With the explosive growth of applications such as AI chips and high-performance computing, advanced packaging is becoming the most dynamic segment in the semiconductor industry chain. Nvidia's latest AI chip power has reached 2700 watts, requiring about 100 amps at 12 volts—this poses unprecedented challenges for advanced packaging processes. Traditional packaging technologies can no longer meet demand, and advanced technologies like TSV (Through-Silicon Via), 2.5D, and 3D packaging have become key industry breakthroughs.

However, the complexity of advanced packaging far exceeds that of traditional packaging: it involves both front-end and back-end processes, requiring deep integration of previously separate wafer manufacturing and packaging/testing stages; process flows are nonlinear, with frequent rework and reconfiguration operations; it also faces multiple challenges such as equipment diversity, data silos, and rapid mass production. Against this backdrop, how to achieve cost reduction, efficiency improvement, and quality enhancement through digital transformation has become a core issue the industry urgently needs to address.

Recently, at the 2025 China Semiconductor Advanced Packaging and Testing Conference, Yang Jun, head of GETECH's advanced packaging and testing business line, shared the company's in-depth practices and innovative insights in semiconductor digital transformation under the theme 'From Automation to Autonomy: The New Generation Evolution Path of AI-Empowered Advanced Packaging and Testing CIM.'

Multiple Challenges Facing Advanced Packaging

In Yang Jun's view, the advanced packaging industry is currently facing an unprecedentedly complex situation.

"Advanced packaging differs from traditional packaging, with the biggest challenge lying in process complexity," Yang Jun pointed out. "It includes both front-end and back-end processes, which is completely different from past management models. How to integrate digital management across front and back ends is a major challenge for many clients."

From the demand side, emerging applications like AI, IoT, 5G, and ADAS continue to drive the semiconductor industry's development but also bring exponential growth in process complexity. From the supply side, factors such as the slowing of Moore's Law, the rise of custom chip design, and geopolitics are reshaping the industry landscape. "Our clients start advanced packaging development simultaneously with new GPU R&D; how to achieve the fastest time-to-market to capture the market is a significant challenge," Yang Jun emphasized.

He summarized five major challenges currently faced by advanced packaging: flexible manufacturing, multi-site collaboration, policy environment, integration difficulties, and market pressure. These technical issues are essentially management and strategic problems.

In fact, there is a hidden, easily overlooked contradiction: the more 'advanced' the technology, the faster it needs mass production; the more complex the process, the more it demands flexible manufacturing. The traditional 'steady and gradual' approach is increasingly struggling to keep pace with the rapid development of advanced packaging today.

From Technology-Driven to Business-Driven

Facing these challenges, GETECH proposed a systematic solution. Yang Jun emphasized in his speech that the core of digital transformation is not technology itself, but how to deeply integrate technology with business to truly solve clients' practical problems.

"When enterprises undergo smart manufacturing transformation, what they often lack is not technology, but methodology," Yang Jun said. "In TCL's practice, we found that shifting from technology-driven to business-driven is a very important first step in transformation."

In 2016, TCL formally incorporated smart manufacturing into its group strategy, establishing a 'three integrations and four steps' approach to smart manufacturing. The 'three integrations' refer to the fusion of lean, automation, and informatization; the 'four steps' include data connectivity, business process informatization, actionable digital operations, and intelligent data insights.

This methodology has yielded significant results. By 2025, TCL had built one pilot factory and five smart manufacturing maturity level four factories.

In subsequent exchanges, Yang Jun further elaborated on GETECH's core philosophy: "We shift from technology-driven to business-driven, prioritizing the activation and optimization of existing assets—including human and hardware assets—rather than blindly pursuing new investments."

The value of this philosophy was validated in a project with a leading packaging and testing plant in Xi'an for an automotive-grade packaging line. Instead of recommending full-line automation, GETECH used lean analysis to plan four automation units, then used logistics systems to bridge business process gaps, ultimately helping reduce equipment investment by one to two units per workstation.

Currently, there is a common misconception in the industry where some enterprises equate digital transformation with 'equipment upgrades' or 'system implementation,' resulting in huge investments with minimal effects. In contrast, GETECH identifies the most critical pain points, bottlenecks, and value points, truly making technology work for them.

AI Is Not a Panacea

After years of accumulation, GETECH officially launched the first Fully Auto CIM solution in China specifically for advanced packaging and testing in 2024, with a core of the 'AI+CIM+AMHS' trinity architecture.

In AI applications, GETECH's industrial AI follows the 'ABCDE' model: Algorithm, Big data, Computing power, Domain knowledge, and Equipment—five dimensions that integrate with each other. The most distinctive feature is its 'small model first' strategy—implementing AI small models in scenarios like quality management, energy and carbon management, and equipment management before the large model trend. This 'lightweight AI' path is more practical in industrial settings with limited computing power.

In CIM, addressing the complex processes of advanced packaging, GETECH introduced the industry's first full-process CIM suite, covering the complete business chain from wafer input management to yield analysis. Key breakthroughs include productized delivery and equipment interconnect optimization—moving beyond the fragmented limitations of past custom development, and achieving rapid deployment and multi-device compatibility through a central equipment template library and no-code programming tools. This gives advanced packaging clients the first opportunity to achieve 'one platform managing the entire factory.'

Finally, in AMHS, automated logistics is the 'arterial system' of smart manufacturing. Through strategic acquisitions and the establishment of a new company, Wilson Optimus, GETECH built a complete AMHS product line, from OHT (Overhead Hoist Transport) to Stocker to MCS control systems, covering both hardware and software stacks. In the highly competitive and funding-constrained domestic AMHS track, GETECH, leveraging TCL's support and AI algorithm advantages, chose to 'compete upward': co-establishing an Industrial AI Joint Laboratory with the University of Hong Kong, with the first collaboration project focused on enhancing OHT intelligent scheduling capabilities to achieve thousand-unit-level coordination.

