Silicon photonics the key to unlocking AI’s full potential
The large language models that enable generative artificial intelligence (AI) are driving an increase in investment and an acceleration of competition in the field of silicon photonics, a technology that combines silicon-based integrated circuits (ICs) and optical components to process and transmit massive amounts of data more efficiently. Top-rank designers and manufacturers of ICs, AI systems and telecommunications equipment have all joined the race, including NVIDIA, TSMC, Intel, IBM, Cisco Systems, Huawei, NTT and imec, the Interuniversity Microelectronics Centre headquartered in Belgium. These and other organizations have been working on silicon photonics for many years, some of them (including Intel and NTT) for nearly two decades. On September 5, at the Silicon Photonics Global Summit in Taiwan, TSMC’s vice president in charge of pathfinding for system integration, Douglas Yu, told the Nikkei newspaper “If we can provide a good silicon photonics integration system … we can address both critical issues of energy efficiency and computing power for AI. This is going to be a new paradigm shift. We may be at the beginning of a new era.” In its invitation to the Summit, semi (the microelectronics industry association) noted that “Silicon Photonics has become a prominent buzzword in the semiconductor industry” due to its high bandwidth, high-speed data transmission, extensive transmission distance, low power consumption and applicability to advanced networking, computing architectures, cloud computing, data centers, autonomous vehicles and smart transportation systems. In other words, right across the leading edge, high-tech industry will benefit from silicon photonics in a way that will improve device and system performance while lowering energy consumption. According to semi, “the global silicon photonics market is projected to reach a value of US$7.86 billion by 2030, with a compound annual growth rate (CAGR) of 25.7% from its valuation of $1.26 billion in 2022.” About a year ago, in September 2022, DigiTimes reported that NVIDIA and TSMC had launched a joint R&D project called COUPE, which stands for Compact Universal Photonic Engine. The goal of the project is to combine multiple AI processors (GPUs) using NVIDIA’s silicon photonic (SiPh) technology. “Source revealed,” wrote DigiTimes, “that SiPh chip and CMOS processes go through co-packaged optics (CPO) technology integration, which can connect multiple advanced GPUs with chip-on-wafer-on-substrate (CoWoS) 2.5D IC packaging.” The combination of low-latency optical data transmission and advanced packaging technology results in a significant reduction in signal loss, making it possible to create “an ultra-large GPU set,” DigiTimes wrote. Silicon photonics is an old technology with big new cutting edge potential. Image: Twitter This technology may not be ready “until the SiPh ecosystem matures,” which helps explain why Taiwan’s leading Outsourced Semiconductor Assembly and Test (OSAT) company ASE Technology and Japan’s Advantest were also prominent at the Silicon Photonics Global Summit. At the Semicon Taiwan trade show, which followed the silicon photonics event, TSMC told the press that its ability to supply NVIDIA with AI processors is likely to be constrained until the end of next year by a shortage of CoWoS packaging capacity. By then, capacity should double. Intel defines silicon photonics as “a combination of two of the most important inventions of the 20th century—the silicon integrated circuit and the semiconductor laser. It enables faster data transfer over longer distances compared to traditional electronics while utilizing the efficiencies of Intel’s high-volume silicon manufacturing.” Intel explains that by combining “the power of optics and the scalability of silicon… optical transceivers are the optical interfaces for Ethernet switches, routers and transport networking equipment, providing connectivity for large-scale cloud and enterprise data centers.” Dylan Patel of SemiAnalysis has written that “Intel manufactures silicon photonics at the largest scale in the world. They lead in market share for manufacturing optical networking transceivers… Intel’s scale and the integrated nature of their solution has led to them leading the industry with… [a failure rate] two orders of magnitude better than the competition.” Intel plans to increase its advanced IC packaging capacity by about four times by 2025, including a new 3D packaging facility in Malaysia. In late August, Intel’s corporate vice president for manufacturing supply chain and operations, Robin Martin, told Tech Wire Asia that “Malaysia will eventually become Intel’s largest production base for 3D chip packaging.” In addition to ramping up its own production of optical transceivers and other silicon photonics products, Intel is providing the technology to others. Last March, China’s FAST Photonics Technologies announced plans to manufacture optical transcei

The large language models that enable generative artificial intelligence (AI) are driving an increase in investment and an acceleration of competition in the field of silicon photonics, a technology that combines silicon-based integrated circuits (ICs) and optical components to process and transmit massive amounts of data more efficiently.
Top-rank designers and manufacturers of ICs, AI systems and telecommunications equipment have all joined the race, including NVIDIA, TSMC, Intel, IBM, Cisco Systems, Huawei, NTT and imec, the Interuniversity Microelectronics Centre headquartered in Belgium.
These and other organizations have been working on silicon photonics for many years, some of them (including Intel and NTT) for nearly two decades.
On September 5, at the Silicon Photonics Global Summit in Taiwan, TSMC’s vice president in charge of pathfinding for system integration, Douglas Yu, told the Nikkei newspaper “If we can provide a good silicon photonics integration system … we can address both critical issues of energy efficiency and computing power for AI. This is going to be a new paradigm shift. We may be at the beginning of a new era.”
In its invitation to the Summit, semi (the microelectronics industry association) noted that “Silicon Photonics has become a prominent buzzword in the semiconductor industry” due to its high bandwidth, high-speed data transmission, extensive transmission distance, low power consumption and applicability to advanced networking, computing architectures, cloud computing, data centers, autonomous vehicles and smart transportation systems.
