Skip to main content

d-Matrix

Category: AI Infrastructure

d-Matrix develops AI inference hardware and software platforms, focusing on making generative AI commercially viable through in-memory computing technology d-Matrix was founded in 2019. The company is led by Sid Sheth. Based in Santa Clara, United States (offices in Toronto, Sydney, Bangalore, Belgrade). Team size: 250+. Total funding raised: $450M. Latest round: Series C (November 2025). Key investors include ["Temasek","Playground Global","M12 (Microsoft)","Qatar Investment Authority","Singapore's EDBI","SK Hynix","Marvell Technology","Bullhound","Triatomic Capital"].

Founded
2019
Headquarters
Santa Clara, United States (offices in Toronto, Sydney, Bangalore, Belgrade)
Team size
250+
Total funding
$450M

Value proposition

Ultra-low latency batched inference for generative AI, combining memory-compute integration and in-memory computing to deliver cost-effective, high-performance AI solutions with 10x faster performance and 3x better energy efficiency than GPU systems

Products and solutions

["Corsair™ (world's most efficient AI computing platform for inference in datacenters)","JetStream™ (custom I/O card designed for AI inference workloads)","Aviator™ (inference-optimized software stack)","SquadRack™ (rack-scale solution for AI inference)"]

Unique value

Pioneering in-memory computing architecture for AI inference, eliminating data movement between memory and compute units to drastically reduce latency and energy consumption, enabling up to 30K tokens/second at 2ms/token on Llama 70B models

Target customer

Hyperscalers, enterprises deploying AI at scale, and data center operators requiring efficient AI inference solutions

Industries served

["Artificial Intelligence (Generative AI, Large Language Models)","Data Center Infrastructure","Cloud Computing","Enterprise AI Solutions"]

Technology advantage

Proprietary digital in-memory computing (IMC) technology enables 40% higher tokens per dollar compared to competitors like NVIDIA's B200 for inference tasks, with 10x faster performance and 3x better energy efficiency

How they differentiate

d-Matrix differentiates itself by utilizing in-memory computing architecture, which eliminates the traditional von Neumann bottleneck, enabling faster and more efficient AI inference compared to competitors relying on conventional GPU/CPU architectures

Main competitors

["NVIDIA","AMD","Intel","Cerebras","Groq","Graphcore"]

Key partnerships

["HTEC (strategic partnership for AI inference solutions)","GigaIO (collaboration to deliver inference solutions for enterprises)","Microsoft (funding and venture capital support)","Qatar Investment Authority and Singapore's EDBI (major investors in $275M Series C round)","Arista Networks, Broadcom, Supermicro (SquadRack partnership)"]

Notable customers

["Major hyperscalers via previous ventures of CEO Sid Sheth"]

Major milestones

["2025: Achieved $2B valuation after $275M Series C fundraising","2025: Closed oversubscribed Series C round led by global consortium","2024: Secured $110M+ in prior funding rounds","2019-2024: Built supply relationships with hyperscalers through previous ventures","Innovation: Pioneered in-memory computing techniques for AI inference"]

Growth metrics

$65.3M revenue in 2025, with 250+ employees

Market positioning

Positioned as a challenger to established players like NVIDIA in the AI inference market, with a focus on commercializing generative AI through novel in-memory computing solutions

Geographic focus

Global market focus with notable investor and regional ties in Singapore and the Middle East; operations in the United States, Canada, Australia, India, and Serbia

Patents and IP

No specific public patents listed; technological differentiator lies in proprietary IMC architecture and chiplet-based design

About Sid Sheth

20+ years in semiconductor industry, former SVP & GM at Inphi Corporation where he built a $1B+ networking business, serial entrepreneur with previous ventures supplying chips to major hyperscalers

Official website: