As the world continues to advance the capabilities of Artificial Intelligence, its networks are straining the limits of interconnectivity and power consumption. The rapidly increasing need for larger switches, lower latency at higher data rates while also slowing the rate of power consumption appear to be the only path forward for the future growth of AI. However, we are stuck in a decade’s old paradigm of shrinking transistor sizes with more dense microchips, operating ever faster with each node improvement. This in turn requires more circuitry and more power to maintain the fidelity of the electrical signaling, but where all the “intelligence” is concentrated down into a chip no bigger than a postage stamp, with the added uncertainty of just how small a transistor can get before physics or cost finally ends that progression. Axonal Networks proposes a new way to interconnect the network through the development of paraDOX – the parallel Digital Optical Cross (X) Connect architecture. Unlike circuit-switching concepts, paraDOX uses a new way to build an Ethernet Packet Switch while being able to scale to thousands of ports. paraDOX is a truly distributed system that scales almost linearly in terms of resources, while also eliminating the need for optical transceivers by making the light part of the digital logic itself. This talk will outline, at a high level, the general architecture along with the proposed advancement of distributed logic structures that are being developed.
Event:
30 October 2025 03:00 PM
Time: 03:00 PM EDT
