Intel unveiled new technical details for its GPU D-series, formerly code-named Crescent Island, along with new Xeon-based rack designs and open software initiatives aimed at supporting large-scale agentic AI infrastructure.
The announcements were made at the Open Compute Project Global Summit and reflect Intel’s strategy of combining general-purpose CPUs, purpose-built accelerators, open software, and standards-based systems.
Intel said agentic AI is turning inference into a multi-step process involving retrieval, reasoning, orchestration, tool use, and execution.
That shift is increasing pressure on data centers to balance performance, memory, power consumption, and cost while supporting a wider variety of AI workloads.
Intel is positioning its x86 CPUs, GPUs, and specialized accelerators as components of a heterogeneous architecture that can match different workloads with the most appropriate compute resources.
The company also introduced new Xeon-based rack designs developed with ecosystem partners.
Many of the designs use the Open Compute Project’s Open Rack v3 standard to simplify deployment and integration within open data center environments.
Select Xeon 6+ configurations can scale beyond 46,000 CPU cores per rack while operating at approximately 133.45 kilowatts of total rack power.
Intel said Xeon processors can handle orchestration, data processing, I/O management, and other general-purpose tasks that keep AI accelerators operating efficiently.
The company also detailed its Intel GPU D-series, a purpose-built accelerator designed for sustained, memory-intensive AI inference.
Based on Intel’s Xe3P architecture, the GPU D-series can support up to 480 GB of LPDDR5X memory, 1.5 TB per second of memory bandwidth, and 1.3 petaflops of FP8 compute within a 350-watt power envelope.
Intel said the architecture is designed for large language models, long-context workloads, and concurrent inference services.
The GPU D-series can support models with as many as 300 billion parameters on a single card and more than one trillion parameters in an eight-GPU configuration.
Potential applications include multi-tenant AI services, robotics, and other latency-sensitive workloads.
The accelerator uses a standard PCIe form factor and air-cooled design, which Intel said should allow enterprises to deploy it in existing data center environments without specialized rack or cooling systems.
Customer sampling of the GPU D-series is already underway.
Intel is also working with OEMs, ODMs, neocloud providers, independent software vendors, appliance providers, and AI-native companies on validated platforms and deployment-ready systems.
The company said these partnerships are intended to reduce complexity and help organizations move AI workloads from evaluation to production more quickly.
Intel is pairing its hardware strategy with open software designed to support heterogeneous AI environments across CPUs and GPUs.
The company said the GPU D-series uses a common Xe software foundation, allowing developers to reuse applications across Intel GPU platforms.
Intel is also supporting widely used AI frameworks, inference engines, development tools, performance libraries, and programmable software layers.
The company is working to provide Day 0 support for emerging open AI models across Intel GPU D-series and Xeon platforms.
Intel’s AI for Enterprise Agent Toolkit is already optimized for Xeon processors and provides orchestration capabilities for production agentic applications.
The toolkit’s architecture is also designed to extend to Intel GPUs, providing a path for enterprises to scale multimodal and agentic workloads across different types of compute.
Intel continues to support open data center standards through the Open Compute Project, where it has made 116 contributions since 2011.
The company is also contributing to the Open Adapter Card workgroup, which is developing a standardized sub-rack accelerator form factor intended to improve mechanical, electrical, and management interoperability.
Intel said its broader goal is to give enterprises the flexibility to expand AI systems without replacing existing infrastructure as workloads and technologies evolve.
KEY QUOTES:
“Open models are expanding the options enterprises have for deploying AI. 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.’”
Nicolas Dubé, Corporate Vice President, Data Center Systems and Solutions Group at Intel
“Inference is becoming the defining AI workload, and enterprises need infrastructure that delivers the right performance at the right cost. With Intel GPU D-series, we’re focused on making high-capacity inference practical to deploy and economical to scale.”
Anil Nanduri, Vice President, AI Products & GTM, Intel Data Center Group
“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.”
Kevork Kechichian, Executive Vice President and General Manager, Data Center Group at Intel
