When AI models &#34;sink&#34; from the cloud to terminal devices such as mobile phones, watches, and glasses, local operation is becoming increasingly common. Behind this lies a breakthrough in a key technology—WoW stacking (Wafer-on-Wafer) technology.
According to the Chasing Wind Trading Desk, Morgan Stanley's latest research report deeply analyzes the revolutionary impact of WoW (Wafer stacking) technology on edge AI devices. It employs a 3D packaging solution that allows chips to be &#34;stacked vertically,&#34; enabling terminal devices to have sufficient computing power and bandwidth to run lightweight AI models, truly achieving an AI experience that is &#34;available anytime, anywhere.&#34;
Through this solution, WoW can achieve a 10-fold increase in memory bandwidth and a 90% reduction in power consumption, which is expected to break the bottleneck in the development of edge AI. Morgan Stanley predicts that the WoW market size will surge from $10 million in 2025 to $6 billion in 2030, with a compound annual growth rate of 257%.
Morgan Stanley believes that WoW technology is expected to significantly accelerate the adoption of edge AI devices, particularly in application areas such as AI PCs, smartphones, and AI glasses. As the technology matures and supply chain collaboration strengthens, memory manufacturers with first-mover advantages will benefit the most from this wave of technological transformation.
<h2>What is WoW stacking? What problems does it solve?</h2>
In simple terms, WoW is a 3D wafer packaging technology, similar to TSMC's SoIC packaging, which directly stacks logic chips (such as processors) and memory chips (such as RAM) like a sandwich, greatly shortening the &#34;distance&#34; between the two, thereby achieving faster bandwidth, lower power consumption, and smaller size.
Traditional AI computing, especially generative AI, has extremely high requirements for &#34;bandwidth&#34; and &#34;energy efficiency.&#34; For example, the mainstream cloud training solutions currently use a high bandwidth memory (HBM) + GPU combination, which not only consumes a lot of power (often several hundred watts) but also has complex packaging and high costs, making it impossible to fit into mobile phones or glasses.
&#34;Terminal AI,&#34; also known as &#34;local AI&#34; or &#34;edge AI,&#34; allows devices to compute instantly and obtain results immediately, such as: real-time voice translation on mobile phones without needing an internet connection; smart glasses recognizing street objects and providing timely route prompts; AI laptops running large models locally, writing and programming without delay.
But to achieve all of this, there must be a more powerful and energy-efficient AI chip behind it, and WoW stacking is one of the key breakthroughs:
<ul>
<li>
Memory bandwidth can be increased by 10-100 times
</li>
<li>
Power consumption can be reduced by up to 90%
</li>
<li>
Packaging size is smaller, making it easier to &#34;fit&#34; into edge devices
</li>
</ul>
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/36bf5586-0bbd-4e3c-a0db-d8ad177b5cbf.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/36bf5586-0bbd-4e3c-a0db-d8ad177b5cbf.png"/>
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/029b027b-924c-4657-b986-504343277a12.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/029b027b-924c-4657-b986-504343277a12.png"/>
Taking the VHM technology of domestic Aipu Semiconductor as an example, its WoW solution has achieved mass production in cryptocurrency mining machines and will expand to AI smartphones, PCs, smart glasses, and more in the future.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/297264e9-0f8e-4693-a236-e6a8ef00dbdd.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/297264e9-0f8e-4693-a236-e6a8ef00dbdd.png"/>
<h2>Huge Market Potential: Scale May Reach $6 Billion by 2030</h2>
Morgan Stanley predicts that the WoW technology market will experience explosive growth.
Under the baseline scenario, the total addressable market (TAM) for WoW is expected to grow from $10 million in 2025 to $6 billion in 2030, with a compound annual growth rate of up to 257%. The year 2027 will be a critical turning point, with the TAM expected to reach $622 million.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/23793135-0125-423a-ba89-6db4c5925834.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/23793135-0125-423a-ba89-6db4c5925834.png"/>
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/382701e7-5a3f-4cb5-93c2-17abc31156d1.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/382701e7-5a3f-4cb5-93c2-17abc31156d1.png"/>
Morgan Stanley points out that 2027 will be a key year for the initial volume release of WoW technology, with major catalysts including the adoption of major brand smartphones, PCs, and automotive applications.
Automotive applications are expected to start incorporating WoW technology in the second half of 2026, with other applications following in 2027. Under Morgan Stanley's baseline scenario, WoW will double the TAM of the segmented memory market by 2030; in an optimistic scenario, it will achieve a threefold increase.
<h2>Technological Development and Supply Chain Collaboration Becoming More Mature</h2>
Morgan Stanley emphasizes that the WoW technology ecosystem is becoming more mature. Several companies in the Greater China region have applied for WoW patents and announced their own solutions:
Aipu Semiconductor has collaborated with foundries such as TSMC and Powerchip Semiconductor Manufacturing Corporation to achieve mass production; Winbond's two-layer stacking technology has matured and can support models from 3B to 7B; Gigadevice and Changxin Memory Technologies are collaborating on the development, with four-layer stacking already matured and eight-layer stacking on the roadmap.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/e4a061d6-ca5d-4d4b-81be-0e11c168799c.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/e4a061d6-ca5d-4d4b-81be-0e11c168799c.png"/>
Morgan Stanley notes that supply chain collaboration is tighter than in the past, primarily driven by SoC customers who need more effective solutions to address edge AI bottlenecks. The design process requires close cooperation among memory, logic chip design companies, and foundries, as DRAM needs to be stacked above or below the XPU (logic chip).
<h2>WoW vs HBM: Not Competition, but Complementary?</h2>
Morgan Stanley also points out the main risks faced by WoW technology.
First is the competition with mobile HBM, but Morgan Stanley believes that the application scenarios of the two are different. Mobile HBM is mainly used for the main processors of edge devices, while WoW is primarily used for independent neural processing units (NPU) that complement the main processor, so they will not compete directly
Secondly, there are risks related to technological development, including slower-than-expected progress in packaging technology, delays in SoC development, and delays in supply chain partnerships.
In terms of market risks, weak consumer demand and prolonged adjustments in the semiconductor market may affect the adoption speed of WoW technology. Tariff impacts may also increase the costs of AI devices, thereby slowing down the proliferation of edge AI

Through 3D packaging solutions, WoW can achieve a 10-fold increase in memory bandwidth and a 90% reduction in power consumption, which is expected to break the bottleneck in the development of edge AI. Morgan Stanley expects the WoW market size to surge from $10 million in 2025 to $6 billion in 2030, with a compound annual growth rate of 257%

- The article discusses the revolutionary impact of Wafer-on-Wafer (WoW) technology on edge AI devices, enabling local processing on devices like smartphones and smart glasses.  
- Morgan Stanley predicts that the WoW market will grow from $10 million in 2025 to $6 billion by 2030, with a compound annual growth rate of 257%, driven by advancements in memory bandwidth and power efficiency.  
- The technology is expected to enhance the adoption of edge AI applications, particularly in AI PCs, smartphones, and smart glasses, while also highlighting the importance of supply chain collaboration and potential market risks.

WoW Stacking - The Breakthrough Technology that Ignites "Terminal AI"