D-Matrix, which is building its chiplet-based Nighthawk architecture for faster matrix math calculations, raises a $44M Series A led by Playground Global
Context & Ripple Effects
D-Matrix's $44M Series A, led by Playground Global, funds Nighthawk, a chiplet-based architecture aimed at speeding up matrix math — the core operation in AI inference. The round fits Playground Global's stated strategy of backing hardware startups, and it plants a flag in workload-specific silicon just as rivals are forming: Lightmatter had earlier raised for AI-specific photonic chips, and MatX would later raise a ~$80M Series A for LLM training silicon.
The arc since has validated the bet: D-Matrix went on to a $110M Series B led by Temasek in 2023, then a $275M round at a $2B valuation in 2025, and by 2026 Nvidia itself was combining its hardware with d-Matrix's in a new system — a rare instance of the incumbent partnering with a chip rival.
First-order effects
- D-Matrix now has the capital to take Nighthawk from design toward silicon, targeting the inference side of AI compute rather than training.
- Playground Global converts its hardware-startup thesis into a marquee position in one of the few funded AI-chip challengers of 2022.
Second-order effects
- MatX's ~$80M Series A for LLM training chips shows the funding pattern spreading across the workload split — inference and training silicon each attracting dedicated capital.
- Specialized inference chips pressure Nvidia's all-in-one positioning, and the eventual Nvidia–d-Matrix system pairing shows the incumbent responding by co-opting a challenger rather than only competing.
Third-order effects
- If the pattern holds, AI compute stratifies into workload-specific silicon — training, inference, and interconnect specialists — with incumbents absorbing the best of them through partnerships or acquisitions.
- Chiplet-based architectures lower the barrier for startups to ship competitive AI silicon, shifting industry structure away from monolithic designs held by a few vendors.
The trend: Venture capital is funding a generation of workload-specific AI silicon startups, and incumbents like Nvidia are increasingly partnering with the survivors instead of competing head-on.