Optoelectronics
chip communication. By leveraging optical rather than electrical signalling, CPO can significantly reduce power consumption, latency, signal distortion, and thermal challenges associated with high-bandwidth data transfer — an increasingly critical factor in large-scale AI infrastructure. NVIDIA unveiled its latest CPO network switches “Spectrum-X” in March 2025, boasting 3.5 times the power efficiency, 63 times better signal integrity, and 10 times the network resiliency of traditional systems. Meanwhile SiPh optical switches also present a compelling alternative to traditional electronic switches by utilising light to route data within integrated circuits (ICs). This optical approach enables high-speed data transfer with reduced latency and lower energy consumption. In AI data centres, the adoption of SiPh optical switches could significantly enhance data processing, addressing the growing demands for bandwidth and energy efficiency.
Finally, at the edge, SiPh-based sensing technologies are emerging as a compelling option for real-time AI applications. These sensors, which leverage the high sensitivity and bandwidth of photonics, could enhance autonomous systems, smart city infrastructure, and medical diagnostics by providing accurate and low-power data acquisition. While SiPh-based sensors promise superior signal fidelity and reduced energy consumption compared to conventional electronic sensors, their CMOS-compatible design can also deliver significant cost reductions compared to more complex electronic platforms.
Role of C-PIC and research in helping SiPh meet the needs of AI Successfully integrating new technologies like SiPh into complex fields such as AI necessitates a unique blend of industry and academic collaboration to foster research and technological scaling. Helping to realise this task is C-PIC.
Established with the support of the Engineering and Physical Sciences Research Council (EPSRC), C-PIC is a joint effort involving the Universities of Southampton and Glasgow, as well as the UK's Science and Technology Facilities Council (STFC). Building on the success of the CORNERSTONE Foundry, it will serve as a linchpin for SiPh development, supporting a broad range of academic research projects and industry R&D through its open foundry service and advanced technology platforms. These open platforms are designed to facilitate applications across diverse sectors, from telecoms
www.cieonline.co.uk
and LiDAR to sensing, AI, and quantum technologies. Each platform is equipped with a standard component library that significantly lowers the barrier to entry for non-photonics experts, fostering innovation and accelerating the adoption of SiPh into new applications.
Advancing SiPh for AI with academic research
Professor Frederic Gardes from the University of Southampton has investigated how SiPh could benefit AI hardware. His work includes developing novel optical modulator systems that leverage the plasma dispersion effect — where free carriers in a semiconductor change its refractive index — to achieve high-speed weight matrix multiplication and create optical feedback loops for solving linear equations in a CMOS-based device. While not currently commercialised, this type of technology could offer significant improvements in AI processing speed and energy efficiency over today’s electronic systems.
Furthermore, Professor Gardes’ research is investigating the potential of non-volatile photonic memory elements to operate as optical memory switches or storage devices – similar in principle to CD-ROM technology but utilising heat pulses to modulate a SiPh device phase. These non-volatile memories could prove complementary to AI processing and traditional electronic memory and provide improved power efficiency and increased communication speed.
CORNERSTONE’s advanced fabrication and design capabilities have provided the system architecture necessary for some of these research projects, and C-PIC’s expanded services will similarly support future endeavours. Although some challenges like packaging complexities and managing noise in analogue components remain, research projects like this help to establish the potential pathway toward energy-efficient, ultra-fast SiPh-based processing that could underpin the next-generation of AI hardware.
Insights from industry from Salience Labs
C-PIC had the chance to speak to Dr. Xia Chen, head of photonics at Salience Labs. A key focus for him and his team is developing SiPh-based optical switches as a replacement for copper interconnects in AI networks. These optical switches are engineered to provide reduced latency, lower power consumption, and minimal signal distortion – advantages critical for high-bandwidth AI applications. The system uses optical input and output, maintaining an all-optical data path and eliminating the need for costly transceivers.
Figure 2: Integrated photonic device mounted and electrically connected for testing or deployment (Source: University of Southampton).
Salience Labs is dedicated to developing integrated photonic chips for AI, focusing on components that address the limitations of traditional electronic switches. As AI models are largely limited by data transfer speeds, a Salience switch with its all-optical networking eliminates bottlenecks, resulting in low-latency, high-bandwidth communication, leading to faster operation completion with reduced power consumption and costs. Dr. Xia Chen believes SiPh is the ideal solution for AI hardware, and the industry must continue to work to achieve low-loss manufacturing and expanded SiPh functionality. It is here where C-PIC continues to work alongside industry to drive SiPh into AI applications by providing open technology platforms, low- cost rapid prototyping, as well as training and expertise – areas that Dr. Xia Chen believes will be key as both SiPh and AI continue to mature.
Conclusion: symbiotic success The observed rapid growth of AI requires increasingly faster processing speeds, improved energy efficiency, and lower costs for hardware. Traditional electronics are increasingly strained by these requirements, particularly as data
References: 1
2 3
centres and edge devices must handle enormous volumes of information. In this challenging environment, SiPh emerges as a potential solution, offering the advantages of high-speed optical communication and low energy loss while addressing cost constraints.
As well as supporting SiPh’s integration into AI, C-PIC is also actively exploring how AI can fuel further innovation in SiPh. By making key metrics available to AI, C-PIC’s teams are hoping to facilitate a new generation of AI-enhanced SiPh development that can drive hardware improvements capable of continuously supporting ever-growing AI data centres. Ultimately, managing the challenges of escalating power consumption and hardware costs while increasing processing speed is not merely a technical necessity – it is a gateway to transforming society. AI could drive profound benefits across sectors – from healthcare and autonomous vehicles to smart urban infrastructures – delivering innovations that are both environmentally responsible and socially transformative, but to do so it will require new and more effective hardware solutions, something that SiPh and C-PIC aim to deliver.
https://www.cornerstone.sotonfab.co.uk https://www.iea.org/reports/electricity-2024/executive-summary
https://www.iea.org/commentaries/what-the-data-centre-and-ai-boom-could-mean-for-the-energy-sector https://www.iea.org/commentaries/what-the-data-centre-and-ai-boom-could-mean-for-the-energy-sector
Components in Electronics September 2025 35
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60
