Product Launchin memory computeinference acceleratord matrix
D-Matrix Builds In-Memory Inference Chip Corsair
8.1
D-Matrix, which recently raised a $275 million funding round, is launching Corsair and Jetstream inference hardware that uses Digital In-Memory Compute and chiplet scaling to reduce memory bandwidth bottlenecks for transformer models. The platform pairs DIMC-powered Jetstream PCIe cards with a RISC-V control core and precompiled kernels, targeting sub-frontier and on-prem inference workloads and planned Triton integration.
Key Points
- 1Implements Digital In-Memory Compute to perform matrix multiplications directly inside memory cells.
- 2Breaks memory bandwidth bottlenecks for transformer inference, reducing weight movement and latency.
- 3Enables cost-efficient, scalable on-prem and edge inference via Jetstream cards and chiplet scaling.
Scoring Rationale
Strong hardware innovation and $275M funding, but incremental versus established accelerators and limited proven market deployment.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
