At the Open Compute Project (OCP) Global Summit today, Intel Corporation shared new technical details on the Intel GPU D-series. The company formerly code-named the GPU Crescent Island. Intel also introduced new Intel Xeon-based rack designs. In addition, it highlighted open software and ecosystem innovations that strengthen Open Infrastructure to Power the Agentic AI.
Open models are reshaping the economics of agentic AI. As a result, demand is growing for more flexible, cost- and power-efficient infrastructure. Together, these advancements reinforce Intel’s approach to open, heterogeneous AI infrastructure. The approach combines general-purpose compute, purpose-built acceleration, open software and standards-based systems. Consequently, it optimizes performance and efficiency across AI workloads.
Agentic AI Turns Inference Into a Systems Challenge
Agentic AI is transforming inference into a multi-step workflow. That workflow spans retrieval, reasoning, orchestration, tool use and execution. Notably, x86 CPUs play a central role in it. Open models give enterprises greater choice. However, power constraints, memory bottlenecks and cost pressures make scaling AI increasingly a systems challenge.
Meeting these demands requires turnkey, heterogeneous infrastructure that is easy to deploy and manage. Intel builds on the established x86 software ecosystem. Through open, interoperable software, it brings CPUs, GPUs and specialized accelerators together. This setup matches the right compute resources to each task. Therefore, utilization, performance and efficiency improve across the entire workflow.
“Open models are expanding the options enterprises have for deploying AI,” said Nicolas Dubé, corporate vice president, Data Center Systems and Solutions Group, Intel. “To realize their full potential, organizations need infrastructure that can match different workloads with the right intelligence, right compute resources and right economics. That’s what we mean by ‘put every prompt in its place.’”
Intel Xeon: The Compute Foundation for Agentic AI
At OCP, Intel showcased new Xeon-based rack designs. The company developed them with ecosystem partners to meet the growing compute demands of agentic AI. Many designs leverage the Open Compute Project’s Open Rack v3 (ORv3) standard. This standard simplifies deployment and integration within open data center infrastructure.
Agentic workflows grow more complex. Accordingly, they place greater demands on general-purpose compute. That compute orchestrates workloads, processes data, manages I/O and executes tasks that keep accelerators productive. To address these needs, select Xeon 6+ configurations scale beyond 46,000 cores per rack at 133.45 kW total rack power. These configurations deliver greater CPU compute density to support agentic AI at scale.
Intel builds on two decades of rack-scale innovation and the broad x86 software ecosystem. Xeon provides the high-density, general-purpose compute foundation needed to scale agentic AI. Alongside GPUs and specialized accelerators, Xeon orchestrates workloads, executes tasks and manages data movement. Thus, the entire system stays productive. As AI infrastructure evolves, performance is increasingly measured by the number of concurrent agents a rack can sustain at target latency. Xeon helps maximize that delivered work by keeping data flowing, tools executing and accelerators fed.
Intel GPU D-Series: Purpose-Built for Efficient Inference at Scale
Intel GPU D-series, code-named Crescent Island, aims to improve the economics of sustained, memory-intensive AI inference. It runs on Intel’s latest Xe3P architecture. The GPU combines up to 480 GB of LPDDR5X memory, 1.5 TB/s of memory bandwidth and 1.3 petaflops of FP8 compute. All of this fits within a 350-watt power envelope. This balance of memory capacity, compute performance and power efficiency reflects Intel’s approach to real-world inference workloads.
Intel designed the GPU for Open Infrastructure large language models, long-context workloads and concurrent inference services. It supports models with up to 300 billion parameters on a single card. Moreover, an eight-GPU system handles more than one trillion parameters. Target applications range from multi-tenant AI services to robotics and other latency-sensitive workloads.
The GPU uses a standard PCIe form factor, an air-cooled design and a common Xe software foundation. Hence, it simplifies deployment within existing data center environments. It also supports application portability across Intel GPU platforms. Customers can scale inference without specialized rack or cooling infrastructure.
“Inference is becoming the defining AI workload, and enterprises need infrastructure that delivers the right performance at the right cost,” said Anil Nanduri, vice president, AI Products & GTM, Intel Data Center Group. “With Intel GPU D-series, we’re focused on making high-capacity inference practical to deploy and economical to scale.”
To bring Open Infrastructure capabilities into production, Intel collaborates with a broad ecosystem. The group includes OEMs, ODMs, neoclouds, independent software vendors, appliance providers and AI-native companies. Together, they focus on validated platforms, software integration and deployment-ready solutions. These efforts reduce complexity. They also help customers move faster from model evaluation to production. Customer sampling is underway. Meanwhile, Intel continues to expand its hardware, software and open-source ecosystem. Enterprises gain greater flexibility to deploy and scale inference across existing and new data center infrastructure.
Open Software for Heterogeneous AI
Heterogeneous AI infrastructure needs open software to work at scale. That software connects compute resources, simplifies deployment and lets developers run workloads on the right architecture. Intel’s open software strategy emphasizes native upstream integration with industry-standard AI frameworks, inference runtimes and deployment environments across CPUs and GPUs. Intel builds on tools developers already use. In this way, it simplifies development, optimizes performance and accelerates deployment through several key capabilities:
Simplified deployment across Intel GPUs: Intel GPU D-series builds on a common Xe software foundation. It supports software portability across Intel GPU platforms. Developers can reuse existing applications and reduce the effort required to deploy workloads on new hardware.
Open frameworks and performance software: Intel supports the AI frameworks, serving engines and development tools already used across the industry. Performance libraries and programmable software layers complement them to optimize inference throughput. These same frameworks support inference and agentic applications on Intel Xeon processors. Developers can therefore work across CPU and GPU architectures.
Day 0 support for emerging open models: Through upstream framework integration and model enablement, Intel works to support new open models as they become available. Support covers Intel GPU D-series and Xeon processors. Recent efforts include Day 0 support for Meta’s Muse Glimmer. This helps developers evaluate and deploy emerging agentic AI workloads more quickly.
Built for real-world AI applications and agents: Intel’s software stack supports diverse workloads. These range from high-throughput token generation to coding agents and complex enterprise workflows. The Intel AI for Enterprise Agent Toolkit is optimized and deployable on Xeon processors today. It provides orchestration capabilities for production agentic applications. Its provider-based architecture is designed to extend to Intel GPUs. Enterprises thus gain a flexible foundation for multimodal workflows and a path to scale across heterogeneous compute resources.
Open Infrastructure Through the Open Compute Project
Intel continues to advance open infrastructure through its work with the Open Compute Project. It has made 116 contributions since 2011. This year, Intel is showcasing its contribution to the Open Adapter Card workgroup. The company developed it with industry partners to define a standardized sub-rack accelerator card form factor. The effort aims to improve mechanical, electrical and management interoperability across system designs. As a result, system builders and cloud providers get a more consistent path for integrating accelerators into open infrastructure.
For customers, the value lies in the freedom to evolve AI systems without starting over. They can build on existing investments and add new compute technologies as workloads change. Consequently, enterprises scale inference more efficiently, manage costs and keep choice over how their infrastructure evolves.
Underlining this, Kevork Kechichian, executive vice president and general manager, Data Center Group, Intel said: “Agentic AI is changing what customers need from their data centers. Success will depend not on any single component, but on open, heterogeneous infrastructure that brings together Xeon, purpose-built acceleration and interoperable software. Intel is helping customers build on their existing investments while creating a flexible foundation to scale emerging AI workloads efficiently.”
Explore IT Tech News for the latest advancements in Information Technology & insightful updates from industry experts!
News Source: Businesswire.com
