2017/05/22 Data centers and AI are driving 3D TSV and heterogeneous integration growth
Through-silicon via (TSV) technology has been forecasted to impact the memory and logic industry for years. Die partitioning was developed to optimize technology nodes and improve device yields for high performance technology. Super high memory density to satisfy the need for high bandwidth data transfer between memory and logic devices was also a key driver. TSVs’ role became noticeable a few years ago with the launch of the hybrid memory cube (HMC) interface and high bandwidth memory (HBM), however they only entered small volume production.Now, a major trend is changing the landscape radically. Artificial intelligence (AI) and deep learning needs the right high performance devices and modules to be developed and implemented in the real world. As we at Yole Développement reveal in our latest report, “3D TSV and 2.5D Business Update – Market and Technology Trends 2017”, this means 3D TSV integration is needed.
Source: 3D TSV and 2.5D Business Update - Market and Technology Trends 2017, Yole Développement
Since their introduction, TSVs have clearly held designers’ attention. The main reason is the performance potential that they offer, which is unequalled by any other packaging platform today. If in the past high-performance was synonymous with limited volume and reserved for data analytics and defense applications this is no longer the case. High-end applications like deep learning, data center networking, augmented and virtual reality (AR/VR), and autonomous driving are becoming real, thereby pushing the limits of other current packaging platforms. Last year we saw data centers receiving networking processing units and memory cubes packaged on an interposer. Fueled by increasing bandwidth needs for moving data in cloud-computing and supercomputing applications, performance-driven markets have adopted 3D stacked technologies. Imaging, as the first market adopter of 3D integration, is propelling the market with an increasing number of sensors in smartphones and tablets, including 3D imaging. For these high-end devices, the production of 3D systems on chips (SoCs) is enabling low-latency computing for consumer devices.
Altogether, such trends mean the high-end market has the potential to reach the equivalent of more than 600,000 300mm TSV-based wafers processed annually by 2022. This is a significant volume of high-value wafers, generating very high revenues.
Beyond high-end applications, existing devices are also driving the growth in TSV adoption. There are ever more imaging sensors in smartphones, and needs in computing are still growing, with data center memories the other big driver for TSV adoption. The volume of wafers using the 3D TSV integration manufacturing platform will increase at a 20% compound annual growth rate (CAGR) over the next five years. From the equivalent of 1.3M 300mm TSV wafers in 2016, the overall market will reach 4M TSV wafers in 2022. TSV’s penetration rate in low-end products including MEMS and fingerprint sensors will remain stable. The main source of growth will be radio-frequency filters in smartphone front-end modules, whose volumes will keep increasing in order to support the different frequency bands used in 5G mobile communication protocols. The wafer count for imaging devices will also be boosted by the increasing number of image sensors in mobile phones. Dual cameras will emerge as the de facto standard and there is growing interest for ToF devices for 3D imaging – which, fittingly, are being enabled by 3D integration.
The evolution of the supply chain is also very interesting. Device manufacturers like nVidia, AMD, and Intel are proposing devices and modules to support the AI revolution. However, web companies are positioning themselves downstream in the supply chain to gain better control over outsourcing IC manufacturing and assembly, as well as a greater share of the value chain. One reason is the number of issues attached to 3D integration for high-performance applications. Being present in data centers and the high-performance computing business is key. It is a way for these companies to maintain their leading technology position.
Also, giant web companies are looking to move their activity from being ‘just’ service providers to become IC designers and develop their own brand-labelled chips, leaving manufacturing to the foundries. In developing customized, home-designed dies, web companies are looking to take business from chip sellers. Die design is central to machine learning because of the need for custom solutions to enable and secure applications like connected objects, voice recognition, and autonomous driving. In developing custom, in-house dies, web companies hope to win the race towards autonomous driving, the Internet-of-Things, and AR/VR.
The 3D TSV manufacturing platform is enabling AI, deep learning, AR/VR and new effective data centers. But it is also changing the existing supply chain by making direct links between web companies like Google, Amazon, Microsoft, Baidu and Tencent and the foundries, potentially bypassing the device makers and taking their profit margin. The coming years will see TSV at the heart of some of the world’s most important technology trends.
Source: 
