Chinese semiconductor design firm Berain Technology has unveiled the next generation of “super node” solutions – systems designed to connect thousands of artificial intelligence chips across a single cluster – using optical data transmission to push the limits of current hardware.
The launch underscores how hyper-connected server systems have become one of the newest battlegrounds for AI infrastructure companies. The industry is currently racing to increase raw computing power as AI models advance to trillions of parameters and AI applications gain widespread adoption.
To break through this ceiling, Perrin is betting on distributed and discrete supernode architecture and near-packaged optics (NPO), which integrates optical fibers closer to the chips to boost data speeds. The layout can allow Biren to scale up to 1,024 cards across the deck.
“Optical interconnection has become an inevitable option to break through this bottleneck,” Ding said.
Shanghai-based Perrin is the latest local company to announce progress in addressing infrastructure constraints.
