BroadPak as key provider of innovative total solutions for 2.5D/3D products

Initially present in niche markets like MEMS devices, 3D integration is entering in a new era. Fueled by increasing bandwidth needs for moving data in cloud-computing and supercomputing applications, performance-driven markets have adopted 3D stacked technologies in a row.

New functionalities such as voice/image recognition, language processing… all these new needs directly contribute to the development of deep learning, datacenter networking, AR/VR, and autonomous driving applications, all based on 3D integration technologies.

According to Yole Développement’s “3D TSV and 2.5D Business Update – Market and Technology Trends 2017”, the number of 12” equivalent wafers will increase at a CAGR of 20% over the next five years, going from 1.3M in 2016 to 4M in 2022.

Yole Développement had the opportunity to speak with Farhang Yazdani, CEO of BroadPak. We talked about the company’s services, position in the industry and outlook on the market. The discussion follows.

Yole Développement: BroadPak provides 2.5D/3D integration technologies, could you tell us more about your current activity and BroadPak Corporation business model?

Farhang Yazdani: As a 2.5D/3D Integrator, we provided total solution and technologies to help our clients develop and launch 2.5D/3D products, basically, a one stop shop for all your 2.5D/3D needs from concept to delivery. With our vertically integrated turnkey model we deliver 2.5D/3D products to our customers, as far as we know, our cost model is by far one of the most competitive in the industry.

We are providing solutions and technologies for our customers in the machine learning industry, in particular for the deep learning applications. BroadPak is also very much active in enabling heterogeneous IP chiplets based integration. We provide support for the IP providers by guiding and helping them to prepare 2.5D/3D version of their embedded IPs for interposer based modular integration. For many of our clients, we provide interposer architecture and manufacturing services which include fine pitch Cu pillar bumping, both at the die level and interposer level. We provide interposers with Cu Damascene layers as well as plating layers on both 200mm and 300mm wafers.

We are known in the industry for our end-to-end co-design methodology and signal/power integrity expertise. In particular, we provide partitioning schemes for 2.5D/3D integrations as well as system level pathfinding and co-optimization. Our unique cross-domain co-design methodology integrates multi-physics simulation and analysis of electrical-thermal-mechanical-stress-reliability to produce high yield products. This fully integrated multi-physics co-simulation has proved to be necessary for successful 2.5D/3D integration.

Full listing of our 2.5D/3D services is available on our website at 

PathFinding BroadPak Yole 2017

Path finding (Courtesy of BroadPak)

YD: You, Farhang Yazdani, founded BroadPak Corporation in 2007 which makes you a pioneer player in the 2.5D/3D integration field. When you look in the mirror, how would you describe those last 10 years in 2.5D/3D field?

FY: It’s been very rewarding to see 2.5D/3D coming to life and proliferate in various markets and applications. At the time, it was clear that limitations in memory-logic bandwidth was to become a show stopper, similarly, availability of silicon proven IPs in specific node, including challenges in IP integration was causing a lot of delays, thus, the need for 2.5D/3D heterogeneous type integration was eminent. I founded BroadPak on the foundation of 2.5D/3D heterogeneous integration. When the cost per transistor increased below 22nm node, we had a surge of inquiries about the 2.5D/3D options and remedies. Looking back 10 years ago, it is rewarding to see that our prediction back then and the investment has paid off.

YD: 3D IC technology has known ups and downs, how hard was it to convince device makers of TSV-based package benefits?

FY: At the very beginning the responses that we received was like, “who else is doing it?” and then there were excuses on the fragmented supply-chains and costs, including lack of field proven reliability data and ESD requirements and tests. Once convinced of the benefits of the modular integration and after resolving all concerns, we found out that customers had no expertise in designing for 2.5D/3D integration and were very much afraid of the risks and the outcome, it was a new territory for them. So, we had to mentor, guide and work interactively with the customers on all aspects of 2.5D/3D flow.

YD: Your customers are large companies but also start-up. When start-up come to you with a project, they might not be much aware of the opportunities going with 2.5D? What does it look like to teach and being an ambassador of 2.5D?

