Previous Page  14 / 76 Next Page
Information
Show Menu
Previous Page 14 / 76 Next Page
Page Background

12

Chip Scale Review January • February • 2017

[ChipScaleReview.com]

including arrayed waveguide grating

(AWG), and transceivers [3]. Then in

2006, variable optical attenuators (VOA)

were commercialized by Kotura, which

has since been acquired by Mellanox.

Today, the number of companies

shipping SiPh products has increased

wi t h Lux t e r a , Ko t u r a /Me l l anox ,

Cisco/Lightwire, Intel, Acacia and

STMicroelectronics. We are also seeing

new startups and more and more products

reaching the market, mostly for 100GB,

but soon for 400GB. The rise of startups

in the field is also a very encouraging sign

of growing investments from the venture

capital (VC) community—and we all

know how tough a domain photonics is for

any startup.

According to our research, today’s

chip market value is small: estimates

were less than US$40 million in 2015.

But analysts believe it holds great

promise, with a market value of US$1.5

billion in 2025 for chips, and more than

US$6 billion in 2025 for transceivers [3].

The horizon, however, is not so clear,

as there are still challenges to overcome:

• Lase r sour ce i n t egr a t i on and

competition with vertical cavity

surface emitting lasers (VCSELs)

(actually SiPh’s best bet in the

500m–2km range);

• The photonics supply chain has

yet to be formed, with possible

oppo r t un i t i e s f o r ou t sou r ced

s emi conduc t o r a s s emb l y and

test suppliers (OSATS) and the

emergence of SiPh foundries

(

Figure 1

); and

• As always in optics: packaging and

assembly.

Compared to semiconductors, where

packaging accounts for 10% of the final

die cost, packaging and assembly is

generally 80-90% for a photonic die [4].

Fiber alignment, thermal management

and non-standardization mostly account

for the large percentage.

Though SiPh has the advantage of an

integrated circuit’s (IC) infrastructure,

function integration, low manufacturing

cost and high density, packaging still

contributes to a large portion of the cost.

Will the game end here?

First off, SiPh uses silicon as the

medium for optical signals, allowing

much faster digital signaling than is

currently possible with traditional

electron-based semiconductor devices.

It is a disruptive technology meant

to achieve a new breed of monolithic

optoelectronic devices for a potentially

low-cost Si process, with the ultimate

goal of delivering optical connectivity

everywhere, from the network level

to chip-to-chip. As SiPh evolves and

chips become more sophisticated, Yole

expects to see the technology used

more often in processing tasks, such

as for interconnecting multiple cores

in processor chips to boost access

to shared cache and buses. As more

and more bandwidth will be required

in data centers (representing 75% of

total data flow), the technology will be

necessary for 100GB and 400GB.

Secondly, silicon photonics is a mix

of several technical blocks: optical, IC

for processing, MEMS for packaging,

copper pillars, through-silicon vias

(TSVs) and more. Advanced packaging

developments, such as interposers,

copper pillars and TSVs (used as

light guides and no longer as electrical

interconnects), could be the solution

to more integrated chips. In terms

of fiber connections, there is also a

clear need for technology to move to

mass production at a low cost. New

technologies for interconnects with

integrated optics could be a viable

solution. Laser integration also needs

to be considered; some companies

(including Kaiam, Luxtera and Kotura/

Mellanox) have developed very clever

laser assembly solutions that will be

beneficial in the short term. Other

solutions using epi bonding (from

Leti, Intel and Aurrion) are also viable

solutions for the future.

Yole’s analysts believe that the

growing need for optical interconnects

for data centers will pave the way for

new developments to achieve lower

cost and higher integration. This in

turn will accelerate mass-volume

production. Innovative approaches

will come from developments in the

field of advanced packaging, such

a s i n t e r pos e r s . Th i s cou l d so l ve

the current challenges surrounding

heterogeneous integration.

On the business side, Yole sees

oppo r t un i t i e s f o r t h e pho t on i c s

supply chain to consolidate—the way

the IC supply chain has today. In

particular, OSATS could play a key

role by offering dedicated photonics

application services, for example, small

packaging houses.

In the end, new packaging and

assembly developments will allow more

functionalities to be integrated into a

single package, similar to what we are

seeing in the industry with the system-

in-package (SiP) trend. Are we seeing

the rise of the photonic-in-package era?

Yole’s analysts think so!

List of references

1. Silicon Photonics for Data Centers

and Other Applications 2016 report,

Yole Développement, October

2016.

2. Silicon Photonics for Data Centers

and Other Applications 2016 report,

Yole Développement, October

2016.

3. Silicon Photonics for Data Centers

and Other Applications 2016 report,

Yole Développement, October

2016.

4. Silicon Photonics for Data Centers

and Other Applications 2016 report,

Yole Développement, October

2016.

Biographies

As Sr. Technology & Market Analyst

of the MEMS & Sensors team at

Yole Développement, Eric Mounier

performs numerous technical and

market analyses focused on MEMS

and sensors, visible and IR imagers,

semiconductors, printed electronics and

photonics, including silicon photonics;

Emai

l: mounier@yole.fr

Thibault Buisson serves as Business

Unit Manager of Advanced Packaging

& Semiconductor Manufacturing

activities at Yole Développement;

Email

: buisson@yole.fr