Intel

Intel—now Intel Foundry—has been a leader in industrial automation for decades, long before commercial analytics platforms could meet our highly complex needs. Therefore, we developed many of our own analytics solutions. Since then, the software industry has matured significantly and presents an opportunity to shift from a build-to-suit model toward a buy-and-standardize approach. We are now adopting a third-party analytics platform that offers several advantages, including a common user interface and the ability to deliver new analytics services 30–40% faster. It also lays the foundation for the next generation of manufacturing analytics, including unified data and AI.

Intel IT developed a unique framework that treats the digital employee experience (DEX) as a measurable, managed capability rather than an abstract sentiment. We defined what experience means at the point of work, instrumented it with a combination of objective telemetry and employee sentiment, and used the resulting signal to drive continuous improvement across three domains: the devices employees use, collaboration tools, and the IT support experience when issues arise. The framework includes dashboards, weekly prioritization meetings, and governance. It enables us to focus on the most impactful DEX areas and helps us—and employees—be more productive.

Every enterprise has vast stores of historical data. This data, if mined effectively, can help improve operational efficiency. However, traditional methods of automated analysis, called clustering, rely on manual labeling and lack sufficient granularity to create meaningful results. Our innovative, domain-agnostic analysis framework can efficiently analyze an unlabeled corpus of unstructured content and automatically discover a hierarchical taxonomy of content categories. We applied this adaptive recursive density-based clustering framework to our IT incident dataset. The resulting multi-level hierarchical view of all IT incident tickets—augmented by visualization tools and GenAI—boosts IT productivity and enables proactive efficiency improvements.

From Data Fragmentation to Strategic Intelligence: An AI-Driven Framework for Enterprise Data Asset Optimization

Intel IT runs Intel data center services like a factory, accomplishing change in a disciplined manner and applying breakthrough technologies, solutions, and processes. This enables us to meet Intel’s business requirements while providing our internal customers with effective data center infrastructure capabilities and innovative business services. Building on previous investments and techniques, our data center strategy has generated savings exceeding USD 13.36 billion from 2010 to 2025. Using disruptive server, storage, network, infrastructure software, and data center facility technologies can lead to unprecedented quality-of-service levels, TCO reduction, operational efficiency, and environmental responsibility.

As AI becomes a mission-critical capability across industries, successful enterprise deployments increasingly depend not just on powerful models, but on the infrastructure, operational frameworks, and expertise required to deliver them reliably and at scale. This white paper offers insights into systems integration challenges that span hardware design, software orchestration, data engineering, network optimization, and continuous operations. This paper also describes evolving AI infrastructure landscapes, from training clusters to production inference platforms, and maps the diverse requirements that define real-world use cases.

The reference AI inference stack presented in this paper illustrates the layered ecosystem required to operationalize AI effectively, while the exploration of deployment challenges reveals why specialized proficiency is indispensable for success. Intel IT technical leadership has the relevant practical expertise to partner with technical leads in Intel Sales and Marketing as well as with Intel business units (such as Data Center Group, AI Group, and Platform Engineering and Architecture) to help Intel customers with their AI initiatives, accelerate time-to-value, and unlock the full potential of their AI investments.

For each AI use case, a “minimum scalable unit” is proposed, which includes cluster size, specifications for server infrastructure (CPUs, GPUs, and DRAM), network (frontend and backend), storage (local and shared), and data center (space and power). It also introduces the applicability of disaggregated servers, patented by Intel, along with a full software stack that includes IT manageability capabilities and operation lifecycle automation.

The paper concludes with an exploration of real-world challenges in deploying AI inference solutions, including latency optimization, scaling strategies, and lifecycle management, followed by the articulation of the strategic value proposition of the Intel IT organization’s technical expertise to operationalize and optimize these complex systems.

Two market trends are driving the need for a higher CPU:GPU ratio in AI infrastructure. First, the inferencing market is accelerating. A higher CPU:GPU ratio is required for inferencing compared to training, especially as agentic AI becomes prevalent. Second, Reinforcement Learning is now being used in complex 3D simulation environments that require a high number of CPUs. Together, these two trends demonstrate that CPUs play a strategic AI infrastructure asset, acting as the control plane for AI systems. Future AI infrastructure planning must explicitly model CPU growth as a first-class driver of cost, performance, and energy-per-token economics.

Intel IT plays a pivotal role in Intel’s business transformation, especially in our role of driving data quality across both the Foundry and Product domains and maintaining master data. In 2025, we evaluated our data, analytics, and AI (DAAI) maturity level. The evaluation revealed that, while our work to date (such as master data governance, a modern data lakehouse, and a centralized data marketplace) represents best-in-class traditional data management capabilities, we have significant work ahead to truly enable AI-ready data. Areas of focus include improving data lineage observability, increasing AI literacy across the enterprise, and establishing better unstructured data management.

With agentic AI, Intel IT is pursuing self-optimizing, human-aligned facility intelligence. To keep up with the pace of business, Intel’s Industrial Automation team, working with the Facility Operations team, is helping Intel facilities transition from mere automation to complete autonomy. Agentic AI combines machine learning, digital twins, and Human-in-the-Loop (HITL) oversight to orchestrate intelligence across building systems, utilities, and sub-fabs. This distributed, context-aware, multi-agent architecture integrates perception, reasoning, and action. Our vision for facilities automation extends far beyond predictive maintenance to achieve self-optimizing, autonomous facilities by the end of the decade, setting a new standard for Industry 5.0 operations.

Intel IT is expanding and evolving our “Citizen AI” framework—named iGPT, which enables Intel employees to use Generative AI (GenAI) to streamline processes and enhance efficiency. This democratized AI platform empowers users to build chatbots and AI agents using intuitive, no-code tools. This enables them to create solutions tailored to their specific needs and quickly adapt to a rapidly changing business landscape. To improve scalability and potentially reduce cost per token, we are deploying our Citizen AI platform, both on-premises and in the cloud, using a mix of newer Intel Xeon processors and Intel Gaudi accelerators.

Intel IT now runs part of our database as a service (DBaaS) in an open-source containerized environment, in response to vendors’ move to per-core licensing. The new bare-metal infrastructure is driving cost efficiencies and future-proofing IT investments. We have also added important new capabilities that improve resource utilization and enhance the user experience. The new infrastructure avoids per-core virtualization license fees, lets us provision resources on demand, and eliminates several thousand cores. We expect the new infrastructure to save USD 2.5 million in hypervisor license costs annually, with an estimated return on investment period of just six months.