Technology Solutions for Essential Chips: From Cloud to the Edge

Abstract

The rapid evolution of artificial intelligence (AI) demands advanced semiconductor solutions that deliver high performance, energy efficiency, and scalability - from cloud servers to edge devices. This keynote explores cutting-edge technologies from GlobalFoundries enabling next-generation of AI chips, with a focus on Low-power embedded CMOS, Power-efficient GaN devices for energy delivery for Servers, Silicon photonics & RF Solutions for high-bandwidth connectivity, and Embedded non-volatile memory (eNVM) for on-chip intelligence.

I will examine how 3D integration and chiplets overcome traditional scaling limits, enabling heterogeneous architectures that combine logic, memory, and analog/RF components. Technology innovations in integrated voltage regulators (IVR) and wide-bandgap (GaN) power devices are critical for dynamic power management, reducing energy waste in AI data centers. Meanwhile, silicon photonics & SiGe addresses interconnect bottlenecks, enabling ultra-fast data movement between chips and systems.

For edge AI, I highlight ultra-low-power embedded CMOS designs with wireless connectivity and eNVM technologies (e.g., MRAM, RRAM) that enable always-on intelligence with minimal energy consumption. The convergence of these essential technologies from GlobalFoundries - Power GaN for data center efficiency, Si-Photonics & RF Technologies for bandwidth, 3D stacking for density, and IVR for power delivery paves the way for sustainable, high-performance AI hardware. .