The world of artificial intelligence (AI) and hyperscale data centers is evolving at a breakneck pace. AI is growing into every part of our lives. For example, the U.S. Food and Drug Administration has approved hundreds of AI-powered medical devices, while Waymo provides more than 150,000 autonomous rides each week via its self-driving cars.
A 2025 study from Stanford University’s Institute for Human-Centered AI found that AI models’ performance on demanding benchmarks sharply increased in 2024. In one benchmark, AI systems could solve 71.7% of coding problems in 2024, a tremendous improvement over just 4.4% in 2023. As workloads become more complex and the demand for performance, efficiency, and scalability grows, the need for custom, high-performance silicon solutions is growing as well.
At Ayar Labs, we believe that the future of AI infrastructure will be shaped not by any single company, but by a robust ecosystem of partners working together to push the boundaries of what’s possible. Building an ecosystem of application-specific integrated circuit (ASIC) providers is essential for delivering purpose-built multi-chip packages (MCPs) that power the next generation of AI and hyperscale computing.
Why Hyperscalers are Turning to Custom MCPs
Hyperscalers like Amazon, Google, Microsoft, and Meta need advanced AI accelerators and platforms that deliver enhanced performance, efficiency, and scalability. They are increasingly seeking to lower their TCO (total cost of ownership) in AI infrastructure. As inference workloads scale rapidly alongside continued growth in training, hyperscalers are focused on power and utilization to control operational costs and boost efficiency. Custom silicon chips tailored to these specific workload needs can help realize these savings.
Increasingly, hyperscalers are working with ASIC design services partners to balance the design and integration workflow for purpose-built MCPs. Many are adopting a hybrid approach, taking on more front-end design responsibilities while leveraging the expertise of back-end ASIC services partners. Co-packaged optics (CPO) solutions, like those from Ayar Labs, are designed to be easily integrated into a wide range of ASIC design flows to provide greater flexibility and value.
Custom MCPs represent a significant evolution. By integrating multiple specialized chips—each optimized for a particular function—MCPs equipped with Ayar Labs’ optical engines can hit target performance levels at lower TCO while also unlocking new architectural possibilities. For example, in AI inferencing, different ASIC designs can be tailored to support a diverse range of workloads and model requirements. Optimizing silicon for specific AI models allows hyperscalers to improve power efficiency and throughput, which ultimately improves the return on invested capital.
As compute and package size become limiting factors, integrating CPO solutions becomes a necessity. CPO makes it possible to cluster resources into unified, high-bandwidth pools that support massive compute fabrics for a wide variety of AI models and real-time inferencing. This approach can enable entirely new classes of applications and business models. As AI workloads diversify and scale, the ability to rapidly iterate and deploy custom silicon with CPO is likely to be an important competitive advantage.
The Ecosystem Approach
Working with a broad set of ASIC partners offers several benefits to Ayar Labs and our customers:
- Flexibility and Risk Mitigation: Partnering with multiple providers enables us to service specific customer needs across a dynamic business environment.
- Supply Chain Resilience: Expanding our ecosystem strengthens supply chain resilience, helping us navigate global disruptions in an increasingly unpredictable world.
- Faster Innovation: Collaborating across companies accelerates innovation, technical learning, and knowledge sharing so new solutions reach the market faster.
- Competitive Differentiation: Building a diverse partner community drives competitive differentiation, fostering healthy competition, improving TCO, and maximizing return on investment for our customers.
We see this ecosystem approach not just as a strategy, but as our core business philosophy. We believe that open collaboration, mutual learning, and shared success are critical to unleashing the full potential of AI and silicon photonics. That’s why we’re building a robust ecosystem of ASIC providers, foundry partners, and technology collaborators, laying the groundwork for the next wave of AI infrastructure.
Key Partnership with Alchip
A recent example of our ecosystem approach is our partnership with Alchip, which is known for its advanced ASIC designs and high-volume manufacturing. At the 2025 TSMC North America Open Innovation Platform (OIP) Ecosystem Forum, Alchip and Ayar Labs unveiled a CPO solution that is designed to efficiently extend scale-up network connectivity beyond traditional rack boundaries.
This new joint solution features Ayar Labs’ TeraPHY™ optical engines co-packaged with Alchip’s advanced solutions on a common substrate, bringing optical I/O directly to the AI accelerator interface. This integration enables more than 100 terabits per second (Tbps) of scale-up bandwidth per accelerator and supports more than 256 optical scale-up ports per device. With support for the Universal Chiplet Interconnect Express™ (UCIe) standard for die-to-die interconnect, and flexible protocol endpoints at the package boundary, design teams can integrate the scale-up fabric solution from Alchip and Ayar Labs alongside existing compute tiles, memory stacks, and accelerators while meeting performance, signal integrity, and thermal requirements at the package level.
Figure 1. A CPO solution in collaboration with Alchip Technologies featuring eight Ayar Labs’ TeraPHY™ optical engines and two AI accelerators.
The Importance of CPO to the Future of AI
Traditional copper-based interconnects can no longer keep pace with the speeds and architectures demanded by AI workloads. But our optical I/O technology can. Together with Alchip, for example, we can enable multi-rack scale-up system architectures designed to overcome the legacy performance bottlenecks created by copper. These systems aim to dramatically improve interactivity while reducing power consumption.
“AI cluster scale-up networks are bounded by copper connectivity distances. At the same time, network power efficiency is limited by power density and cooling. This means that we want to optimize space efficiency, but the challenge is that copper doesn’t have the I/O connectivity to put everything together. CPO will break these boundaries.”
—Erez Shaizaf, Chief Technical Officer, Alchip Technologies
A Market Ready for Disruption
Industry observers are taking notice as well: according to Yole Group, CPO is on the cusp of a significant market opportunity and becoming a vital part of the AI ecosystem. Their latest research forecasts the CPO market will grow from $46M in 2024 to $8.1B by 2030, with a 137% CAGR. Some experts are even more bullish, with recent estimates suggesting the total addressable market opportunity could grow more than 10x as optical engines scale next-gen compute architectures. This surge highlights why the industry is rapidly embracing CPO as a foundational technology for future AI systems.
Figure 2. Hyperscalers and global players are driving a new era of AI infrastructure spending. (Sources: Company reports, BBG, MSUSA, Morgan Stanley)
Of course, our strategy is built on alignment and close collaboration with leading partners across the Taiwan semiconductor ecosystem. Advanced packaging and process technologies from this region are enabling new levels of integration, performance, and efficiency for high-volume MCPs. By combining compute, memory, and photonics in ways previously not possible, we’re opening the doors to new architectures like AI scale-up and memory disaggregation. This means faster time-to-market, greater scalability, and a clear path to manufacturing readiness for our customers.
Scaling Success Through Strategic Partnerships
Working with multiple partners allows Ayar Labs to scale more rapidly, address a wider range of customer needs, and capture a larger share of the growing AI infrastructure market. Ayar Labs is actively engaging with select customers to integrate CPO into AI accelerators and scale-up switches. Our solutions are backed by a robust ecosystem of key strategic top-tier semiconductor leaders, including AMD, Intel, and NVIDIA, underscoring our position as a recognized industry leader.
The future of AI infrastructure depends on collaborative innovation and robust ecosystems. Ayar Labs is committed to leading this transformation with its partners.
- Access the Ayar Labs/Alchip TSMC OIP Ecosystem Forum presentation: Enabling Next-Gen AI Infrastructure with Co-Packaged Optics and Advanced Packaging
- And if you’ll be at SC25 in St. Louis, MO on Nov. 16-21, visit us at booth #2148
