D-Matrix, which is building its chiplet-based architecture Nighthawk for faster matrix math calculations, raises a $44M Series A led by Playground Global
Context & Ripple Effects
This April 2022 round is the opening move of what became one of the longer-running challenges to Nvidia's dominance in AI silicon. The $44M from Playground Global — a firm whose stated mandate is backing hardware startups — funded Nighthawk, a chiplet-based architecture aimed specifically at matrix math, the core operation of neural networks.
The bet aged well on paper: D-Matrix narrowed its focus to inference processing by the time of its $110M Series B led by Temasek in 2023, then raised $275M at a $2B valuation in late 2025, taking total funding to $450M. The endgame so far is coexistence rather than displacement — by mid-2026 Nvidia was combining its hardware with d-Matrix's chips in a shared system.
First-order effects
- D-Matrix gets the capital to take Nighthawk from architecture to silicon, targeting the matrix-multiplication workload that dominates AI computation.
- Playground Global converts its hardware-startup thesis into a marquee position in a company that becomes a repeat fundraiser.
Second-order effects
- The raise helps legitimize inference-specialized silicon as a fundable category, a lane MatX entered two years later with a ~$80M Series A for LLM-training chips founded by ex-Google chip alumni.
- Nvidia's calculus shifts over time from pure competition to selective partnership, as the 2026 joint system with d-Matrix shows — incumbents can absorb challengers' innovations without acquiring them.
Third-order effects
- If the pattern holds, AI compute stratifies into specialized players — training versus inference, chiplets versus monoliths — with the incumbent anchoring ecosystems through partnerships rather than owning every layer.
- Venture capital keeps underwriting multi-hundred-million-dollar chip bets against an entrenched incumbent, making semiconductor startups a durable asset class rather than a one-cycle phenomenon.
The trend: AI silicon is fragmenting into well-funded inference and training specialists whose eventual relationship with Nvidia is partnership as often as rivalry.