In other words, right across the leading edge, high-tech industry will benefit from silicon photonics in a way that will improve device and system performance while lowering energy consumption.
According to semi, “the global silicon photonics market is projected to reach a value of US$7.86 billion by 2030, with a compound annual growth rate (CAGR) of 25.7% from its valuation of $1.26 billion in 2022.”
About a year ago, in September 2022, DigiTimes reported that NVIDIA and TSMC had launched a joint R&D project called COUPE, which stands for Compact Universal Photonic Engine. The goal of the project is to combine multiple AI processors (GPUs) using NVIDIA’s silicon photonic (SiPh) technology.
“Source revealed,” wrote DigiTimes, “that SiPh chip and CMOS processes go through co-packaged optics (CPO) technology integration, which can connect multiple advanced GPUs with chip-on-wafer-on-substrate (CoWoS) 2.5D IC packaging.”
The combination of low-latency optical data transmission and advanced packaging technology results in a significant reduction in signal loss, making it possible to create “an ultra-large GPU set,” DigiTimes wrote.
This technology may not be ready “until the SiPh ecosystem matures,” which helps explain why Taiwan’s leading Outsourced Semiconductor Assembly and Test (OSAT) company ASE Technology and Japan’s Advantest were also prominent at the Silicon Photonics Global Summit.
At the Semicon Taiwan trade show, which followed the silicon photonics event, TSMC told the press that its ability to supply NVIDIA with AI processors is likely to be constrained until the end of next year by a shortage of CoWoS packaging capacity. By then, capacity should double.
Intel defines silicon photonics as “a combination of two of the most important inventions of the 20th century—the silicon integrated circuit and the semiconductor laser. It enables faster data transfer over longer distances compared to traditional electronics while utilizing the efficiencies of Intel’s high-volume silicon manufacturing.”
Intel explains that by combining “the power of optics and the scalability of silicon… optical transceivers are the optical interfaces for Ethernet switches, routers and transport networking equipment, providing connectivity for large-scale cloud and enterprise data centers.”
Dylan Patel of SemiAnalysis has written that “Intel manufactures silicon photonics at the largest scale in the world. They lead in market share for manufacturing optical networking transceivers… Intel’s scale and the integrated nature of their solution has led to them leading the industry with… [a failure rate] two orders of magnitude better than the competition.”
Intel plans to increase its advanced IC packaging capacity by about four times by 2025, including a new 3D packaging facility in Malaysia. In late August, Intel’s corporate vice president for manufacturing supply chain and operations, Robin Martin, told Tech Wire Asia that “Malaysia will eventually become Intel’s largest production base for 3D chip packaging.”
In addition to ramping up its own production of optical transceivers and other silicon photonics products, Intel is providing the technology to others. Last March, China’s FAST Photonics Technologies announced plans to manufacture optical transceivers based on Intel technology.
Fast Photonics’ headquarters and factory are located in Shenzhen. It also has a sales and technical support center in San Jose, California, near Intel’s headquarters in Santa Clara.
In May 2023, Japan’s national telecom carrier and telecommunications technology developer NTT announced plans to manufacture a series of increasingly sophisticated photonics-electronics convergence devices over the next several years.
These devices, which eliminate the need to convert signals from optical to electrical and back again, should open the way to a radical reduction in the power consumption of telecommunications networks and data centers.
NTT is aiming for a 100-times increase in energy efficiency, a 125-times increase in transmission capacity and a 200-times reduction in end-to-end delay (latency) in mobile and optical networks by 2030.
Along with Intel and Sony, NTT has established the Innovative Optical and Wireless Network Global Forum to “to accelerate the adoption of a new communication infrastructure that will bring together an all-photonics network infrastructure including silicon photonics, edge computing, and wireless distributed computing.”
Huawei and imec added silicon photonics to their joint research on optical data link technology in 2014. That followed Huawei’s acquisition of Caliopa, a developer of silicon photonics optical transceivers spun out of imec and Ghent University, the previous year.
But Huawei-imec collaboration was shut down after Huawei was put on the US Commerce Department’s Entity List and shipments of ASML’s EUV lithography systems to China were banned in 2019. ASML has a close relationship with imec.
Huawei has continued with its own research, which is important both for its telecom equipment business and its efforts to escape sanctions imposed by the US government. In a video posted on YouTube in October 2022, it declares that: “Huawei intends to produce a photonic chip in order to circumvent American chip restrictions.”
It goes on to say that photonic chips outperform silicon-based ICs, which will be nearly impossible to make at process nodes below 2nm… Furthermore, [the photonic] chip manufacturing process is completely different… It will use brand-new manufacturing technology and will not require a lithography machine at all…”
“The most important point,” the video concludes, “is that research in this field is still in its early stages and European and American countries have not yet established a monopoly.”
This assessment appears to look at least a decade ahead. While it is optimistic but plausible, and given the recent launch of the 5G Mate 60 Pro smartphone with its 7nm processor in spite of US sanctions, Huawei’s intent should not be discounted.
On August 23, Bloomberg published an article under the headline, “Huawei is Building a Secret Network for Chips.” According to the article, the US Semiconductor Industry Association (SIA) has warned that Huawei has acquired at least two fabs and is building three others, supported by $30 billion in Chinese government funding.
If the report is accurate, it would be folly to assume that Huawei is doing this simply to import standard semiconductor production equipment under different names. Huawei and China have every incentive to invest heavily in silicon photonics.
Follow this writer on Twitter: @ScottFo83517667