d-Matrix has closed $275 million in Series C funding at a $2 billion valuation, bringing total capital raised to $450 million. The Santa Clara-based company will use the investment to advance its product roadmap and support large-scale deployments of what it claims is the world's highest-performing data center inference platform.
The oversubscribed round attracted investors from Europe, North America, Asia, and the Middle East. New participants include the Qatar Investment Authority and EDBI, while M12 (Microsoft's Venture Fund), Nautilus Venture Partners, Industry Ventures, and Mirae Asset provided follow-on investment.
The AI industry is entering what d-Matrix calls the "Age of AI Inference"—a period where running trained models at scale becomes more important than the training itself.
This isn't just a technical shift; it's a massive economic one. The market for AI inference is already huge—valued at $91.43 billion in 2024, according to a recent report from Fortune Business Insights. That same report projects the market will grow to $255.23 billion by 2032, showing just how much is at stake. As companies move from experimenting with AI to deploying it in production, they're hitting infrastructure bottlenecks that traditional GPU-based systems weren't designed to solve.
d-Matrix addresses this with a full-stack platform that combines compute-memory integration, high-speed networking, and inference-specific software. The company says its technology delivers 10× faster performance, 3× lower costs, and up to 5× better energy efficiency compared to GPU systems. Its Corsair inference accelerators, JetStream NICs, and Aviator software can produce up to 30,000 tokens per second at 2 milliseconds per token on a Llama 70B model.
"From day one, d-Matrix has been uniquely focused on inference," said Sid Sheth, CEO and co-founder. "We predicted that when trained models needed to run continuously at scale, the infrastructure wouldn't be ready. We've spent the last six years building the solution: a fundamentally new architecture that enables AI to operate everywhere, all the time."
The new funding will accelerate global expansion and support multiple large-scale deployments for hyperscale, enterprise, and sovereign customers. d-Matrix's compute-dense design allows customers to run models up to 100 billion parameters in a single rack, which the company says addresses growing AI sustainability concerns by enabling one data center to handle the workload of ten.
The capital will also advance d-Matrix's product roadmap, which includes 3D memory-stacking innovations. The company recently announced the d-Matrix SquadRack open standards-based reference architecture in partnership with Arista, Broadcom, and Supermicro, signaling its strategy to build an ecosystem around its technology.
d-Matrix was founded in 2019 by Sid Sheth (CEO) and Sudeep Bhoja (CTO), two industry veterans who saw this problem coming. Both came from Inphi Corporation (now Marvell), where Sheth built and scaled the cloud data center networking business and Bhoja served as CTO, developing key technology for data centers.
They made a contrarian bet: while the rest of the AI chip industry focused on training workloads, Sheth and Bhoja saw inference as the real long-term challenge. Six years later, that thesis is playing out as companies struggle to deploy AI models economically at scale.
The company now employs more than 250 people across offices in Santa Clara, Toronto, Sydney, Bangalore, and Belgrade. Its technology has attracted customers looking to run AI inference workloads more efficiently than current GPU-based infrastructure allows.
Investor confidence stems from d-Matrix's early focus on a problem that's only now becoming visible. The oversubscribed round was co-led by BullhoundCapital, Triatomic Capital, and Temasek. This group brings a powerful mix of expertise: Bullhound focuses on category-leading AI software, Triatomic is a founder-driven firm specializing in deep-tech computing, and Temasek is a major global investor known for scaling established companies.
Their backing validates the company's strategy as AI moves from research labs to production environments, where inference costs are becoming the dominant expense.
"As the AI industry's focus shifts from training to large-scale inference, the winners will be those who anticipated this transition early and built for it," said Per Roman, Founder of BullhoundCapital. "d-Matrix stands out not only for its technical depth but for its clear strategic vision."
"AI inference is becoming the dominant cost in production AI systems, and d-Matrix has cracked the code on delivering both performance and sustainable economics at scale," said Jeff Huber, General Partner at Triatomic Capital.
Michael Stewart, Managing Partner at M12, added context about the company's market position: "The explosion in AI inference demand shows us that efficiency and scalability can be key contributors to revenue capture and profitability... d-Matrix is the first AI chip startup to address contemporary unit economics in LLM inference."
Morgan Stanley served as exclusive placement agent, and Wilson Sonsini Goodrich & Rosati provided legal counsel.
