Breaking

Nvidia invests in inference chip startup d-Matrix to strengthen its AI infrastructure platform

October 11, 2026 at 15:00 EDT

Foundation Models
Infra & Chips
Sales & Business

Nvidia plans to invest in d-Matrix, a startup building AI chips optimized for inference workloads, signaling a strategic shift to lock in competing hardware makers within its own interconnect ecosystem. The investment, disclosed by The Information, comes weeks after d-Matrix announced integration of its Raptor processors into Nvidia's NVLink Fusion rack architecture, enabling alternative silicon to connect directly into Nvidia-based AI factories.

October 2026 · AI Infrastructure

Nvidia Backs Competing Inference Chip to Lock In Custom Silicon

As enterprises seek alternatives to GPU-only inference, Nvidia invests in d-Matrix and binds competing chips to its NVLink Fusion platform, maintaining control over the AI infrastructure layer.

$2B

d-Matrix Valuation

100

TB/s Memory Bandwidth (Raptor)

10×

Speed vs GPU (Claimed)

80–90%

Nvidia Market Share

Inference Efficiency Comparison

1×

GPU Baseline

10×

d-Matrix (Corsair)

d-Matrix claims 10× faster inference speed and 5× better energy efficiency for decode workloads

The Strategy: Ecosystem Lock-In

← Competing Chips

d-Matrix, Marvell, MediaTek

↓ Platform Layer

NVLink Fusion interconnect

↓ Nvidia Revenue

NVLink switches, BlueField DPUs, networking

Nvidia maintains infrastructure revenue regardless of which accelerator customers deploy, addressing market demand for inference alternatives while keeping competitors tethered to the Nvidia platform.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.