A bright future for the global PIC market
The latest report from IDTechEx predicts that the global PIC market will see robust growth, reaching $22 billion by 2034. But which applications are likely to grow the most? And what commercialisation challenges do PICs still face?
By James Falkiner, Technology Analyst at IDTechEx
From sending and receiving billions of bits of information in a package the size of a candy bar to detecting different molecules in the air with remarkable sensitivity, PICs are a powerful technology that can be adapted for a wide range of uses. These tiny optical systems, made of materials such as silica (glass), silicon, and indium phosphide (InP), have already experienced commercial success in telecoms and datacoms, and show plenty of promise for future growth.
By leveraging the billions of dollars of investment in CMOS chips, for instance, PICs can unlock new potential for scaling information processing beyond Moore’s law.
However, there remain significant challenges for the PIC market, such as material limitations, integration complexity, and cost management. Large demand volumes are required to offset the high initial costs of designing and manufacturing PICs, while production lead times can take months.
IDTechEx’s new PIC report thoroughly investigates the market for this technology and identifies photonic transceivers for AI as an emerging segment that will soon be the largest source of demand for photonic chips. This is because PICs can overcome some of the challenges faced by traditional optical transceivers, including data limitations associated with bandwidth, modulation speeds, and noise. PIC-based transceivers can thus send more data in a similar form factor.
Figure 1 shows a traditional pluggable optical transceiver on the left with large independent transmitter optical sub-assembly (TOSA) and receiver optical sub-assembly (ROSA) circled in orange. On the right is a PIC-based transceiver of a similar size, but which offers 10 times the performance thanks to the large PIC-based TOSA under a heatsink. Under that heatsink is a PIC that controls the transmitting laser (bonded to the PIC), modulator systems, and passive optical components used for wavelength-division multiplexing (WDM), a method used to increase the data rate.
Figure 1: A traditional pluggable optical transceiver (left) compared with a PIC-based transceiver (right). The PIC-based transceiver offers around 10 times the performance of the traditional transceiver, despite being a similar size, thanks to a transmitter optical sub-assembly (TOSA), which is situated under a heatsink and contains passive optical elements that increase the data rate through wavelength-division multiplexing.
Materials on the horizon
Although silicon photonics has generally been the most widespread platform for commercial PIC devices, researchers in both academia and industry are investigating alternative materials for various applications.
As part of our analysis of the PIC market, IDTechEx has evaluated some emerging and popular materials. We compare important performance metrics, such as cost, scalability, losses (power lost per metre), and modulation performance. The latter is defined as how quickly a material changes when exposed to an electric field, and is a critical factor for high-performance transceivers.
Although most of the current market uses silicon- and silica-based PICs for light propagation, silicon, as an indirect semiconductor, is not an efficient light source or photodetector. Therefore, it is usually combined with III-V materials, like InP, for photon emission and photodetection functionalities.
Silicon’s market dominance looks set to continue. However, thin-film lithium niobate (TFLN), with its moderate Pockels effect and low material loss, is emerging as a strong contender for applications that require high-performance modulation, such as quantum systems and potentially also future high-performance transceivers. Monolithic InP continues to be a major player due to its light detection and emission abilities. Additionally, innovative materials like barium titanate (BTO), electro-optical polymers, and rare-earth metals are being explored for their potential in quantum computing and other cutting-edge applications.
AI impacts
The rise of artificial intelligence (AI) has spurred an unprecedented demand for high-performance transceivers capable of supporting the massive data rates required by AI accelerators and datacentres. Silicon photonics and PIC technologies are at the forefront of this revolution, with their ability to transmit data at speeds of 1.6T and beyond.
As shown by Nvidia’s latest Blackwell CPUs, which, according to our research, require approximately two 800G transceivers per GPU, the need for efficient, high-bandwidth communication is becoming more critical for AI, positioning silicon photonics and PICs in general as essential components in the AI-driven future. Indeed, the biggest driver of PIC-based transceiver development is AI, since higher-performance AI accelerators will require higher-performance transceivers, with 3.2T transceivers expected to arrive by 2026, as indicated in Figure 2. Based on past trends, we forecast that future AI accelerators will require two optical transceivers per accelerator on average on launch, with in-generation transceiver improvements resulting in a lower transceiver per accelerator ratio throughout the estimated eight-year lifespan of an accelerator.
Figure 2: A roadmap for PIC-based pluggable transceivers. 3.2T transceivers are expected to arrive by 2026.
The AI accelerator market is set to see significant short-term growth with the H100 and B100 GPUs, alongside Google, Microsoft, and Amazon’s upcoming dedicated products. Beyond 2026, IDTechEx forecasts that the market will not see the same level of demand that we see in 2024, due to a slowdown in datacentre growth, and satisfaction with and consolidation of the current emerging AI and large language model (LLM) market.
In the short term, we expect to see rapid growth in global datacentre numbers. However, datacentre spending is set to slow beyond 2026 as datacentres start to satisfy global AI demand. Beyond 2030, datacentre performance and capability will be improved through the renovation and refitting of existing datacentres with faster AI accelerators, satisfying AI demand and leading to reduced numbers of new datacentres.
According to our analysis, datacentres already use about 1 percent of the world’s energy, so there is a limit to how much they can grow before they start running into wider electrical infrastructure limitations. Datacentre growth is likely to have a significant impact on the PIC market, as PIC-based transceivers are a critical component for transferring data to, from, and within this infrastructure. IDTechEx forecasts that by 2034, there will be over 15 000 datacentres globally.
Emerging applications
Silicon photonics and other kinds of PICs are being used to innovate new solutions in a wide range of fields, as illustrated in Figure 3. IDTechEx has explored several of the most prominent applications and the advantages they can offer within their respective industries.
First, within the area of photonic engines and accelerators, photonic components such as Mach-Zehnder Interferometers and low-loss waveguides can be used to design and manufacture high-performance photonic processors and programmable PIC devices. This could unlock higher performance than is possible with electronic accelerators alone. These photonic accelerators can perform certain mathematical functions extremely quickly, potentially having applications in machine learning and AI matrix operation.
In the sensor market, PIC materials, such as silicon nitride, can enable a range of different technologies, from gas sensors to “artificial noses.” The healthcare sensor industry may be able to take advantage of the miniaturisation of optical components into PIC devices, which could see applications in point-of-care diagnostics or wearable tech. Meanwhile, PIC-based frequency-mode continuous-wave (FMCW) LiDAR has the potential to transform the automotive and agricultural industries, with applications in drones and autonomous vehicles. As the quantum revolution continues to unfold, companies investing in quantum computing based on trapped ions and photonic qubits are looking to PICs for more stable and scalable quantum systems. PICs are used in photonic quantum systems to achieve the precise control of photons necessary for quantum computation.