FY: One of the reasons that many customers come to us is for brainstorming and to investigate the optimal 2.5D/3D solution and options. One of our greatest asset is to guide and offer a competitive solution to our clients. During the initial joint brainstorming process, startup clients often realize a new architectural approach, including new markets and opportunities. It takes team work among experts from both side of the aisle to propose a competitive solution.

Over the years we have realized that customers seeking 2.5D/3D path need intimate support and comprehensive solution spanning all the way from IPs to co-design/partitioning and more importantly the right supply chain. Our service offering includes determining the optimal 2.5D/3D solution and final delivery of the 2.5D/3D products to our clients. Many other customers also come to us for our very competitive and low cost interposer manufacturing and bumping services.

Pathfindingsandcost BroadPak Yole 2017

Path finding and cost optimization across multiple devices and packages (Courtesy of BroadPak)

YD: You offer turnkey solutions from design and IP to manufacturing and assembly, how important it is for customers to being offered a complete solution?

FY: Unlike conventional products which involves packaging of one or more monolithic die(s), a complete turnkey solution is mandated for launching a successful 2.5D/3D product. Many aspects of 2.5D/3D integration are highly interrelated, thus a total turnkey solution is needed to mitigate risks and to produce better performance, cost and yield. A profitable 2.5D/3D products start with smart planning, partitioning and pathfinding which bridges the design, foundry, assembly, bumping, basically the entire supply chain. A segmented supply chain for 2.5D/3D products is highly unlikely. Many of the successful 2.5D/3D products are launched by a single large corporate which owns and manages the entire flow, including IPs, design, supply chain and logistics. BroadPak offers the same turnkey solution similar to large corporates for the rest of the market.

YD: Major TSV foundries are blue chips companies, how hard is it for start-ups and mid-sized companies to get access to the 3D TSV manufacturing supply chain and convince big players to offer services for limited volumes? Do you believe that mid-sized TSV foundries have a bright future ahead?

FY: In the absence of high volume products, it’s been very difficult for start-up and mid-sized companies to get access to interposers and the associated supply chains. It is no secret that 2.5D/3D supply chains are very selective and difficult for everyone to get access to. Tier-one fabs only support their tier-one customers with high volume, they are not set up to support startups. Not that every TSV based product requires a large foundry, in fact, smaller and mid-sized TSV foundries with the right expertise in TSV manufacturing and Damascene patterning capabilities are vital and play an important role in the development of TSV based products. Often, many of the TSV based products with low to mid volume can very well be supported by smaller and mid-sized foundries.

YD: In the 2017 update of 3D TSV and 2.5D business report, Yole has reviewed the applications where 3DIC and 2.5D make the difference. Among them, machine learning using deep learning algorithms is the big trend where it seems that 3DIC and 2.5D have found a sweet spot. It seems that deep learning and 2.5/3D hardware have benefited from each other. Supercomputing using deep learning has been made possible only because high bandwidth hardware is available and TSV-based hardware were put on stage because of this new emerging field. Do you see machine learning and high performance computing as sustainable markets segments for 3DIC and 2.5D technologies? Is it a market of interest for BroadPak Corporation?

FY: In a connected world, managing big data and cloud computing is indeed on the rise. It so happens that for deep learning applications, in particular, during the training stage massive memory bandwidth is required. Many of the deep learning platforms to date use HBM2 and logic on an interposer. BroadPak is very active in the deep learning markets and its applications and is providing customized deep learning platforms for many of our customers. We see machine learning and high performance computing markets not only as sustainable, but also a market with considerable growth.


 (Source: 3D TSV and 2.5D Business Update – Market and Technology Trends 2017, May 2017, Yole Développement)

YD: Yole estimates the total TSV-based wafers market for high-performance applications around 230 000 12’’ wafers in 2017. Almost half of them are aimed for cloud computing in datacenters, networking industry and super big datacenters seem to be primary beneficiaries of 2.5D integration. What are your perspectives on 2.5D integration for the networking industry? Is it a market of interest for BroadPak Corporation?

FY: Cloud computing, super big datacenters and networking industries are the traditional markets where 2.5D/3D integration proved to be successful. These industries heavily depend on the low power, high bandwidth memory and logic integration. It is clear that the connected world will further increase the demand for these markets and along with it 2.5D/3D type integration will be required to scale up the networking and communication infrastructures. BroadPak has been very active in these markets and will continue to play an active role in enabling a wide range of customers in these markets.

