In Pursuit of 1000x and beyond
We build silicon photonic solutions in pursuit of 1000X advances in data connectivity bottlenecks and next-generation AI compute.
3E8 develops lightspeed silicon photonic solutions to achieve breakthroughs in optical connectivity with a vision of 1000X advances in system level AI compute throughput and efficiency.
Digital electronic chips are facing the sunset of Moore’s law and the end of Dennard Scaling (transistor energy density) which, combined with extreme challenges and costs of shrinking transistors below 5nm means that a new paradigm is needed to help meet the massive and growing AI compute gap over the coming decades.
Furthermore, the physical limits of trying to push data at higher speeds efficiently over copper are unsolvable which together with the afore-mentioned chip scaling challenges results in unsustainable growth in datacenter AI power consumption and associated costs. Latest AI processors consume 700W+ which requires exotic water/nitrogen cooling.
Apart from the chip level efficiency challenge – The major bottleneck today for AI scale out is the bandwidth and energy dissipation limitations of Datacenter (DC) connectivity. From Intra-DC to Rack to Card to chip level the energy penalty kills system level efficiency and prevents composable datacenter architectures where compute, networking and memory resources are globally accessible and optimizable. The solution lies in enabling optical communications to extend from DC to Rack and right down to chip level ( Co-packaged optics).
3E8 is developing breakthrough connectivity technologies to advance the key electro-optical conversion bottleneck and advance higher levels of efficient optical modulation schemes.
The long term AI compute challenge will be addressed likewise using photons with a unique Photonic Tensor Core architecture with breakthrough system frequency. The architecture is based on novel photonic memory and electro-optical modulation technologies to achieve orders of magnitudes advances in speed and efficiency for the matrix math which dominates Neural Network computing.
We have a mission and clear roadmap to exploit the “unfair” advantages of photons over electrons for transport and beyond that for AI compute. In this mission, 3E8 partners with and builds on over a decade of cutting-edge research at George Washington University, McGill and other leading research institutions.