, and other data-heavy workloads. Rising processor core counts and increasingly large application memory footprints are intensifying demand for additional server capacity and bandwidth. Cloud providers are also evaluating CXL memory pooling to reduce stranded memory across data center fleets.

Support from current Intel and AMD server processors, combined with continued progress from CXL 2.0 through CXL 3.x and CXL 4.0, is strengthening the multi-vendor ecosystem. Adoption is also becoming more urgent as DRAM supply tightens. Samsung reported price increases of up to 60% for selected memory chips in December 2025, while data centers are expected to consume as much as 70% of global memory output in 2026.

Key restraints include additional latency compared with processor-attached DRAM, limited DRAM availability, rising prices, and the prioritization of HBM production for AI accelerators. Wider adoption also depends on improved operating system, hypervisor, application, hot-plug, tiering, and pooling support. Interoperability across devices and specification versions, data placement across memory tiers, reliability, security, and measurable return on investment remain important deployment considerations.

High-Growth Market Opportunities

Memory pooling for AI inference represents a major opportunity as key-value caches expand with context length and concurrent user demand. At the 2025 OCP Global Summit, Astera Labs demonstrated three times as many concurrent model instances using CXL memory, together with higher throughput. Additional opportunities include reusing DDR4 modules during server refreshes, deploying rack-level memory pools, extending AI inference cache capacity, and introducing NAND-backed & hybrid devices.

Cloud and sovereign operators developing new AI infrastructure can incorporate CXL into facilities from the planning stage. The availability of CXL memory through public cloud services is also expected to broaden access and accelerate enterprise testing, platform validation, and commercial adoption.

Segment Outlook

CXL memory expansion modules are expected to account for the largest component share in 2026 because production deployments are initially focused on direct-attached devices. Switches & pooling systems are projected to record the highest CAGR as operators move toward shared rack-level capacity. Controllers will remain strategically important, with suppliers including Astera Labs and Montage Technology supporting connections between DRAM and CXL platforms.

DDR5-based devices are expected to lead the memory type segment in 2026, supported by products from Samsung, SK hynix, and Micron. NAND-backed & hybrid devices are forecast to grow fastest as customers seek larger memory tiers at a lower cost per gigabyte. DDR4-based products will support cost-conscious deployments and the reuse of memory removed from older servers.

EDSFF E3.S modules are projected to hold the largest form factor share in 2026, reflecting adoption by server manufacturers and compatibility with front-accessible drive bays. CXL memory boxes & appliances are expected to achieve the highest CAGR as shared capacity expands across multiple servers. Add-in cards will remain relevant for development platforms and servers without E3.S bays.

CXL 2.0 is expected to lead the specification segment in 2026 because it is supported by current server processors and validated commercial modules. CXL 4.0 is projected to grow fastest as future platforms adopt higher bandwidth, bundled ports, and memory sparing. CXL 3.x will support the transition through enhanced fabric and memory sharing capabilities.

In-memory databases & analytics are expected to represent the largest application segment in 2026. AI inference & training is forecast to record the highest CAGR as key-value caches exceed available GPU and host memory. Hyperscale cloud providers are expected to lead by end user, while enterprises are projected to grow fastest as CXL-capable servers become more widely available for on-premises AI and database infrastructure.

Regional Market Analysis