100G Price Challenge Winning Metro 100G Daryl Inniss, PhD daryl.inniss@ovum.com ECOC 2013, Market Focus 23 September 2013 1 Copyright Ovum. All rights reserved. Ovum is a subsidiary of Informa plc.
Outline Executive summary History: 100G so far Why is metro so important? What are the new opportunities? Examples of winning strategies What about beyond 100? 2
Executive summary The 100G long-haul market is in good shape as the 100G DWDM line card has reached price parity with 10G for long-haul transmission and the bill of materials costs have tracked The 100G LR4 client is an exception it has not yet reached parity with 10G Metro is the next DWDM 100G market opportunity Metro demand is critically important because it drives up volume, drives down cost, and the solutions improve and ultimately support the long-haul market. Moreover, the metro solution is the platform for next-generation coherent transmission A completely different design is being developed for metro due to cost sensitivity Pluggable modules are being developed in the CFP and CFP2 form factor The transmitter, receiver, and ICs are all being redesigned CFP2 will likely be the volume market, but the products will take longer to develop Module with the DSP on the board is gated by the need for a robust high-speed analog link between the transceiver and the DSP chip Module with the DSP inside is gated by the need for a low-power-consumption chip Winning component strategies include vertical integration; photonic-integrated products; building volume; and solutions that work for both CFP and CFP2 3
Outline Executive summary, Ovum view, and Recommendations History: 100G so far Why is metro so important? What are the new opportunities? Examples of winning strategies What about beyond 100? 4
Component and long-haul line-card price declines track ~50% drop in 2 years 100G LH system cost has reached parity with 10 x 10G this drives volume Normal price declines resume Component costs track system costs NB: Scenario discussion only; not an official Ovum forecast 5
System vendors forward pricing strategy worked to launch the long-haul 100G market 100G is now cost-effective compared with 10G 6x and volume has accelerated 6
Metro places new cost pressures on the market This is the estimated price of the metro line card. The target is 50% of the price of the long-haul line card. How can components cost track metro line card? 7
Outline Executive summary History: 100G so far Why is metro so important? What are the new opportunities? Examples of winning strategies What about beyond 100G? 8
Metro solutions become pervasive Compared to long-haul metro is cost-sensitive metro has shorter distance requirements Components are specially designed to support metro requirements DSPs with lower dispersion compensation Modulators for shorter reach Photonic integration of transmitters and receivers to support the lower cost and shorter distances Multiple 100G line-card ports for lower system cost The high volume pushes the technologies designed for metro to become mature The metro solutions improve over time and are ultimately used for regional and long-haul applications 9
Metro can drive high volumes! Adjusted metro volumes assuming low prices Ovum s latest volume forecast (TE003-000570) shows higher metro volumes than LH but assumes high metro prices Low prices can drive metro volumes to levels 3 to 5 times greater than LH volumes 10
Metro technology is ultimately used in long-haul System designers want cost reduction but can be resistant to change DSP complexity increases resistance to changing 100G coherent Therefore, 100G metro design window is critical for introducing newer, better technology Technical advances can trickle up to LH, delivering smaller size, modularity, and cost reduction Over time, these new technologies can be brought to bear on LH For example, 10G DWDM is now in XFP/SFP+ these form factors and components were originally for datacom, then moved to telecom client-side, then telecom line-side LH Metro + LH 11
Metro components will help drive down long-haul cost All parts of the line card will see major changes that decrease both size and cost Many of these changes can be applied to LH 168-pin coherent module or equivalent Client CFP transceiver Coherent transmit electronics Coherent receive electronics DWDM transmit optics DWDM receive optics 5 x7 4 x5 4 CFP CFP 4 Coherent Tx+Rx IC Coherent Tx+Rx IC DWDM Tx optics DWDM Rx optics DWDM Tx optics DWDM Rx optics LH line card today Future LH line card 20?? CFP Coherent Tx+Rx IC DWDM Tx optics DWDM Rx optics CFP2 coherent module ( analog ) CFP Coherent Tx+Rx IC DWDM Tx optics DWDM Rx optics Target metro line card 2014 CFP coherent module ( digital ) 12