Although this solution is comprehensive, Yang Jun has his own reflections:

"In the first half of this year, the questions I was asked most by clients were: How to calculate the ROI of AI projects? What is a reasonable ROI cycle? But disappointingly, among the clients we visited, 60%-70% of AI projects actually did not find good value scenarios or lacked high-quality data; these projects typically showed no value."

Another 20%-30% of AI projects are under continuous investment, with cycles of one to three years, but have not yet materialized. "Only less than 10% are truly successful, but once successful, the benefits grow very quickly, with huge increases."

He cited a leading semiconductor materials company as an example: by introducing AI algorithms and using RTS+RTD for intelligent dispatch and real-time scheduling, annual revenue approached 10 million in the first year after implementation, already exceeding the entire project investment. "So, for AI projects that truly find the right direction, implementation can be completed in three to six months."

For many enterprises, AI seems to have become a panacea, as if all problems will be solved with AI, but the reality is not so. Finding the right value scenarios for application is key to leveraging AI's role.

For GETECH, there is no shortage of proven cases: AI-FDC (Fault Detection and Classification) systems monitor CMP equipment in real-time, successfully providing early warnings for issues like filter blockages and reducing quality losses; Luban Assistant integrates equipment knowledge, helping engineers with 1-2 years of experience quickly reach the level of 3-4 year engineers; AI energy and carbon optimization assists factories in reducing peak electricity costs...

Identifying suitable value scenarios for use, rather than treating AI as a panacea, is precisely why GETECH's Fully Auto CIM solution for advanced packaging and testing stands out among numerous production solutions for the advanced packaging industry.

The Calm and Restraint of Moderate Intelligence

Amid the AI and automation boom, Yang Jun also proposed a reflective concept: moderate intelligence.

He presented a curve: in the early stages, automation/digital investment significantly reduces manufacturing costs; but when climbing from L3 (integration level) to L4 (optimization level) and L5 (leading level), continued investment leads to slower declines in manufacturing costs, or even increases due to overinvestment.

"So today, we shouldn't blindly invest in digitalization, but find the optimal value point suitable for your factory and seek a balance," Yang Jun said. "Man-machine ratio is not what we pursue; we pursue optimal factory operating costs. It's not about blindly pursuing 100% automation rate or 100% yield rate, but finding the balance point."

In an industry filled with anxiety and a pervasive pursuit of 'extremes,' GETECH's calm and restraint are particularly valuable. Over-automation and over-intelligence are essentially resource waste and may even become a burden for enterprises.

Autonomy and Control: Not Just About Security, But a Prerequisite for Innovation

Beyond technical and commercial value, GETECH's full-stack domestic AI+CIM+AMHS solution also carries deeper strategic significance.

"The first thing to discuss must be security," Yang Jun emphasized. "In semiconductor factories, there are two core systems: one is the brain CIM system, and the other is the arterial AMHS system. Both were highly dependent on foreign suppliers in the past, not only costly but also facing risks of supply chain disruption and information security at any time."

GETECH's full-stack domestic solution, supporting domestic databases and operating systems, but also uses domestic and open-source technologies in the underlying tech stack—from databases to message middleware to programming languages—avoiding chokehold risks.

"The second value is system synergy," Yang Jun continued. "Traditionally, building a Fully Auto factory required multiple suppliers: one for the CIM system, and two to three for automation. This easily creates data silos, making it hard for the informatization brain and intelligent units to collaborate efficiently. Our purpose in acquiring two automated logistics companies is to bridge IT and OT."

The deeper significance lies in industrial synergy and innovation breakthroughs. "For Mainland China's advanced packaging technology to do better, it requires the synergy of the entire industrial system—including semiconductor materials, equipment, manufacturing processes, and the industrial software we provide," Yang Jun said. "If we still use foreign solutions, it's like playing Go: no matter how high our skills, we're ultimately confined within this board."

Within the board, all is constraint; only outside the game can innovation occur—this is the fundamental reason for adopting domestic overall solutions, promoting industrial synergy, and creating solutions suited to China's semiconductor industry.

Conclusion

At the end of his speech, Yang Jun summarized four core experiences for enterprise smart manufacturing transformation: shifting from technology-driven to business-driven, pursuing optimal operating costs, integrating the three transformations with lean thinking, and leveraging AI technology to accelerate transformation—seemingly simple, but actually refined from years of practice.

Advanced packaging is in a golden period of rapid development, but its complexity brings unprecedented challenges. Based on TCL's over forty years of deep accumulation in smart manufacturing transformation, GETECH, through the innovative 'AI+CIM+AMHS' architecture, provides the industry with a feasible path from automation to autonomy, from experience-driven to intelligent decision-making.

"Advanced packaging is a very important strategic track for GETECH's future," Yang Jun said. "We hope to explore advanced packaging digital transformation together with the industry, promote our best solutions to more clients, and contribute to the high-quality development of China's semiconductor industry."

Against the backdrop of increasing geopolitical uncertainty, the concepts advocated by GETECH—autonomy and control, industrial synergy, and intelligent decision-making—not only concern enterprise cost reduction and efficiency improvement but also relate to the entire industry's innovation breakthroughs and long-term development. In this sense, the digital transformation of advanced packaging is not just a technological revolution, but a strategic choice concerning the industry's future.

*This article is authorized for reprint from the WeChat public account 'Semiconductor Chip News' (ID: MooreNEWS).