Mahdi Nikdast Leads NSF-Backed AI Chip-Link Research to Cut Data-Center Energy Use
Updated
Updated · engr.source.colostate.edu · May 27
Mahdi Nikdast Leads NSF-Backed AI Chip-Link Research to Cut Data-Center Energy Use
1 articles · Updated · engr.source.colostate.edu · May 27
Colorado State University researcher Mahdi Nikdast is leading a new NSF ASCENT collaboration to develop photonic interconnects that move AI data faster while reducing the power needed in data centers.
The project targets a growing AI bottleneck: shuttling huge volumes of data between compute and memory nodes now consumes significant computing power and energy as models scale up.
Nikdast's team is combining network architecture, chip packaging and device design to link many small chips in stacked and side-by-side configurations, creating more communication pathways than conventional packages.
Light-based links built directly into the chip package would replace some electrical wiring, aiming to improve speed, scalability and efficiency for next-generation AI systems.
The effort includes the University of Michigan and industry partners, with the researchers saying the work could support faster, lower-power AI hardware from large data centers to edge-computing systems.
Could photonic interconnects and 3D chip packaging truly solve AI's data bottleneck, or will new thermal and reliability challenges create fresh limits?
With AI hardware racing ahead, how might these innovations reshape the balance between centralized data centers and real-time edge computing?
Powering AI’s Future: How Photonic Interconnects Will Transform $4.1B Data Centers by 2034
Overview
Artificial Intelligence is expanding rapidly, creating an urgent need for new ways for AI systems to communicate and process data. At Colorado State University, Mahdi Nikdast’s research, supported by the National Science Foundation, is leading the development of faster and more energy-efficient AI hardware. This work is crucial for solving major challenges in the AI industry, such as the high energy use and data bottlenecks found in modern data centers. By advancing these technologies, Nikdast’s research aims to transform both large data centers and advanced edge-computing systems, paving the way for the next era of AI.