HyperLight, UMC, Wavetek, and Jabil join forces to scale next-generation optical networking for hyperscale AI infrastructure
HyperLight Corporation, United Microelectronics Corporation (UMC), and its subsidiary Wavetek Microelectronics Corporation have announced a strategic collaboration with Jabil Inc. to accelerate the deployment of thin-film lithium niobate (TFLN) photonics technology for hyperscale artificial intelligence data center interconnects. The partnership aims to bring advanced optical communication solutions from research and development into large-scale manufacturing, enabling faster, more energy-efficient networking infrastructure for the rapidly expanding world of AI computing.
As global demand for artificial intelligence services continues to grow, the computing infrastructure supporting these technologies is also expanding dramatically. Modern AI systems rely on massive clusters of GPUs and specialized processors working together to train models and process enormous datasets. These clusters require high-speed data connections capable of transmitting information across servers, racks, and entire data centers with minimal delay and energy consumption.
Optical interconnects—communication links that use light instead of electrical signals—have become essential to meet these performance requirements. However, as the scale and speed of AI systems increase, traditional optical networking technologies face new challenges related to power consumption, bandwidth limitations, and manufacturing scalability. The collaboration between HyperLight, UMC, Wavetek, and Jabil aims to address these challenges by advancing the commercialization of TFLN photonics technology.
Bringing Advanced Photonics to Hyperscale AI Infrastructure
Thin-film lithium niobate is an emerging photonics technology known for its exceptional optical performance. Lithium niobate has long been used in optical modulation systems, but advances in thin-film manufacturing techniques have enabled the material to be integrated into compact photonic chips suitable for modern data center applications.
HyperLight has been at the forefront of developing TFLN-based photonic devices capable of supporting extremely high data transmission speeds while maintaining low power consumption. The company’s technology is designed to serve as the foundation for next-generation optical modules used in AI data centers.
Through the new collaboration, HyperLight’s photonics innovations will be combined with UMC’s advanced semiconductor manufacturing capabilities and Jabil’s large-scale production and system integration expertise. Together, the companies aim to create a comprehensive pathway that takes TFLN technology from laboratory development into full commercial deployment within hyperscale data center environments.
Combining Expertise Across the Semiconductor Ecosystem
Each partner in the collaboration brings specialized expertise that is critical for scaling photonics technologies to global data center markets.
HyperLight contributes its proprietary TFLN photonic platform, which enables high-performance optical modulation and communication at significantly reduced energy levels. The company’s technology is built around its TFLN Chiplet™ Platform, designed to integrate seamlessly into optical transceiver modules used in networking equipment.
UMC and its subsidiary Wavetek provide the semiconductor manufacturing infrastructure required to produce photonic devices at scale. Their fabrication facilities support both 6-inch and 8-inch wafer production, enabling high-volume manufacturing while maintaining the quality and precision required for photonic components.
Jabil, a global manufacturing and supply chain services provider, contributes expertise in large-scale system integration, optical module assembly, and high-volume electronics production. The company has extensive experience building complex technology products for data center operators and telecommunications providers.
By combining these capabilities, the partnership creates a vertically integrated manufacturing ecosystem that supports the entire lifecycle of TFLN-based optical technology—from chip fabrication to module integration and final deployment.
Addressing the Growing Demands of AI Clusters
Artificial intelligence workloads require enormous computing resources. As organizations train increasingly sophisticated AI models, they rely on massive clusters containing thousands of GPUs or specialized accelerators working simultaneously. These computing clusters must exchange data continuously at extremely high speeds.
Traditional electrical connections are often unable to keep up with the bandwidth requirements of such systems. Optical interconnects provide a more efficient alternative by transmitting information through light signals rather than electrical currents. However, even optical technologies must continue evolving to meet the demands of next-generation AI infrastructure.
One of the biggest challenges facing data center operators is managing power consumption. AI clusters already consume significant amounts of electricity, and networking infrastructure adds to this energy demand. If optical interconnects require excessive power, they can become a major bottleneck limiting the scalability of AI systems.
TFLN technology addresses this issue by enabling optical modulators that operate with significantly lower power requirements than many existing alternatives. This improvement allows data center operators to increase network bandwidth while maintaining manageable energy consumption levels.
