Intel’s journey towards Edge-Native AI

Reflecting on 2023, I can’t help but be amazed at the milestones we have achieved in Intel's Network and Edge Group (NEX) just over the last year. I've been so proud to watch our commitment to developing open, edge-native platforms and solutions that bring AI everywhere become even more tangible with Project Strata and the announcement of our?edge-native software platform that will enable developers and customers to build, deploy, run, manage, connect, and secure distributed edge infrastructure and applications.?

Reflecting on 2023 innovations?

Infusing AI at the edge has become a vital business imperative. Many of the partners I talk with want to add AI to their operations as fast as possible. To “bring AI everywhere,” the NEX team was involved in key initiatives announced this year that positioned Intel to make AI accessible and scalable anywhere our customers can imagine it. In addition to announcing Project Strata, NEX:??

• Released our 4th Gen Intel Xeon Scalable Processors with Intel vRAN Boost , which provides twice the vRAN capacity and 20% compute power savings. This is critical as we move AI to the edge as it integrates vRAN acceleration directly into the processor, eliminating the need for external accelerators and reducing system complexity and cost.?
• Expanded our strategic collaboration with Ericsson to enable higher 5G network capacity, energy efficiency, and scalability through our Intel 18A process technology. This is the most advanced node on Intel’s ‘5 nodes in 4 years’ roadmap and is key to building sustainable and resilient networks supporting edge and hybrid AI technologies.?
• Supported Google Falcon with our IPU E2000 series. Falcon focuses on evolving Ethernet to address low tail latency and congestion handling at scale to support critical AI and HPC workloads. This commitment of an industry leader highlights the importance of open standards such as Ethernet for AI networking fabrics.?
• Announced OpenVINO will be Intel’s AI inferencing and deployment runtime of choice for developers on client and edge platforms. OpenVINO has seen tremendous traction, increasing 90% since last year in developer downloads, reaching 2.7M in total through November.??

Moving to the edge?

Innovation has always been a business differentiator. Using edge AI, businesses can adapt new processes in days rather than months, find more efficient ways to operate, and be able to react to customer needs more dynamically. What stands in the way? We see three key challenges:??

1. Reducing operational complexity. Modern edge environments are a complex mix of disconnected—and often legacy—systems and software that often do not communicate or interoperate with one another.??
2. Conquering development complexity. At the edge, we encounter a multitude of nodes, operating systems, connectivity protocols, compute and storage needs, energy, cost and size constraints, and compliance restrictions built for central clouds. Unfortunately, the technologies usually employed are not optimized for the edge where they must function securely and at machine speed.???
3. Addressing the unique nature of AI at the edge. Edge AI primarily involves inference and customizing pre-trained models for quick impact. It requires running AI alongside existing software on heterogeneous hardware, often in low-power and space-constrained environments.?

Industry analysts see edge intelligence evolving through three phases—1) from highly custom, bespoke use cases that have been used for decades 2) to the industry-specific solutions that have interoperability and inefficiency challenges 3) to a foundational cross-industry platform that addresses common integration challenges every industry faces. They predict the industry won’t reach this final stage until 2030. The NEX team plans to beat that estimate by many years with our current development roadmap.?

Putting AI anywhere and everywhere?

Tightly integrating software, hardware, and platform solutions explicitly developed for AI at the edge creates a digital ecosystem that supports AI capabilities wherever they are needed.?Our modular, unified platform approach includes four aspects:?

1. Silicon built for edge: Our architectures with built-in networking and AI acceleration are designed to operate at the edge to improve performance, efficiency, and security.??
2. Infrastructure software: A software-defined approach makes it easier to deploy, manage, orchestrate, and secure the edge and the applications that run on top of it. This software is critical to abstracting the complexity inherent in heterogeneous edge silicon.??
3. Common AI application runtime from edge to cloud: Cloud-based AI models must be fine-tuned for specific industry needs and be able to run on the smallest of devices. OpenVINO enables seamless experiences with AI across any model, any hardware, anywhere, without rewriting the application.??
4. Vertical-specific solutions: Offered through our partners or directly from Intel, vertical-specific solutions solve complex challenges specific to different industries. These solutions—from healthcare to manufacturing to retail to logistics—result from years of repeatable success driving clear proof of business outcomes and often providing greater ROI than horizontal solutions.??

Looking forward to 2024??

At the edge, the network and application are inexorably intertwined. To realize global network and edge transformation, our mission is to bring AI everywhere. This means developing solutions that shift the world’s physical infrastructure into software with ubiquitous and seamless connectivity everywhere.??

I’m an engineer at heart. I delight in building the best solutions that technology will allow and making them available to everyone. The lifeblood of an AI future is an open ecosystem. Empowering the ecosystem is a core tenet of Intel’s culture, and it’s why we’re a leading voice in every major edge standards body and driving initiatives like DPDK, IPDK, oneAPI, and OpenVINO to enable choice and help developers port applications across boundaries, and across vendors.??

For everyone to succeed, we are meeting developers and enterprises where they are?on their journey. The foundation is a unified, modular platform built on open standards. The platform abstracts the complexity of deploying AI everywhere. Software building blocks like low-code Intel Geti put AI in the hands of domain experts for smarter solutions.?Innovations in silicon—like Infrastructure Processing Units that transform the power, performance, and cost efficiency of data centers, Ethernet Controllers and Adapters that transport AI data seamlessly from cloud to edge with reduced latency and responsive throughput, and technologies like Precision Time Protocol?for higher accuracy synchronization signaling to the edge—contribute to our goal getting AI everywhere the enterprise needs it.??

Join us at the edge?

In today’s hypercompetitive environment, analog-only business will not survive. The ability to operate with speed is just as crucial as innovation. Putting AI-based automation at the edge as quickly as possible is a business imperative.??

I am excited about the planned launch of our edge-native software platform in the first quarter of 2024. Infrastructure operators, developers, and enterprise verticals will be able to utilize it to accelerate the deployment of edge AI solutions. Intel’s mission is to lift the world’s infrastructure into software, enable pervasive and seamless connectivity, and bring AI everywhere. The NEX team believes our work will help the company meet this goal by building a unified platform of software, hardware, and ecosystem solutions. For updates on our progress, follow me on LinkedIn and visit our websites Edge Computing Solutions and Technologies and Powering 5G Networks and Applications from Cloud to Edge .???