Intel Core Ultra Series 3 Drives the Next Wave of Physical AI and Intelligent Robotics

As physical AI and robotics continue to evolve, the Intel Intel® Core™ Ultra Series 3 processors are emerging as a key computing platform for next-generation intelligent machines. One notable example is Ella, an AI-powered barista robot that now runs exclusively on Intel’s latest processors. Around the world, dozens of robotics developers are also evaluating and adopting these chips as an alternative to large, expensive, and heat-intensive discrete GPUs traditionally used for robotic intelligence and real-world decision-making.

The new architecture will be fully showcased at Computex 2026 in Taipei this June. During the event, Ella will demonstrate its capabilities by preparing up to 200 drinks per hour while introducing three newly developed AI service agents powered entirely by Intel Core Ultra Series 3 processors. These demonstrations highlight the chip’s heterogeneous computing architecture, allowing different parts of the processor to handle separate workloads simultaneously without relying on remote cloud servers, significantly reducing latency.

Building a Smarter Robot Business Model

Ella was created by Keith Tan, a former café owner in Singapore who faced persistent labor shortages and inconsistent beverage quality in the food service industry. After spending years training baristas only to see them leave shortly afterward, Tan turned to robotics as a long-term solution. However, early robotic systems introduced a new challenge: the need for costly standalone GPUs to manage order processing, beverage preparation algorithms, and robotic arm coordination.

In many cases, the GPU alone cost more than the rest of the system combined, making profitability nearly impossible for businesses selling affordable beverages such as a $5 latte.

Tan explained that earlier systems used Intel CPUs alongside separate GPUs to manage workloads, but the overall cost was unsustainable. He realized that any commercially viable robotic café solution needed to deliver a realistic return on investment while keeping deployment and operational costs manageable.

Today, Ella has evolved into a fully Intel-powered platform designed specifically for the service economy. By leveraging Intel Core Ultra 3 processors, Sensory AI can deploy agentic AI directly at the edge inside physical robots, enabling features such as fleet management, AI-optimized operations, and business intelligence across multiple store locations.

Since the launch of Intel’s Core Ultra Series 3 processors, Sensory AI has completely abandoned discrete GPU architectures in favor of an all-Intel system design within Ella.

Advancing Edge AI for Robotics

The shift toward Intel’s latest system-on-chip (SoC) architecture allows robots to efficiently run reasoning-focused workloads. Once trained, robots no longer require large gaming-class GPUs to perform daily tasks. Instead, they only need the ability to execute learned behaviors quickly and efficiently.

By integrating the CPU, GPU, and neural processing unit (NPU) into a single chip, Intel has significantly reduced both the thermal output and cost of robotic computing systems. This enables Sensory AI to run three specialized AI agents simultaneously:

• The Avatar agent handles customer interactions.
• The Ella agent analyzes business and store-level patterns.
• The Guardian agent monitors system health and resolves operational issues.

For example, if cups become stuck together, the Avatar agent can notify the customer while the Guardian agent determines an appropriate corrective action. Sensory AI’s orchestration software then directs the robotic arm to fix the issue automatically. Each AI agent operates on the section of the SoC best suited to its workload.

This architecture allows Sensory AI to manage vision, language, and motion control without requiring additional graphics hardware, resulting in lower ownership costs and simplified maintenance for self-service robotic kiosks.

Robotics Companies Embrace Integrated AI Architectures

Sensory AI is not alone in moving toward integrated edge AI systems. Several robotics innovators are already testing Intel’s processors for next-generation applications.

Trossen Robotics, based in Illinois, has begun evaluating Intel Core Ultra Series 3 processors for robotic arms used in food service and industrial automation. The company also develops machine-learning research kits for universities and engineering teams.

According to Trossen’s Chief Solutions Architect Marc Dostie, x86 architecture remains highly attractive because of its broad developer adoption, extensive software ecosystem, and rapid support for new frameworks. He noted that the combination of strong CPU performance, integrated graphics, and high-performance I/O capabilities makes the platform a highly effective development environment.

Meanwhile, in South Korea, Circulus is advancing physical AI and humanoid robotics through a modular operating system optimized for Intel Core Ultra Series 3 processors. The company develops products including the AI companion robot Pibo and enterprise AI workstations designed for humanoid platforms.

Circulus focuses heavily on human-robot interaction (HRI), prioritizing low-latency, privacy-focused on-device AI. By processing critical information locally, its systems can continue operating reliably even without internet connectivity, enhancing both autonomy and safety in industrial and collaborative environments.

In Italy, Oversonic Robotics is also transitioning away from discrete GPUs. The company develops humanoid and centaur-style robots for manufacturing and healthcare applications, including rehabilitation support systems.

Although Oversonic originally relied on standalone graphics processors for training and operation, it now uses Intel Core Ultra Series 3 processors with integrated acceleration capabilities. This approach enables real-time processing for language understanding, computer vision, reasoning, and motion control directly on the device, reducing both costs and dependence on cloud infrastructure.

Together, these companies are demonstrating that the future of robotics is not simply about increasing computational power, but about building intelligent, efficient, and adaptable machines capable of operating effectively in real-world environments using compact, integrated computing platforms.