Improving Efficiency Through Advanced Materials
Lithium niobate is widely recognized as one of the most effective materials for optical modulation due to its excellent electro-optic properties. When used in thin-film form, the material enables extremely efficient conversion between electrical signals and optical signals.
This efficiency translates into several important advantages for data center networking systems:
- Reduced power consumption in optical transceivers
- Lower requirements for laser power within networking modules
- Improved signal performance at higher transmission speeds
- Greater scalability for large-scale deployments
As optical interconnect speeds continue to increase, these benefits become even more significant. Faster transmission rates typically require more complex optical designs and higher power levels. TFLN technology helps mitigate these challenges by enabling simpler architectures with improved performance.
Enabling Mass-Market Adoption
A key objective of the collaboration is to transition TFLN technology from innovative research into practical, large-scale commercial deployment. While the advantages of thin-film lithium niobate have been recognized for years, widespread adoption requires robust manufacturing processes and reliable supply chains.
According to Mian Zhang, the partnership represents a major step toward achieving this goal. Zhang emphasized that the fundamental material advantages of TFLN become increasingly valuable as optical interconnect speeds continue to rise.
By combining HyperLight’s photonics expertise with Jabil’s manufacturing capabilities and the foundry support provided by UMC and Wavetek, the collaboration creates a scalable production pathway for next-generation optical modules.
Meeting the Needs of Hyperscale Customers
Large technology companies operating hyperscale data centers require networking technologies that can be produced and deployed reliably at enormous scale. These organizations often build infrastructure consisting of thousands of servers and networking devices distributed across multiple facilities worldwide.
For such customers, innovation alone is not sufficient. New technologies must also be manufacturable in large quantities and integrated seamlessly into existing hardware platforms.
According to Jason Wildt, hyperscale customers are specifically seeking optical solutions that combine high performance with manufacturing scalability. The collaboration between HyperLight, UMC, Wavetek, and Jabil aims to deliver exactly that combination.
By aligning photonics innovation with proven high-volume manufacturing capabilities, the partnership makes it possible to deploy TFLN-based networking solutions at the rack level within large AI data centers.
Strengthening Semiconductor Manufacturing Capabilities
UMC has played an important role in advancing TFLN technology from early research stages to qualified semiconductor manufacturing processes. The company’s foundry expertise ensures that photonic devices can be produced consistently and reliably across large wafer batches.
G C Hung noted that expanding the collaboration to include system-level integration with Jabil helps establish a complete ecosystem for TFLN deployment. This ecosystem supports the reliability, scalability, and production capacity required for modern AI infrastructure.
The involvement of UMC and Wavetek also ensures that the technology can benefit from established semiconductor fabrication processes, enabling cost-effective production as demand grows.
Unlocking Data Center-Level Benefits
While individual optical modules may deliver modest improvements in power efficiency, the impact becomes far more significant when multiplied across an entire data center.
Hyperscale facilities can contain tens of thousands of networking modules connecting servers, storage systems, and compute clusters. Even small improvements in energy efficiency per module can translate into substantial power savings at the facility level.
HyperLight’s TFLN technology offers several design advantages that help achieve these improvements. In some architectures, the technology allows optical modules to operate with fewer lasers than traditional designs. This reduction lowers both power consumption and system complexity.
By decreasing the power requirements of networking infrastructure, data center operators can allocate more energy resources to computing workloads. This additional power headroom can support higher GPU densities, larger AI clusters, and expanded workloads without requiring additional electrical capacity.
Supporting the Future of AI Infrastructure
As artificial intelligence continues to reshape industries ranging from healthcare to finance and transportation, the computing infrastructure supporting these applications must evolve rapidly.
The collaboration between HyperLight, UMC, Wavetek, and Jabil reflects the growing importance of advanced photonics technologies in enabling the next generation of data center networks. By improving the efficiency and scalability of optical interconnects, TFLN technology could play a key role in supporting the continued expansion of AI computing capabilities.
Through their combined expertise in photonics innovation, semiconductor manufacturing, and large-scale system integration, the partners aim to transform TFLN technology from a promising research concept into a foundational component of modern AI data centers.
As the deployment of AI infrastructure accelerates worldwide, innovations like these will be essential for ensuring that data center networks can keep pace with the increasing demands of artificial intelligence workloads.
Source link: https://www.businesswire.com