Don t confuse line-side CFP and CFP2 with client side, but common footprint expands coherent transceiver market Source: CFP-MSA.org Client or datacom module world Source: Molex CXP QSFP28 CFP CFP2 CFP4 Just one step in continuous migration to smaller size Telecom line-side world A different relationship between optics module and system vendor customer Merchant CFP2 In-house or partner Coherent Tx+Rx ASIC CFP2 Merchant DWDM Tx Optics DWDM Rx Optics CFP2 Coherent Tx+Rx ASIC CFP DWDM Tx Optics DWDM Rx Optics Merchant module vs merchant optics + in-house ASIC 13
Module choice is linked to ASIC for example, CFP2 with ASIC is best today as it can sit in any CFP2 slot A different relationship between optics module and system vendor customer Telecom line-side world But transition to module opens old wounds: who owns the ASIC? 14
Outline Executive summary History: 100G so far Why is metro so important? What are the new opportunities? Examples of winning strategies What about beyond 100G? 15
New opportunity exists for metro 100G component Pluggable coherent transceivers are being developed for the line side These will be the first pluggable coherent transceivers introduced to the market The CFP includes the optics and the DSP chip in the module There are two CFP2 options one with the DSP in the module and the second with the DSP on the board Low-power ASIC is required due to the low target CFP power consumption; an even lower-power ASIC is needed for the CFP2 High-speed analog link is required between the CFP2 transceiver and the ASIC sitting on the board. This will be the first coherent connector at these data rates Lower-cost client-side transceiver is needed to support the metro cost target The coherent transceiver module needs integrated transmit and receive optics to support the small size, low power, and low cost New technologies are being developed to support the transmit and/or receive optics 16
CFP path is clear, but there is uncertainty in the CFP2 solution CFP has the DSP in the module. Low power consumption is required. The DSP is most likely developed and owned by the module provider CFP2 has numerous potential paths DSP on the board is a solution supported by the OEMs as they control it. Availability is gated by the need for a high-speed connector to the module DSP in the module is supported by module vendors as they would control the optics and all the electronics. Availability is gated by the need for a low-power DSP and understanding of the application space it can support History teaches us that the lowest-cost solution includes the IC on the board and the optics in the module. We cannot discount that coherent may usher in a new era, with the lowest-cost solution being a common transceiver form factor suitable for all client- and line-side applications 17
CFP and CFP2 have different start dates and volume outlooks CFP Available sooner, may use existing components Limited addressable market size Contains DSP chip, so may not serve customers who have own DSP already Pursue module as incremental revenue if you already have components that fit CFP2 Requires new components High-speed analog electrical connection needed between DSP and module Addresses entire customer base Leading proponents are customers with own DSP (which do not buy modules today) Caution: these customers may still want discretes, not modules Addressable application space to be determined Limited reach due to lower power, smallest modulator All new component development should be CFP2-capable, since it can address the CFP, CFP2, and discretes opportunities with the same design 18
Metro component challenges and opportunities ASIC needs very low power for CFP, but same design should support CFP2 and ideally LH CFP2 CFP2 Coherent Tx+Rx ASIC Coherent Tx+Rx ASIC CFP2 requires high-speed connector, line drivers, and very good TIA in Rx CFP CFP2 DWDM Tx Optics DWDM Rx Optics DWDM Tx Optics DWDM Rx Optics CFP2 optics can be used in CFP but not vice versa Client LR4 CFP has not reached parity with 10G CFP LR4 is same price as DWDM non-coherent LR4 Introduction of CPAK splits nextgen (CFP2) market volume 19
DWDM transmitter optics technology changes are in progress Semiconductor modulator instead of LiNbO InP and silicon modulators are being considered What reach is possible? Cost savings Laser integrated with modulator Local oscillator and transmit laser combined Transmit optics Tunable CW laser 4 modulators Integrated Tx Transmitter laser to receiver to serve as local oscillator 20
