Scaling AI with Chiplets & CPO
By Dave Kulansky, Director of Product Management at Qualcomm
AI's explosive growth is straining datacenter connectivity across scale-up and scale-out networks, pushing optics ever closer to the compute. This talk traces that shift from pluggable modules to near-packaged optics (NPO) to co-packaged optics (CPO), showing how chiplet architectures and heterogeneous integration enable denser, lower-power optical I/O. We examine direct-drive versus digital-drive CPO, co-design, and weigh the opportunities and challenges shaping optical interconnects for upcoming generations.
Related Chiplet
- Optical Chiplet
- Reverse Gearbox Chiplet
- FlexGen Multi-Die Smart Network-on-Chip (NoC) IP
- Ncore Multi-Die Interconnect IP
- Integrated voltage regulator (IVR) chiplet
Related Videos
- Rearchitecting AI Infrastructure with General Purpose and Accelerator Chiplets
- Inside the AI Bottleneck: Data Movement, Chiplets, and System Scaling
- Building Interoperable AI Systems with Chiplets Arm Neoverse and Cadence in an Open Ecosystem
- Getting Moore with Less: How Chiplets and Open Interconnect Accelerate Cloud-Optimized AI Silicon
Latest Videos
- Physical AI Addressing the barriers to interoperable chiplets in automotive and robotics
- Letizia Giuliano, Qualcomm & Sophie Wang, Marvell | GSA Women's Leadership Initiative 2026
- Advancing Chiplet Flexibility: Updates in 2.0 OCE Transaction and Link Layer Specification
- AI native Virtual Chiplet Eco-systems: industry best practices to accelerate Chiplet adoption
- Introduction to Semiconductor Packaging