Redefining Semiconductor Scaling Beyond Transistors

The semiconductor sector has reached a significant milestone, crossing the $1 trillion mark ahead of its target. However, maintaining this growth trajectory now hinges on systems integration rather than relying solely on transistor advancements.

Industry Milestones and Market Growth

At the SEMICON West conference in San Francisco, industry leaders faced a pivotal challenge: sustaining growth that has already outstripped expectations. SEMI President and CEO Ajit Manocha highlighted ambitious projections, estimating the market could exceed $2 trillion by 2030 and may reach $3 trillion by 2035. With spending on 300mm fabrication equipment estimated at $149 billion this year, this could rise to $227 billion by 2029.

Complex System Integration

The increase in demand for advanced technologies such as AI, electrification, robotics, and autonomous vehicles adds complexity to the semiconductor landscape. As Sassine Ghazi, CEO of Synopsys, noted, optimizing engineering requires a shift towards co-designing across various disciplines, including materials, architecture, and software.

There is a convergence of technologies at SEMICON West, where advanced packaging plays a crucial role as part of the overall system architecture. Innovations in photonics and digital manufacturing are becoming intertwined with other technologies, complicating independent optimization and marking a significant challenge moving forward.

Challenges of Co-optimization

Challenges in scaling are becoming multifaceted. For example, improvements in compute performance heighten the demands on memory bandwidth and thermal management. As noted by Lalitha Immaneni of Intel Foundry, advanced packaging now marks the beginning of system innovation rather than an endpoint; the objective is to fine-tune the entire system from silicon through packaging to platform.

Nitin Deshpande of Marvell emphasized that the integration of computing, memory, and cooling systems into a cohesive package is crucial. However, convincing manufacturing processes to scale reliably and maintain high yields remains a significant hurdle.

AI's Role in Designing the Future

Artificial Intelligence is transforming its role within the semiconductor industry, transitioning from a support tool to a pivotal operator in design and manufacturing. With the advent of generative and agentic AI, engineers are now able to optimize solutions over vast parameter spaces. This transition raises a new set of questions regarding decision-making authority and accountability in manufacturing processes.

AI can navigate complex parameter sets, which is crucial given the sheer volume of possibilities in modern semiconductor manufacturing. However, the nuance lies in determining how much autonomy these AI systems should have, particularly when it comes to modifying production processes.

Growing the Packaging Technology

Advanced packaging technologies are no longer experimental; they need to be manufacturable at scale. Experts like Ravi Mahajan from Intel emphasize the importance of ensuring 'known-good' components through testing before assembly to avoid losses. This approach extends to every stage, incorporating reliable processes to reduce uncertainty in manufacturing.

For ASE, the primary goal is ensuring every part of the assembly process is thoroughly vetted, given the significant value of components, sometimes reaching $30,000. Identifying defects early is key to maintaining efficiency in production.

Testing in a Complicated Ecosystem

As semiconductor packages adopt more complex features, the landscape of testing is shifting. Photonic components present unique challenges that require specialized testing methodologies to manage their optical properties effectively. The demand for dual-sided electrical and optical testing is growing as manufacturers strive to streamline testing processes while ensuring quality and efficiency.

This evolution necessitates a shift from a final pass/fail testing framework to a continuous assessment model that makes use of historical data to trace issues through the manufacturing lifecycle. This approach seeks to identify and mitigate problems at early stages, reinforcing the link between manufacturing and AI.

The Path Ahead for Semiconductors

The semiconductor industry stands at a crossroads, needing to marry traditional transistor scaling with complex systems integration and optimization across multiple levels. As AI continues to reshape how engineering and manufacturing are approached, the future will depend heavily on how effectively these integrated technologies can be managed. The industry is not abandoning traditional methods; rather, it is enhancing them to create sophisticated, high-performing systems that can keep up with soaring demand.