IMPACT Distinguished Lecture - From Chip to AI Factory: A Five-Year Roadmap for Silicon Suppliers and Hyperscalers
Artificial intelligence (AI) infrastructure is moving from accelerator-level optimization toward validated AI-factory capacity. In this regime, peak silicon performance is necessary but insufficient: deployable capacity depends on the coupled behavior of silicon, heterogeneous package, substrate, board, rack, liquid-cooling loop, power train, facility, software observability, service model, and regional infrastructure. This talk proposes a five-year engineering roadmap for converting semiconductor innovation into repeatable AI-factory deployment. The central construct is the validated envelope: a multi-domain operating range in which a chip/package/rack/facility configuration can be manufactured, powered, cooled, serviced, monitored, regulated, and replicated at scale without exposing proprietary design details. The lecture will define validated envelopes across thermal, mechanical, electrical, fluidic, manufacturing, serviceability, telemetry, safety, regulatory, and regional-infrastructure domains; map key technology drivers (chiplets, HBM, optical I/O, power delivery, liquid cooling, mechanical reliability, fleet telemetry); and propose milestone gates for silicon suppliers, hyperscalers, foundry/packaging/OAT ecosystems, cooling suppliers, standards bodies, and policymakers. The roadmap is framed as a decision and validation framework rather than a device forecast and argues that the practical unit of AI infrastructure scaling is now the validated chip-to-region stack: validate once, replicate many times.
Date and Time
Location
Hosts
Registration
-
Add Event to Calendar
- No. 2, Jingmao 2nd Rd.
- Nangang Dist.
- Taipei, T'ai-pei
- Taiwan
- Building: Taipei Nangang Exhibition Center, Hall 2 (TaiNEX 2)
- Room Number: 701A
- Click here for Map
- Starts 30 July 2026 04:00 PM UTC
- Ends 19 October 2026 04:00 PM UTC
- No Admission Charge
Speakers
Gamal Refai-Ahmed
From Chip to AI Factory: A Five-Year Roadmap for Silicon Suppliers and Hyperscalers
Biography:
Gamal Refai Ahmed, Ph.D., FIEEE, LFASEM, Fellow Canadiana Academy of Engineering, Fellow Engineering Insitute of Canada, AE, and a member of the National Academy of Engineering, is a distinguished technical executive renowned for his groundbreaking work in thermal management, silicon and power architecture, and advanced packaging technologies. With over two decades of experience at industry giants like AMD, GE, and Cisco, Dr. Refai-Ahmed has led high-impact projects, including pioneering new silicon and thermo-mechanical architectures for advanced technology nodes and enabling innovative packaging technologies for ASIC and embedded FPGA. His tenure at AMD as Senior Fellow and Chief Architect saw him spearheading thermal management solutions for Xilinx products and fostering research collaborations to expand HPC, NIC, AI, and ML ecosystems in data centers. He enhanced thermal management at GE for intelligent platform control systems and healthcare mobile systems, contributing significantly to GE’s patent portfolio. Gamal has also defined technology directions and led projects that improved thermo-mechanical solutions for LED printing and UV LED products. Dr. Refai-Ahmed’s academic background includes a Ph.D. in Mechanical Engineering from the University of Waterloo, Canada, with numerous publications and over 160 US and international granted and pending to his name, as well as more than 120 publications in IEEE, ASME, AIAA conferences and journals. His exemplary contributions have earned him prestigious accolades, including election to the National Academy of Engineering, IEEE Fellow, and the Presidential Medal from Binghamton State University. Actively involved in professional societies, Dr. Refai-Ahmed continues to influence the field through his roles in IEEE and ASME, shaping the future of high-performance computing and advanced packaging solutions.