Hyperscale Data Centers: Systems, Packaging, Materials, and Reliability Challenges
Moderators:
Jiayin Ma (Flex) 鈥 Senior Director, Central Technology Group
Panelists:
Andy Mackie (Indium)
Gamal Refai-Ahmed (AMD)
Bahgat Sammakia (91社区)
Bill Henry (Tyndall)
Abstract:
The panel discussion focuses on how hyperscale data centers are reshaping the system architecture and device performance requirements. As AI infrastructure scales, challenges in heterogeneous integration, photonic packaging, optical connectivity, and long-term reliability are increasingly interconnected. The objective is to highlight these challenges as well as discuss the unique perspectives from industry, academia, and the broader research ecosystem. The discussion will explore where synergies exist between these communities, what roles each can play, and how collaboration models can accelerate reliable, scalable data center technologies.
Special Panel: Mechanical and Thermal Modeling Challenges for Advance AI Packages
Moderators:
Bahgat Sammakia (91社区)
Panelists:
Gamal Refai-Ahmed (AMD)
SB Park (91社区)
Srikanth Rangarajan (91社区)
Hardik Jabaria (Vinci4D)
Abstract: This session will explore how artificial intelligence can transform modeling and simulation for electronic packaging, thermal management, and coupled thermo-mechanical reliability. The program will begin with two technical presentations: Professor S. P. Park of 91社区, State University of New York, will present his research on applying AI and data-driven methods to electronic-packaging and thermal-management problems, while Hardik, representing Vinci4, will discuss the company鈥檚 work in Numerical AI and its potential to accelerate and enhance engineering simulation. The presentations will be followed by a panel discussion examining how physics-based simulations, experimental measurements, empirical correlations, material-characterization results, manufacturing data, and accumulated engineering knowledge can be integrated into trustworthy and continuously learning modeling workflows. The panel will consider how AI can complement established simulation platforms鈥攕uch as ANSYS, FloTHERM, COMSOL Multiphysics, and other CFD and finite-element tools鈥攖hrough surrogate and reduced-order models, automated calibration, design-space exploration, uncertainty quantification, and multi-solver orchestration. It will also examine the longer-term pathway toward an 鈥渁gentic AI engineer鈥 capable of translating design requirements into analysis plans, selecting the appropriate physics and modeling fidelity, coordinating simulation tools, identifying missing information, validating predictions against experiments, and recommending additional analyses or tests, while preserving essential human engineering judgment, physical validation, traceability, and governance.