Intel has announced a major expansion of its AI and data-centre portfolio, introducing new Xeon 6+ processors, Intel Ethernet E835 networking products and fresh details about its next-generation AI accelerator platform, Crescent Island. The launch shows the company’s broader strategy of competing not just in AI chips, but across the entire infrastructure stack that powers modern AI systems. The firm argues that as AI evolves from chatbot-style models to more autonomous ‘agentic AI’ systems capable of planning, reasoning, and coordinating tasks, CPUs, networking hardware, and memory infrastructure are becoming just as important as GPUs.
The biggest announcement is the new Xeon 6+ processor family, internally known as Clearwater Forest, which becomes Intel’s first data-centre CPU built on its advanced Intel 18A manufacturing process. The chips scale up to 288 Efficient-cores per socket, making them among the highest-core-count x86 server processors ever introduced. Intel claims up to 2.5 times higher performance than the previous generation and as much as 45% better performance-per-thread-per-watt versus competing products.
The processors support 12-channel DDR5 memory, 96 PCIe Gen 5 lanes, Compute Express Link (CXL) support and as much as 576MB of L3 cache. The company is also promoting up to 9:1 server consolidation compared with older 2nd-generation Xeon platforms, a figure aimed at reducing data-centre footprint, cooling requirements and operating costs.
According to the firm, Xeon 6+ was specifically designed for environments where performance-per-watt, rack density and predictable latency matter more than raw clock speeds. Intel is also introducing Application Energy Telemetry, allowing real-time workload-level power monitoring so operators can track energy usage more precisely across AI and cloud deployments.
Along with the processors, Intel expanded its networking portfolio with the Intel Ethernet E835 family. The new controllers and adapters support data rates ranging from 10GbE to 200GbE, with configurations including 2×25GbE, 4×25GbE, 2×100GbE and single-port 200GbE variants. The networking products are designed for AI clusters, cloud infrastructure, edge deployments and high-performance computing systems where data movement increasingly determines overall performance. Intel claims the E835 family focuses on high throughput, lower power consumption, secure communications and improved scalability for large distributed AI systems.
Intel also revealed new technical details about its upcoming Crescent Island AI accelerator, an inference-focused GPU platform based on the Xe3P architecture. Unlike many competing accelerators that rely heavily on expensive HBM memory, Crescent Island uses LPDDR5X memory and is designed around high-capacity, cost-efficient inference deployments. Intel disclosed that reference designs will ship with 160GB of LPDDR5X memory, while partners may eventually offer configurations reaching as much as 480GB. The architecture supports multiple data formats including FP4, MXFP4 and FP64, allowing both AI inference and certain scientific computing workloads. Crescent Island is expected to be delivered as a PCIe accelerator card with roughly a 350W power target.
The development is strategically important because it showcases Intel’s latest effort to re-establish itself at the centre of AI infrastructure. Rather than competing solely with NVIDIA in AI accelerators, Intel is positioning itself against a wider range of rivals across the data-centre ecosystem. The company is taking on AMD’s EPYC server processors and Instinct accelerators, custom AI chips developed by hyperscalers like Amazon’s Trainium and Inferentia, Google’s TPU family, and networking providers like Broadcom.
The Tech Portal is published by Blue Box Media Private Limited. Our investors have no influence over our reporting. Read our full Ownership and Funding Disclosure →
Ashutosh is a Senior Writer at The Tech Portal, largely reporting on new tech, and intersection of technology and business. Ashutosh’s career spans across nearly a decade of technology writing across multiple platforms and languages.
