AI Infrastructure Momentum Drives Memory & Processor Demand; HBM4 Supply Tightens Further
This week's component supplier updates reinforce the industry's aggressive pivot toward AI infrastructure. Memory suppliers are ramping HBM4 production to meet insatiable demand, while CPU vendors (Intel, AMD) are shipping next-generation processors built on cutting-edge process nodes. Storage innovation continues with PCIe Gen5 mainstream adoption. The overarching theme: enterprise data centre builds must lock in component allocations early, validate platform balance (CPU/GPU/memory), and consider refurbished alternatives to manage cost and lead time pressures.
Micron
Micron Powers AI Everywhere at COMPUTEX 2026
Micron showcased its leading-edge 256GB DDR5 RDIMM, capable of speeds up to 9,200 megatransfers per second (MT/s)—a 40% performance boost over prior generations. The company has fully contracted its 2026 HBM output and committed to approximately $200 billion in new capacity to address persistent memory supply constraints. Micron's CEO confirmed that tight industry conditions across DRAM and NAND flash are expected to persist through and beyond 2026, driven by AI workload scaling.
Read sourceServero Note
DDR5 RDIMMs at 9,200 MT/s are now standard for high-performance enterprise builds. Lock in memory orders early; HBM availability remains constrained for AI accelerator platforms. Consider refurbished DDR5 modules as a cost-effective interim solution.
Phison
CES 2026: Phison Showcases Gen5 Storage to Redefine User Experiences
Phison's flagship E28 PCIe Gen5x4 NVMe 2.0 SSD delivers up to 14.9 GB/s sequential reads and 14 GB/s sequential writes, while consuming just 7W power. The company is bringing PCIe Gen5 mainstream and has also unveiled 61.44TB enterprise SSDs for large-scale data centre deployments. Gen5 adoption is accelerating across enterprise platforms, offering significant bandwidth improvements for AI and HPC workloads.
Read sourceServero Note
PCIe Gen5 storage is now production-ready for enterprise builds. The E28's power efficiency (7W) is critical for dense server configurations. Validate motherboard/controller support before specifying; Gen5 is becoming standard for new-build data centres.
Intel
Intel Announces New AI Innovations at COMPUTEX 2026
Intel launched the Xeon 6+ processor (codenamed Clearwater Forest), built on Intel's 18A process and featuring up to 288 Efficiency cores in a single socket. The new Xeon 6+ is designed for high-density, scale-out cloud and data centre workloads, with Supermicro already launching compatible server solutions to reduce TCO and accelerate time-to-online for large-scale deployments.
Read sourceServero Note
Xeon 6+ represents Intel's aggressive re-entry into the high-core-count data centre CPU market. The 18A process delivers density and power efficiency gains. Supermicro's rapid integration signals strong platform readiness. Validate memory and storage pairing for optimal performance.
AMD
AMD Announces Production Ramp of Next-Generation EPYC Processor "Venice" on TSMC 2nm
AMD has begun production ramp of its 6th Gen EPYC CPUs (codenamed "Venice") on TSMC's 2nm process. The Venice lineup includes configurations up to 256 cores and 512 threads, delivering up to 70% performance improvements for AI and enterprise workloads. This marks the first HPC product to ship on TSMC's 2nm node, positioning AMD as a leading platform for AI infrastructure builds.
Read sourceServero Note
EPYC Venice on 2nm is a significant architecture refresh. The 256-core configuration is ideal for large-scale AI training and inference clusters. Early availability is expected mid-to-late 2026. Validate cooling and power delivery for dense deployments.
Samsung Semiconductors
Samsung Showcases Next-Generation AI Semiconductor Innovations at COMPUTEX 2026
Samsung has begun mass production of HBM4 memory chips and is increasing HBM production capacity by 50% during 2026. The company also announced mass production of the PM1763, a PCIe 6.0-based enterprise SSD optimized for next-generation AI and HPC workloads. Samsung's HBM4 delivers 46% faster performance than industry standard, running at 11.7 Gbps.
Read sourceServero Note
HBM4 supply is ramping, but allocation remains competitive. Samsung's 50% capacity increase is welcome but insufficient to meet full demand. Pair HBM4 with PCIe 6.0 storage (PM1763) for balanced AI infrastructure. Early engagement with suppliers is essential.
NVIDIA
NVIDIA Kicks Off the Next Generation of AI With Rubin Platform
NVIDIA announced the Vera Rubin platform, the next-generation AI supercomputer architecture succeeding Blackwell. Vera Rubin delivers 10 times more performance per watt than Blackwell, requires 4x fewer GPUs for equivalent performance, and reduces inference token costs by 10x. Early cloud deployments (AWS, Google Cloud, Microsoft, OCI) are planned for H2 2026. The platform features seven new chips now in full production.
Read sourceServero Note
Vera Rubin's efficiency gains are transformative for large-scale AI deployments. The 10x performance-per-watt improvement reduces operational costs significantly. H2 2026 availability means Blackwell remains the primary platform for immediate builds. Plan Vera Rubin infrastructure for Q4 2026 and beyond; validate power and cooling capacity early.
What This Means for Enterprise Builds
Memory Allocation
HBM4 and DDR5 supply remains constrained. Lock in orders 8–12 weeks in advance. Refurbished DDR5 modules offer cost savings (typically 15–20% below new) with equivalent performance and warranty.
Storage Strategy
PCIe Gen5 is now mainstream for new builds. Validate platform support (motherboard, CPU, controller). PCIe 6.0 (Samsung PM1763) is entering production; plan for 2026 Q4+ deployments.
CPU Selection
Intel Xeon 6+ and AMD EPYC Venice both deliver significant performance gains. Xeon 6+ is available now; Venice ramps mid-to-late 2026. Validate memory and storage pairing for balanced configurations.
GPU/AI Platform
Blackwell remains the production platform through 2026. Vera Rubin deployments begin H2 2026. Plan infrastructure refresh cycles accordingly.
Contact Servero for Configuration Support
Contact Servero for tailored enterprise builds, component allocation, and configuration support. As a Supermicro Authorised Partner with 25+ years of enterprise experience, we help data centre operators navigate supply constraints and deploy production-ready infrastructure.
Contact Servero