From concept to production in advanced packaging

SkyWater Technology outlines how its Technology-as-a-Service model is helping customers bridge the gap between early-stage innovation and scalable manufacturing, as advanced packaging reshapes semiconductor development.

An interview with Phil Alsop, contributing editor of Advanced Packaging, and Percy Gilbert, Senior Vice President of Engineering at SkyWater Technology, explores how the company enables customers to move from early-stage concepts to manufacturable technologies through its Technology-as-a-Service (TaaS) model. The discussion spans advanced sensing, heterogeneous integration and quantum-adjacent systems, while Ross Miller, Chief Strategy Officer, provides additional insight into advanced packaging capabilities, including chiplets, 2.5D and 3D integration, packaged optics and quantum computing, a market projected to reach $80 billion by 2030 as applications extend beyond the data centre.

PA: We are going to cover technology as a service and advanced packaging. Starting with TaaS, SkyWater positions this as distinct from standard manufacturing services. Structurally, what makes it different from joint development agreements or custom process engagements?

PG: The biggest advantages come down to access, speed and end-to-end support. From an access standpoint, we work with materials that larger fabs would not typically allow into production environments. We have the flexibility to bring in these more exotic materials and process them on real production tools.

That leads directly to speed. Customers are developing and experimenting on the same toolsets that will ultimately be used for manufacturing, so learning cycles are faster and more relevant. From an end-to-end perspective, we are not just developing processes. We are enabling platforms with full design infrastructure, including PDKs, so customers can move from innovation to product realisation.

In this space, we are dealing with new materials, new integration schemes and new processes. Our goal is to take these from proof of concept to manufacturable solutions quickly, which differs significantly from traditional joint development alliances that focus on scaling established technologies.

In conventional environments, especially at scale, anything that introduces risk to production is avoided. Exotic materials or unconventional approaches are not allowed because they could disrupt yield or tool stability. Our model is different. We can bring those materials in and apply production-level thinking to early-stage development, helping customers bridge the gap between concept and manufacturability.

PA: Turning to advanced packaging, it is evolving rapidly as a key enabler for next-generation semiconductors. What is driving the shift away from monolithic scaling?

RM: Advanced packaging is a natural evolution as transistor scaling approaches physical limits. Demand for performance has not slowed, so the industry is finding new ways to deliver it.

Packaging enables higher density integration, combining different types of chips into new architectures that improve performance, reduce power and increase flexibility. AI is the primary driver, but we are also seeing strong interest from areas like quantum computing. That will require complex integration and higher qubit counts, which advanced packaging can enable.

PA: Percy, you describe innovation occurring deep in the technology stack before standards or volume production exist. What does that mean in practice?

PG: Traditional foundry models rely on well-characterised processes that can be combined to produce an end product. In our case, those baselines often do not exist.

We may start with entirely new materials or device concepts, working from first principles to understand how they behave in process tools. That includes developing recipes, defining specifications and establishing integration pathways.

A key challenge is translating laboratory concepts into CMOS-compatible flows. For example, developing an etch process for a material that has never been used in a CMOS environment. The question becomes how to integrate that material into a manufacturable process.

Working deep in the stack means solving those integration challenges, ensuring that novel materials and device concepts can ultimately be scaled into real products.

PA: Ross, chiplet architectures are a major focus. What are their key advantages in terms of performance, yield and flexibility?

RM: Chiplets offer clear benefits in density, size, weight and power. They enable tighter integration and shorter interconnects, which improves performance and reduces power consumption.

They also provide flexibility. You can combine different process technologies in a single system, which is not possible with monolithic designs—for example, integrating compound semiconductor devices, advanced logic and sensors into a single heterogeneous architecture.

PA: Percy, when innovation happens at the process and integration level, IP sensitivity increases. What are the main risks?

PG: The two main categories are process knowledge and integration decisions. At this stage, IP includes tool recipes, process flows and interactions between materials and design rules. These are highly sensitive.

Integration decisions are equally critical. Choices around thermal cycles, planarisation or stress management can determine whether a design is viable. These decisions form a key part of the IP.

This differs from traditional design-level IP, such as layouts or system architectures. Here, the IP is embedded in how the process itself is created and controlled.

PA: Ross, what challenges remain for widespread adoption of advanced packaging?

RM: Yield is a major challenge. These processes are still relatively early in their lifecycle, and the business model itself introduces complexity by combining components from different sources.

Ensuring known good die, validating integration and testing full heterogeneous systems are all ongoing challenges. Infrastructure is another issue. There is a need for significant investment to scale advanced packaging capacity, particularly in Western markets.

The third challenge is knowledge. Many customers are still developing the expertise needed to adopt these technologies effectively and cost-efficiently. That is where collaboration becomes critical.

PA: Are you confident that yield challenges will be resolved?

RM: Yes. It is part of the natural evolution of any emerging technology. Advances in metrology, equipment and architecture will improve yield over time. These are engineering challenges, not fundamental barriers.

PA: Percy, how do you manage IP protection without slowing collaboration?

PG: It comes down to structured team environments and need-to-know access. Teams are organised around specific projects with controlled access to data and tools.

Within each team, communication is open and efficient. Across teams, there are clear boundaries to protect IP. This allows us to maintain speed while ensuring proper separation between projects.

PA: Ross, what role do co-packaged optics play in addressing bandwidth and energy constraints?

RM: They are critical. Data centre power consumption trends make it clear that efficiency improvements are essential.

Optical communication is a key lever, and integrating optics at the package level improves both bandwidth and energy efficiency. Co-packaged optics will be fundamental to scaling high-performance and edge computing systems.

PA: Percy, when working on technologies that may not reach volume production for years, how do you define success?

PG: It depends on the project. Process stability, control and repeatability are always essential.

Beyond that, success metrics can include early reliability, specific performance targets or even a single critical capability. In some cases, achieving a defined metric, such as a specific etch quality, is the entire objective.

For projects moving towards production, we also look at readiness for scaling, even if that scale is relatively small in absolute terms.

PA: Ross, how important is advanced packaging for quantum computing?

RM: It is foundational. Scaling quantum systems requires sophisticated packaging solutions.

It is not just about adapting existing architectures; quantum systems operate under entirely different conditions, including temperature and material requirements. Advanced packaging is essential to move from laboratory demonstrations to scalable, utility-level quantum computing.

PA: Finally, how important is collaboration as these technologies evolve?

PG: It is critical. Our model depends on close collaboration with customers and across technical disciplines. Many of the challenges we face are not fully defined at the outset, so solving them requires collective expertise.

RM: Collaboration is at the core of what we do. Unlike conventional foundry models, innovation here spans the entire technology stack. Customers bring device-level expertise, and we bring process and manufacturing expertise. Together, we form integrated teams that turn new ideas into real products.

PG: That collaborative environment is one of the main reasons customers choose to work with us. It enables open innovation while maintaining the structure needed to deliver results.