d-Matrix secured $275 million in Series C funding to advance its full-stack inference platform for hyperscale and enterprise data centers. The round, which values the company at $2 billion, was led by Bullhound Capital, Triatomic Capital, and Temasek, with participation from QIA, EDBI, and Microsoft’s M12 venture fund. The investment brings total funding to $450 million as d-Matrix scales global deployments of its Corsair inference accelerators, JetStream networking NICs, and Aviator software suite.
Founded in 2019, d-Matrix has focused exclusively on AI inference—the stage where trained models run continuously at scale. Its platform integrates compute and memory in a single architecture to deliver up to 10× performance, 3× lower cost, and up to 5× energy efficiency over GPU-based systems. On a Llama 70B model, the platform achieves 30,000 tokens per second with 2 ms latency per token, allowing 100B-parameter models to run in a single rack. The company’s new SquadRack reference architecture—developed with Arista, Broadcom, and Supermicro—extends its open ecosystem approach.
Sid Sheth, CEO and co-founder of d-Matrix, said, “When we started d-Matrix six years ago, training was seen as AI’s biggest challenge, but we knew that a new set of challenges would be coming soon. We’ve spent the last six years building the solution: a fundamentally new architecture that enables AI to operate everywhere, all the time.”
• Series C: $275 million led by Bullhound Capital, Triatomic Capital, and Temasek
• Valuation: $2 billion | Total funding: $450 million
• HQ: Santa Clara, CA | Global offices in Toronto, Sydney, Bangalore, and Belgrade
• Core Products: Corsair inference accelerators, JetStream NICs, Aviator software stack
• Performance: 30 K tokens/s at 2 ms latency on Llama 70B; 100 B parameters in one rack
d-Matrix Portfolio Highlights
🌐 Analysis:
The funding underscores how AI inference has become the next battleground in AI infrastructure as training hardware saturates hyperscaler budgets. d-Matrix’s compute-in-memory architecture tackles the latency and power limits that GPUs face in serving massive language models. The company’s ecosystem alliances with Arista and Broadcom link it to key networking and silicon supply chains, while backing from Microsoft’s M12 suggests future alignment with Azure AI deployments.
🌐 We’re tracking the latest developments in networking silicon. Follow our ongoing coverage at: https://convergedigest.com/category/semiconductors/
🌐 We’re launching the “Data Center Networking for AI” series on NextGenInfra.io and inviting companies building real solutions—silicon, optics, fabrics, switches, software, orchestration—to share their views on video and in our expert report. To get involved, send a note to jcarroll@convergedigest.com or info@nextgeninfra.io.
