The addressable market for chip-level optical interconnects could eventually run into billions of units, as revenues rise to almost $520m by 2019 then $1.02bn by 2021, according to the report 'Revenue Opportunities for Optical Interconnects: Market and Technology Forecast – 2013 to 2020 Volume II: On-Chip and Chip-to-Chip' from industry analyst CIR, which continues the firm's coverage of this market dating back to 2009.
This follows a related report 'Revenue Opportunities for Optical Interconnects: Market and Technology Forecast – 2013-2020. Vol. I Board-to-Board and Rack-Based' issued by CIR in August.
The latest report covers four kinds of chip-level interconnect: optical engines, photonic integrated circuit (PIC)-based interconnects, silicon photonics and free-space optics. It includes nine-year (volume and value) forecasts with breakouts by active components along with fiber and waveguide transmission media. Compound semiconductor, silicon and polymer waveguides are covered, as are vertical-cavity surface-emitting lasers (VCSELs), silicon lasers and quantum dot (QD) lasers. In addition, the report contains assessments of the latest business and technology strategies in the chip-level optical interconnect space.
Companies discussed include Avago, Cisco, Corning, Dow Chemical, Dow-Corning, DuPont, Finisar, Fujitsu, Furukawa, IBM, Intel, Juniper, Kotura, Micron, Novellus, Optical Interlinks, QD Laser, Reflex Photonics, Samtec, Sumitomo, TeraXion, Tokyo Electron, ULM Photonics, and VI Systems.
The growing popularity of parallel computing, and the arrival of multicore processors and 3D chips, are leading to data traffic jams both on-chip and chip-to-chip, says CIR. However, these trends are also creating opportunities for chip-level optical interconnects, adds the report.
Avago, Finisar, IBM and Samtec have all proposed optical engines for chip-level interconnect. These miniaturized optical assemblies are currently the most mature technology available for this application and will generate revenues of $235m in 2019, CIR forecasts. However, with their attached connectors and heat sinks, optical engines may prove too large for complex optical interconnection environments, such as in the coming generation of Exascale supercomputers.
Meanwhile, the arrival of multicore processors and 3D chips means that computer power now depends on how fast each CPU can talk to each other and to memory devices. So reliable, low-loss, high-speed interconnects between chips then become crucial. Interconnect data rate requirements could reach hundreds of times what they are currently.
Because of the limitations of optical engines, there are emerging opportunities for compact PIC-based interconnect devices based on indium phosphide (InP) and gallium arsenide (GaAs). These opportunities will generate $120m in 2019, increasing to $275m by 2021, reckons CIR. However, bonding PIC interconnects onto a silicon processor or memory chip is both technically challenging and expensive. So far, only a few PIC and VCSEL technology companies have pursued the interconnect opportunity.
Although silicon photonics has compelling advantages, firms – especially Intel — have struggled for years to make active optical devices using silicon. A breakthrough in silicon laser technology would be the single most important development in optical interconnects, allowing the full integration of both electronic information processing and optical integration, says CIR. Faster VCSELs will also be important for the development of chip-level optical interconnect. Several firms and research institutes have announced high-speed VCSELs, operating all the way up to 55Gbps, although such lasers await extensive commercialization. Quantum dot-enhanced VCSELs have also been proposed and these, too, may have applications in interconnection, concludes CIR.
