resources | Red Hat and Edgescale: AI physical operations at the edge - HTML

Industries like manufacturing, utilities, and logistics are generating massive amounts of data at the edge—where people and machines work. However, scaling AI in these physical environments is difficult. The data sources are largely disconnected from one another, and integrating the diverse technologies that are already in place is a complex task. Software and AI thrive in public cloud environments, but deploying them on physical systems with intermittent connectivity and demanding performance requirements is a major challenge.

To bridge this gap, Red Hat and Edgescale are working together to extend open hybrid cloud environments directly to operational sites. This joint solution combines Red Hat’s industry-leading enterprise open source platforms with Edgescale’s Cube, a fully managed, virtual connected edge (VCE) appliance, to deliver a unified and scalable foundation for AI in the real world.

Edgescale allows enterprises to outcompete by applying AI to their operations at physical sites. Edgescale’s flagship product, the Cube, is an edge appliance that supports the security-focused operation of cloud-native software in physical places and delivers data sovereignty. By combining the public cloud ecosystem with a security-focused physical appliance, the Cube eliminates the friction of custom engineering, allowing enterprises to use data at its source for real-time operational effect. The Cube is fully local, fully private, air-gapped, and System and Organization Controls (SOC 2) compliant.

Red Hat software powers the infrastructure for Edgescale’s Cube, so Red Hat customers can run modern cloud-native workloads on a familiar, compatible, and highly reliable platform.

Edgescale’s Cube is a fully managed device that combines software and hardware systems to deliver private AI inference at the edge of the network. With the full suite of Red Hat® AI products on board, the Cube acts as an on-site AI brain for various physical environments, including factories, hospitals, labs, power plants, and other critical infrastructure.

Red Hat’s AI suite: vLLM, llm-d, Red Hat Enterprise Linux,and Red Hat OpenShift

For enterprises that need to enforce security and governance standards consistently—whether in public cloud environments or on a factory floor—Red Hat provides the trusted foundation. To power advanced gen AI at the edge, Edgescale’s Cube incorporates vLLM as its dedicated AI runtime, delivering high-performance execution for large language model (LLM)-based inference and model fine-tuning directly on the appliance. Edgescale also integrates llm-d as an intelligent software routing layer that connects Cubes and allows them to interact with one another. This routing layer intelligently directs prompts based on context, sending generic queries to public models while ensuring specific operational questions are routed appropriately, prioritizing security to the local Cube where that proprietary factory data resides. Red Hat Enterprise Linux® and Red Hat OpenShift® respectively serve as the operating system (OS) and orchestration layer for the Cube, ensuring forward compatibility for Red Hat-centric environments and meeting the high reliability and security requirements critical for physical systems.

Edgescale’s VCE, aka The Cube 

The Cube is a fully managed, plug-and-play appliance that comes online in minutes, not months. It integrates high-performance compute (featuring Intel, NVIDIA, and Qualcomm chipsets) with the Red Hat AI software configuration that delivers private AI inference at the edge. The Cube acts as a security-focused bridge, autonomously ingesting data from heterogeneous equipment (sensors, cameras, programmable logic controllers (PLCs), etc.), and allows low-latency decision-making without relying on constant cloud connectivity.

By running Red Hat software on Edgescale’s Cube, organizations can overcome the pilot purgatory that plagues other AI methods at the edge and achieve massive scale.

• Plug-and-play simplicity and speed: The joint solution allows for rapid deployment—online in minutes—and can be scaled across thousands of locations without extensive systems integration work. This approach allows engineering teams to use their existing continuous integration and continuous delivery (CI/CD) processes to push updates quickly and safely.
• Operational autonomy and reliability: The solution offers dedicated local compute for low-latency data processing. It delivers 99.9% uptime and can operate fully offline, ensuring that operations continue even when internet connectivity is lost.
• Data sovereignty and security: The architecture delivers true data sovereignty. Data remains on private instances with local residency, supported by a security-first design that includes secure boot, encryption, and zero trust principles.
• Predictable total cost of ownership (TCO): The solution is up to 60% cheaper than custom alternatives because it helps customers avoid unpredictable engineering costs and excessive cloud egress fees. It is delivered as a managed Infrastructure-as-a-Service (IaaS), allowing for predictable budgeting.

The combination of Red Hat’s software tools with Edgescale’s dedicated AI appliance makes it possible for organizations to apply AI to a variety of performance-enhancing use cases across industries:

• Improve manufacturing yield and quality: Run real-time visual inspection and process optimization to detect defects earlier and reduce waste.
• Reduce downtime: Implement predictive maintenance by continuously monitoring equipment health signals and guiding diagnostics.
• Improve safety: Monitor behavior and system states to provide real-time guidance to operators, reducing safety incidents.
• Optimize energy and resource use: Lower energy use and operating cost by optimizing equipment scheduling and environmental controls based on real-time demand and occupancy.
• Cache content: Improve performance, reliability, and user experience by serving large content and updates locally rather than over constrained networks.

Ready to safely and scalably use AI with your physical systems? Visit redhat.com/industrial to schedule a discovery call with your Red Hat account executive.