Cerebras and manufacturing partner Flex are scaling production of the CS-3 AI supercomputer by approximately 7 times through 2026 at a facility in Milpitas, California, marking a significant shift in where the world's most advanced AI systems are actually manufactured. The expansion signals that despite decades of electronics manufacturing moving overseas, some of the most sophisticated AI infrastructure is being designed, assembled, and tested in the heart of Silicon Valley.

What Makes the CS-3 So Difficult to Manufacture?

The CS-3 is not a typical server or data center component. It is built on Cerebras' wafer-scale engine architecture, featuring a processor physically larger than any conventional artificial intelligence chip. The system integrates advanced liquid cooling, high-density power delivery, precision mechanical assembly, and coordinated networking infrastructure into a single platform designed for large-scale AI training and inference workloads.

Manufacturing this system presents challenges rarely encountered in traditional server production. Each CS-3 requires specialized handling processes, custom tooling, precision calibration, and extensive system-level validation that goes far beyond standard electronics assembly. Flex engineers worked closely with Cerebras to develop dedicated assembly flows, automated test stations, and new manufacturing methodologies tailored specifically to wafer-scale computing systems.

"The CS-3 does not resemble a conventional server or rack-scale compute platform. Every stage of the manufacturing process, from mechanical integration to thermal validation and final system qualification, required deep collaboration between our engineering teams," said Rob Campbell, President of Communication, Enterprise and Cloud at Flex.

Rob Campbell, President of Communication, Enterprise and Cloud at Flex

How Is Flex Scaling Production to Meet Demand?

To support the anticipated 7x production increase, Flex is expanding dedicated manufacturing operations for Cerebras in Milpitas with multiple new assembly and integration lines coming online through 2026. The footprint devoted to CS-3 manufacturing is expected to grow substantially as production accelerates to meet customer demand from AI model developers, cloud providers, and enterprise customers.

Inside the Milpitas facility, production operations now span several specialized areas and processes:

• Precision Mechanical Assembly: Custom tooling and fixtures that enable multiple CS-3 systems to move through integration and testing simultaneously, significantly increasing throughput.
• High-Power Electrical Integration: Advanced power delivery systems that handle the substantial energy requirements of wafer-scale processors.
• Liquid Cooling Installation: Specialized thermal management systems required to keep the large processors operating at optimal temperatures.
• Optical Networking Validation: Testing of high-speed interconnects that allow the system to communicate with other infrastructure.
• Full-Rack System Qualification: Comprehensive burn-in and validation testing to ensure reliability for large-scale AI deployments.

The expanded facility also includes enhanced burn-in and validation areas, additional automated test infrastructure, and increased warehouse and logistics capacity for critical components and finished systems.

Why Does American Manufacturing of AI Hardware Matter?

The partnership challenges a widespread assumption that all electronics manufacturing, including cutting-edge AI systems, happens overseas. Cerebras' Chief Operating Officer emphasized this point, noting that everyday across the United States, teams of American engineers and technicians are building state-of-the-art AI systems that power frontier AI workloads around the world.

"The CS-3 is unlike any computer system ever built, and scaling its production requires an extraordinary manufacturing partner. Flex brings the technical depth, operational rigor, and manufacturing expertise needed to support that scale," stated Dhiraj Mallick, Chief Operating Officer of Cerebras.

Dhiraj Mallick, Chief Operating Officer of Cerebras

The expansion is also contributing to growth in high-skilled manufacturing roles across the region. New positions are being created in manufacturing, systems integration, quality assurance, supply chain management, and testing, representing a meaningful investment in local technical talent.

What Are the Performance Advantages of the CS-3?

The CS-3 is built on Cerebras' Wafer-Scale Engine 3 (WSE-3), which the company describes as the world's largest and fastest commercialized AI processor. The processor is approximately 58 times larger than a leading GPU (graphics processing unit) chip, a conventional processor used for AI workloads. According to Cerebras, the WSE-3 uses a fraction of the power per unit of compute while delivering inference, the process of running trained AI models on new data, up to 15 times faster than leading GPU-based solutions when benchmarked on leading open-source models.

These performance advantages explain why demand for the CS-3 is accelerating among corporations, research institutes, and governments on four continents. The system is available both on premises and through cloud providers, giving organizations flexibility in how they deploy advanced AI infrastructure.

The Flex and Cerebras partnership demonstrates that advanced manufacturing capability remains viable in the United States when companies tackle the engineering challenges head-on. As AI infrastructure demand continues to accelerate globally, the ability to manufacture these systems domestically could become increasingly valuable for supply chain resilience and technological sovereignty.