In November of last year, d-Matrix emerged from stealth, announcing a total of $154M raised and backing from more than 25 different companies. Their recent Series B round was led by Singapore’s Temasek, with support from M12, Microsoft’s venture fund.
As Karl Freund put it in Forbes, “d-Matrix uses a hybrid approach to memory that appears to deliver excellent results, using SRAM as ‘Performance Memory’ and a larger DRAM store for “Capacity Memory”. Use the Performance Memory for on-line operations that require low-latency for interactivity, and use the Capacity Memory for off-line work.”
In this edition of Founders Feature, we connected with Sid Sheth, Founder and CEO of d-Matrix, to learn more about the d-Matrix journey, his vision for the path forward, and how he builds a strong team. Here’s what he shared with us:
What inspired you to start this company?

Sid: d-Matrix was founded in 2019, when there was not yet much emphasis on inference compute. Most of the narrative centered around training AI models. At d-Matrix, we believe that inference would be the larger computing opportunity and focused our efforts on building the world’s most efficient inference compute solution at-scale. On the personal front, coming from a family of business-minded entrepreneurs in India, venturing out on my own was always on my mind. I was waiting for the right opportunity to present itself, and I felt the time had come.
Can you describe a significant pivot or change in strategy that your company has made?
Sid: d-Matrix has had several large and small pivots to get to where it is today. The one that I would highlight is the pivot we made in 2020 to emphasize generative Transformer-based workloads. It was early on, and we were looking for product-market fit. While we had emphasized vision (CNN) and Recommender-type workloads, we felt the market opportunity wasn’t quite there. GPT-3 was just getting launched and BERT was getting deployed in large use cases like search. Transformers seemed to be scaling well across size and modality, so we shifted to emphasize these classes of workloads and focused our software and hardware efforts from the ground-up with that in mind.
How do you see your company evolving in the next 5-10 years?
Sid: Our goal is to make gen AI inference commercially viable for everyone. The industry is now moving into the age of inference – with the arrival of reasoning models, Inference-time-compute and highly interactive video generated content. We believe this trend will carry the company for the next 5-10 years.
The only way we build value is by creating value for our customers. Over the past 5 years, we have purposefully scaled the company by taking this approach. Today we now have 200+ team members across 5 sites globally and have generated revenue the last 2 years. We will continue down this path of value creation across multiple customers, with multiple products. When our customers think of a partner that can help them profitably operationalize inference, they should think d-Matrix.
What qualities do you look for in your team members?
Sid: The chip business is hard, and certainly not for the faint-hearted. We have a saying at d-Matrix: “Stay humble, stay paranoid, and execute.”
I typically look for three things in my team members:
- Willingness to take feedback and learn quickly
- Someone who brings 100% to their craft everyday
- Persistence in the face of setbacks
How do you maintain work-life balance as a founder?
Sid: My work is my life, I don’t compartmentalize. However, I do always make time to change the context in which I view my work, and this provides a balanced perspective. Whether it is hiking in nature near large water bodies, meditating, listening to Indian classical music, or going to the gym with my sons – these all provide a setting to help clear the clutter and bring a fresh perspective to my work.