Yes, we’ve reached Si photonics’ tipping point!
SILICON PHOTONICS TECHNOLOGY HAS REACHED ITS TIPPING POINT, WITH TRANSCEIVERS SHIPPING IN VOLUME!
Fueled by increasing internet traffic thanks to social networks, video and gaming content, increasing bandwidth will be required for inter- and intra-data center communications. As a short term answer to this possible bandwidth congestion, new photonics technology for transceivers is still pulled onto the market by Google, Apple, Facebook, Amazon and Microsoft (GAFAM) and now by Baidu, Alibaba and Tencent (BAT) in China. Indeed, in China data centers are expanding fast, with over 650 million Internet users and three big operators setting up their own fiber networks. In 2018, new hyperscale data centers will be set up in China to expand the growth of optical components, new installed fiber and optical technologies. In US and Europe, big names like Facebook, Google, and Amazon are developing their own optical data center technology in partnership with chip firms, such as Facebook and Microsoft’s partnership with Intel. While traffic continues to increase between users and data centers across the internet, more and more data communication is taking place within the data center. And with existing data center switching and interconnects making it difficult and costly to cope with this increasing flow, new approaches are necessary. As a semiconductor-based technology, silicon photonics’ cost will drop with increasing volume thanks to economies of scale. Silicon photonics components will therefore answer data centers’ requirements in terms of low cost, higher integration, greater embedded functionality and higher interconnect density, lower power consumption and better reliability compared with legacy optics. And more players are entering this industry’s landscape (see figure). After 16 years of development, Intel has successfully penetrated this application and shares the current market with its leader Luxtera. We also see new startups and more products reaching the market, mostly for 100G data rates, but soon for 400G. Besides Luxtera, which did well in 2016 in inter-data center links, Acacia Inc. had success with its single-chip silicon photonics 100G coherent transceiver for metropolitan applications, opening the way for silicon photonics in telecoms.
MARKET WILL BOOM FOR 100G THEN 400G
Silicon photonics is still a small market today, with sales at die level estimated to be $30M in 2016. However, it has big promise, with a 2025 market value of $560M at chip level and almost $4B at transceiver level. It means that silicon photonics technology will grow from a few percent of total optical transceiver market value in 2016 to 35% of the market in 2025, mostly for intra-data center communication. The strongest demand is for 400G, and 200G could be only an intermediate step between 100G and 400G. The next evolution is to develop a 400G optical port over a single fiber across 500m at less than $1 per gigabit and with power <5mW/Gb. One terabit per second rates should follow. Although the wafer area this accounts for will be a minute part of the worldwide silicon-on-insulator (SOI) market, it will represent significant value because of the SOI wafer high price. The main supplier is SOITEC.
WE ARE AT THE VERY BEGINNING OF MASSIVE DEPLOYMENT
We believe we are only at the very beginning as there is massive ongoing development worldwide for further integration. The recent involvement of large integrated circuit foundries, such as TSMC’s relationship with Luxtera, and GlobalFoundries with Ayar Labs, are very encouraging signs showing the big promise for silicon photonics. STMicroelectronics is the major silicon photonics manufacturer today. The “Zero-Change” processes currently in development, manufacturing optical components without making any changes to a CMOS process, are targeting future inter-chip optical interconnects that could represent huge market volumes. Silicon photonics is at the maturity level of the electronics industry in the 1980s and there are still challenges to overcome. For all these challenges, technical breakthroughs will be necessary (see figure).
- Laser source: For low-cost, low data rate <50Gb/s and
- We need smaller modulators: cheaper, better peakto- peak driver performance, less power consumption.
- Alignment and packaging, wafer-scale testing.
- Design and software: Photonics software is very specialized and very custom-designed, with predefined models.
- Supply chain.
- Manufacturability: A new trend is to have a zerochange approach on a CMOS line.
FROM DATA CENTERS TO NUMEROUS OTHER APPLICATIONS!
Besides data centers, some other interesting applications could follow. High performance computing most generally refers to the practice of aggregating computing power in a way that delivers much higher performance than one could get out of a typical desktop computer or workstation in order to solve large problems in science, engineering, or business. Due to the accelerated growth in performance of microprocessors and the recent emergence of chip multiprocessors (CMP), the critical performance bottleneck of high-performance computing systems has shifted from the processors to the communications infrastructure. So, by exploiting the parallelism and capacity of wavelength division multiplexing (WDM), optical interconnects offer a high-bandwidth, low-latency solution that can address the bandwidth scalability challenges of future computing systems. Optically-connected memory systems can enable continued performance scaling through high-bandwidth capacity, energy-efficient bit-rate transparency, and time-of-flight latency. Telecom is also one possible application for silicon photonics, especially metropolitan networks from 80km to 1,500km in length. For example, Acacia Communications is developing coherent-based silicon photonics modules for telecom applications. While coherent was first adopted in long haul, Acacia’s coherent technology reduces polarization dispersion and losses and reduces channel passbands. There is also interest for fiber-to-the-home (FTTH), especially second-generation gigabit passive optical networks (GPON2). Silicon photonics may be useful as FTTH moves from 10G to 25G, because more functionality will be needed, such as wavelength tuning and switching. Silicon photonics is also an interesting platform for sensing applications in biomedical applications and sensors. Future autonomous vehicles will require a high level of security with very low latency for high reactivity speed. Lidar is also an attractive application for silicon photonics, with IMEC developing the necessary technology, and with some R&D projects already having started. For aerospace/aeronautic applications where size and weight are critical parameters for reducing power consumption, silicon photonics is also promising.
Objectives of the Report
- Examine the current status and future challenges for data centers
- Explain why silicon photonics is the answer to future datacenter needs and other possible applications
- Offer 2016-2025 market forecast for silicon photonics in US$M, units, and wafers
- Provide updated market shares for silicon photonics players
- Give a description of the industrial supply chain, with player status
- Present silicon photonics’ technical challenges
- Updated silicon photonics forecast in units, value and wafers for 40G, 100G, 200G and 400G for 2016-2025
- Wavelength division multiplexing (WDM) versus parallel
- Updated 2016 market shares for silicon photonics suppliers
- Supply chains
- Technology and process trends
- Application descriptions: Data centre, HPC, telecom, automotive, medical, sensors, aeronautics/aerospace