Built for AI.
Engineered for Speed.
DreamBig; A silicon platform powering high-performance networking for AI at any scale.
• Our Company
Networking, from the edge to the datacenter.
DreamBig is a networking and chiplet platform company, specializing in AI-native applications, from the datacenter, to automotives and robotics. Our state of the art NIC, Mercury, delivers the fastest throughput speed at unprecedented scale on the planet, while the Mars Chiplet Platform supports both standard and die-to-die interfaces like UCIe and BoW with standardized protocols, giving you the freedom to build the solutions you need.
• Our Impact
The Fastest, the Smartest, and the Most Scalable.
The highest RDMA throughput of any solution on the market.
#1
Future-proof Scaling from 800Gbps to 12.8 Tbps+. The first RDMA architecture defined in the Al era for multiple transports.
800Gbps to
25.6 Tbps+
Max GPU Cluster size: 16 million GPUs operating at 40%+ less power, for optimized efficiency.
16m
• Our Products
Introducing Mercury:
Go faster.
The Mercury SuperNIC delivers 12.8 Terabits throughput per second, and removes scale-out bottlenecks. Your GPUs function at 100% capacity, not 15 or 40%. It’s completely interoperable with NVIDIA, AMD or whatever chips you’re using, and Pytorch and CUDA compatible.
01
Mercury NIC
Enabling up to 12.8 Terabits per second, and removing scale-out bottlenecks, allowing your GPUs to function at 100% capacity, not 15 or 40%.
01
Mars Platform
3D stacked HBM enabling 20 nanosecond round trip time, 80% lower power use, and infinite customizability. Build products in months, not years.
03
Custom Design
Build your own system using the most advanced RDMA on the planet, around the specifications you want.
• Resources