DWDM receiver optics technology changes are also in progress Includes TIA, which strictly speaking is electronics TIA needs bleeding-edge performance due to intimate connection to coherent DSP Additional dynamic range needed for colorless ICR (integrated coherent receiver): higher index waveguides for size also helps wafer cost but hurts link budget 2 TIAs 4 PDs 90-deg optical hybrid Receive optics PBS Rx Waveguides being redesigned to fit in CFP and CFP2 2 TIAs 4 PDs 90-deg optical hybrid BS Transmitter laser for local oscillator 21
Electronics technology changes in progress Coherent DSP: new metro design is opportunity to move to next silicon process, but cost needs to be considerably reduced to justify re-spin Mux already moved from SiGe to CMOS. Next: integrate mux with receiver DSP, whenever yield/process maturity allows Modulator driver: GaAs InP SiGe? Connector, line drivers, etc. to maintain signal integrity for high-speed analog connection is a new requirement Mux Transmit electronics 4 drivers Receive electronics Demux DSP A/D A/D A/D Combine mux and demux in CMOS A/D 22
Silicon photonics competes with InP but no clear winner InP is competitive with LiNbO in the pluggable modules due to its small size; silica receiver struggles with CFP/CFP2; Si getting stronger over time Where Feature Glass LiNbO InP Si ICR Modulator Integrated PD PBS integrated Coupling efficiency No n/a yes ~Yes (Ge) Yes n/a no Yes? Good if low index delta n/a Good with 3D taper Maturity n/a Incumbent Already available Integrated laser n/a No Yes, not always? Research Research 23
Outline Executive summary, Ovum view, and Recommendations History: 100G so far Why is metro so important? What are the new opportunities? Examples of winning strategies What about beyond 100G? 24
Consider winning strategies to attack metro Identify the correct end-goal, play the long game, and drive volumes Nortel invested in coherent DSP ~2003 while the rest of industry cut costs. Ciena still benefits from technology lead 10 years later JDSU, the dominant tunable XFP supplier, acquired Agility in 2005 specifically to make tunable XFP while other vendors tried to move up the food chain Tunable XFP market Acquisition Recognized that OEM customers like transceivers, even for line side Desirable product able to command premium despite cheap form factor Currently pursuing coherent CFP2 First to market, high barrier to entry 25
Other examples of winning strategies Identify what is critical and become the best at it InPhi working on high-speed TIA (40G) in 2000. QPSK requires four TIAs; nextgen 16-QAM will require even more Kyocera high-speed package CFP2 connector and line drivers: who will do? Recognize this may mean accepting a role relative to generational life cycle. Be ready to trade up Semtech dominated mux (transmit electronics) market in early stage at 40G, 100G with SiGe part. Now 100G in CMOS, SiGe is moving on to 400G Fujitsu Optical Components LiNbO3 modulator at forefront of performance from 10G to 40G multi-level, then to 100G coherent. As InP/Si moves into 100G metro/lh, LiNbO3 stays incumbent in ULH 100G and also moves on to 400G Aggregate volumes across applications 26
Outline Executive summary History: 100G so far Why is metro so important? What are the new opportunities? Examples of winning strategies What about beyond 100G? 27
Cost per line card Consider coherent era is just beginning 400G Metro, LH? 200G LH 100G LH 100G Metro 100G LH 1 port per card 2 ports 100G Metro 4 ports 2013 2014 2015 2016 28
Summary Long-haul market is in good shape 100G DWDM line card price has reached parity with 10G for long haul Bill-of-materials costs have tracked 100G LH system costs Exception: 100G LR4 client has not yet reached parity with 10G Metro is a life cycle stage 2014 is the year for new cost-reduced designs motivated by metro demand These cost reductions will help with both metro and long-haul Coherent CFP2 is more than a form factor between CFP and CFP4 As a pluggable module it can address a bigger market The best long-term component investments work for both CFP and CFP2 Examples of winning strategies Pick the winner, do not game the system The long game includes coherent for multiple applications 29
Thank you! 30 Copyright Ovum. All rights reserved. Ovum is a subsidiary of Informa plc.
Disclaimer All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the publisher, Ovum (an Informa business). The facts of this report are believed to be correct at the time of publication but cannot be guaranteed. Please note that the findings, conclusions and recommendations that Ovum delivers will be based on information gathered in good faith from both primary and secondary sources, whose accuracy we are not always in a position to guarantee. As such Ovum can accept no liability whatever for actions taken based on any information that may subsequently prove to be incorrect. 31