What becomes possible when we scale computing vertically rather than relying on ever-smaller transistors?

This talk introduces PhantaField’s approach to connecting atomic-scale materials engineering with AI chip architecture—using low-temperature two-dimensional semiconductors to bring memory and computation closer together. 

The talk begins with NanoGalaxy, PhantaField’s 12-inch plasma- and photon-enhanced metalorganic chemical vapor deposition (MOCVD) platform. Designed to grow two-dimensional transition-metal dichalcogenides, including MoS₂, at temperatures below 450 °C, NanoGalaxy combines plasma activation and in-situ ultraviolet illumination to support low-temperature growth and defect control. We will discuss how this approach addresses the thermal constraints of back-end-of-line integration and opens a path to building successive layers of active devices above existing circuitry. We will then introduce Sophon, PhantaField’s monolithic 3D AI accelerator architecture. Its 64-tier reference design interleaves logic and memory above a mature 28 nm silicon base, targeting 330 GB of embedded DRAM and 2.1 PB/s of local weight bandwidth.

Connecting materials, manufacturing, and architecture, the presentation will examine the device-performance requirements, integration challenges, and potential system-level benefits of this approach. The central theme is a path toward brain-scale intelligence through atomic-scale technology—not simply by scaling with more electricity, but by rethinking how computation and memory are built together. 

Read More: A subthermionic tunnel field-effect transistor with an atomically thin channel
     PhantaField PFG-1 Sophon Whitepaper

 

Speaker:

Dr. Xuejun Xie

Founder & CEO                                                    PhantaField Inc

 AGENDA: