Global Unichip Corp. (fabless ASIC (Application-Specific Integrated Circuit) design service provider) and French fabless semiconductor company VSORA have taken their AI inference collaboration public at the TSMC Europe Technology Symposium, where GUC is showcasing VSORA's Jotunn8 processor, a next-generation data center AI inference chip built on TSMC's 5nm process node and CoWoS-S advanced packaging technology.

A Flagship Processor Targeting the Memory Wall

Jotunn8 is VSORA's flagship AI inference processor, purpose-built for inference workloads in hyperscale data centers. According to the companies, the processor is specifically engineered to address the memory wall bottleneck, a long-standing constraint in AI infrastructure that limits the throughput and latency performance of large-scale model deployment.

By tackling that bottleneck directly, VSORA says Jotunn8 delivers ultra-low latency and very high throughput, enabling cost-efficient deployment of large-scale AI models across cloud and enterprise environments.

VSORA, headquartered in France, describes itself as a fabless semiconductor company building AI inference infrastructure designed to run efficiently at scale. The company's stated mission is to solve the memory wall and enable real-world AI deployment at the cost and performance levels the industry requires.

GUC's Role: Full Turnkey ASIC Implementation

GUC provided comprehensive turnkey ASIC services for Jotunn8, taking responsibility for the full implementation from netlist to manufacturing. The scope of GUC's contribution spanned several technically demanding domains.

The project involved advanced chiplet architecture design and integration, along with high-bandwidth memory integration using HBM3E PHY and controller technology. GUC also implemented 2.5D die-to-die connectivity through its own 17.2 Gbps GLink-2.5D interconnect solution.

The packaging was executed using TSMC's CoWoS-S technology at three times the reticle size, a configuration that reflects the scale and complexity of the chip design.

In addition to the physical implementation, GUC conducted full-system co-optimization covering signal integrity, power integrity, and thermal integrity. Power and IR optimization work was also carried out to enhance overall system efficiency. The chip was implemented on TSMC's 5nm process node.

Patrick Wang, Senior Vice President and Chief Revenue Officer at GUC, described the showcase as a demonstration of the company's ability to bring complex AI processors to silicon.

"This achievement highlights our strong collaboration with VSORA and demonstrates GUC's ability to bring complex AI processors to silicon using advanced process and packaging technologies," Wang said.

Khaled Maalej, CEO of VSORA, credited the collaboration with GUC as central to the project's success. "Jotunn8 is designed to unlock a new level of efficiency and scalability for AI inference. Our collaboration with GUC has been key in successfully translating this architecture into a high-performance silicon solution," Maalej said.

TSMC Ecosystem at the Core

The Jotunn8 project is presented by both companies as a demonstration of what is possible within TSMC's Open Innovation Platform ecosystem. GUC is a key partner in that platform, which is designed to help customers efficiently adopt advanced process nodes and packaging technologies.

By leveraging TSMC's CoWoS-S packaging alongside the 5nm process, GUC's role extended beyond design services to providing what the company characterizes as a reliable path to production for demanding AI applications.

The choice of CoWoS-S at three times reticle size is notable in the context of advanced packaging demands for high-performance computing and AI workloads, where integrating multiple chiplets and high-bandwidth memory into a single package presents significant engineering challenges.

GUC, based in Hsinchu, Taiwan, is publicly traded on the Taiwan Stock Exchange under the symbol 3443 and maintains a presence across China, Europe, Japan, Korea, North America, and Vietnam. The company positions itself as a provider of IC implementation and SoC manufacturing services using advanced process and packaging technology.

Data Center AI Demand Driving the Push

Both GUC and VSORA frame the Jotunn8 collaboration as a direct response to growing demand for high-performance, energy-efficient AI inference in hyperscale data centers. As operators seek to scale AI deployments, the pressure on inference infrastructure, in terms of throughput, latency, power consumption, and cost, has intensified considerably.

Jotunn8's architecture is designed to support large-scale AI model deployment, and VSORA says it enables operators to scale AI deployments with significantly improved infrastructure efficiency.

The processor targets data center operators who require both high throughput and low latency without the cost penalties that have historically accompanied large model inference at scale.

The unveiling at the TSMC Europe Technology Symposium marks a significant milestone for VSORA, which works with TSMC and GUC as leading ecosystem partners to bring advanced AI processors from architecture to silicon.

For GUC, the project reinforces its stated positioning in complex AI and high-performance computing ASIC development, combining its interconnect IP, packaging expertise, and system-level co-optimization capabilities into a single integrated offering.

Winning Semiconductor Contracts Across Europe Starts With Seeing What Others Miss

The semiconductor landscape across Europe is shifting faster than most teams can track. New fabrication investments, strategic partnerships, and government-backed initiatives are reshaping the competitive map at every level. If your business is relying on fragmented sources and delayed intelligence, you are already behind the curve.

Staying ahead in this industry means knowing not just where projects stand today, but where they are heading tomorrow. The Global Project Tracking (GPT) platform by Blackridge Research gives semiconductor professionals a single, structured source of truth covering the full project lifecycle across European markets and beyond.

Whether you are targeting chip manufacturing expansions, research facility developments, or supply chain infrastructure builds, the platform delivers the kind of early-stage visibility that turns market awareness into a genuine competitive advantage.

• Upcoming Projects

• Tender Notices

• Contract Awards

• Projects Under Construction

• Completed Projects

See how the Global Project Tracking (GPT) platform by Blackridge Research can sharpen your pipeline strategy across Europe. Book a Free Demo with our team today.