TSV wafer forecast by market segments


(Source: 3D TSV and 2.5D Business Update – Market and Technology Trends 2017, May 2017, Yole Développement)

YD: TSV manufacturing is now quite well controlled by manufacturers. The consortium JEDEC is now putting on tracks High Bandwidth Memory 3rd generation (HBM3) to feed needs. Are there some remaining elements that could break or block the 2.5D and 3D IC take-off?

FY: Certainly demands for the 3rd generation of HBM is reflective of the ever increasing demand for memory in high performance computing and networking industries as well as the emerging deep learning applications. Similar to its predecessor, 2.5D/3D interposer based approach will enable HBM3 and logic integration. Traditional 2.5D approach will continue to evolve to better serve such integration, with advances in smart partitioning and low power technologies 3D IC will find its space in these evolving markets.

The remaining element that would further solidify 2.5D interposer based integration for the long run is the emerging demand for IP chiplets for the purpose of IP reuse and cost reduction. Heterogeneous IP chiplets based integration is expected to considerably lower the cost of the SOC and will enable a wider market adoption of 2.5D approach.

YD: In parallel OSATs and IDMs work on developing their proprietary TSV-less packaging technologies to compete with silicon interposer package. Companies like TSMC will offer at the same time CoWoS package and high density InFO (Fan-Out WLP). How do you see the differentiation between silicon interposer and TSV-less packages?

FY: Although both technologies are different in terms of physical structure, but silicon interposer and TSV-less packaging is to support their respective markets and applications. One of the major brings about of the TSV-less technologies were to support the fan-out and to reduce the cost of the TSV in interposer based integration. However, over the years the cost of the TSV has considerably diminished and nowadays, more foundries are capable of manufacturing TSVs, on the other hand, the competitive patent landscape of TSV-less approach is expected to further push the usage of the TSV approach for its targeted applications. The other factors are the markets and applications, for example the high performance computing and networking market is not so deterrent from slight variations in cost. Thus, the TSV approach is expected to stay in the long run and is to complement the TSV-less approach, a hybrid integration of the both technologies are to enable a wider markets and applications.

TSV-less interposer & advanced high density substrates

TSV less interposer advanced high density substrates

(Source: 3D TSV and 2.5D Business Update – Market and Technology Trends 2017, May 2017, Yole Développement)

YD: BroadPak Corporation seems to be well positioned to benefit from taking of 2.5D business emergence. How do you see the future of BroadPak Corporation in the next 5 years? What major evolution do you see in the future?

FY: BroaPak will play a critical role in the development and proliferation of 2.5D/3D heterogeneous IP chiplets based products. In addition, BroadPak will continue to play a critical role in the cloud computing and machine learning markets by supporting a wide range of customers as well as providing mid-range and high-end platforms for customized deep learning applications. BroadPak’ s one-stop 2.5D/3D service center will continue to support worldwide clients and evolve with the market requirements and applications. We will see an evolution in the future of SOC development entirely based on heterogeneous IP chiplets integration.


Farhang BroadPakFarhang Yazdani is the President and CEO of BroadPak Corporation. His company provides total solution and technologies to help its clients develop and launch 2.5D/3D products. Through his 18 years with the industry, he has served in various technical, management, and advisory positions with leading semiconductor companies worldwide. He has numerous publications and IPs in the area of 2.5D/3D Packaging and Assembly, serves on various technical committees and is a frequent reviewer for IEEE Journal of Advanced Packaging. He received his undergraduate and graduate degrees in Chemical Engineering and Mechanical Engineering from the University of Washington, Seattle.

Sources:  /



3D TSV and 2.5D Business Update – Market and Technology Trends 2017
3D integration, a leading-edge technology supporting high-performance applications – from imaging to artificial intelligence, an entire ecosystem in motion – More  



BroadPak will be part of a Yole Group of Companies webcast, held on Sept. 7. Focus will be 2.5D/3D update and Fan-Out. More information and registration coming soon or contact Camille (

Related presentations

Cet article vous a plu ?

Partagez-le sur vos réseaux sociaux