CEOInterviews.AI
Start App
Try: AI strategy tariffs earnings guidance layoffs interest rates growth

Sid Sheth

Founder, President and CEO, D-matrix

Search every verified Sid Sheth interview, podcast appearance, and on-the-record quote — each transcript cross-checked by AI and human review to confirm speaker identity. Sid Sheth, founder and CEO of d-Matrix, has recently discussed the company's focus on inference-specific computing and its expansion into systems and software. In September 2026, d-Matrix announced a partnership with Nvidia to integrate its next-generation Raptor XPU, described by Sheth as the world's first 3D-stacked DRAM-based XPU, into Nvidia's rack ecosystem using NVLink Fusion. Sheth stated the partnership aims to provide a faster path to market by leveraging Nvidia's existing infrastructure across data centers. He also noted d-Matrix completed two acquisitions in the past year, one on the software side and one on the system side, and emphasized that "the chip is not the product anymore" but rather the rack and accompanying software. In public appearances, Sheth has argued that AI inference is bifurcating into distinct markets, with a growing "premium token economy" where users pay significantly more for ultra-low-latency, real-time interactivity. He has described GPUs as optimized for throughput rather than latency, citing the memory-compute boundary as a core constraint, and positioned d-Matrix's memory-centric architecture as complementary to GPU-based systems. Sheth has also discussed disaggregated inference techniques, including attention feed-forward disaggregation (AFD), and mentioned a demo with Nvidia Hopper GPUs achieving over 800 tokens per second per user. He has spoken about the company's internal mandate to adopt AI across its organization and has used AI tools, such as Claude, as a sounding board for M&A strategy.

Recent appearances

  • The Path to Infinite Inference presented by d Matrix

    October 13, 2026 · 📺 Open Compute Project

    Sid Sheth CEO, Founder - D-Matrix The world faces insatiable demand for AI inference as compute becomes constrained by cost, latency, and energy consumption. Todays infrastructure was not designed to deliver intelligence at global scale. Sid Sheth, founder and CEO of d-Matrix, will lay out the path to infinite inference AI infrastructure that is fast, efficient, and scalable. He will explain why the memory-compute boundary is a core constraint and how purpose-built architectures can bring compute directly into memory. Sid will share how heterogeneous deployments combining purpose-built infe…

    Read the verified transcript → Watch on YouTube ↗

  • SBT AI Infra Summit Hardware Panel with Sid Sheth, Anton McGonnell, and Naveen Verma

    September 22, 2026 · 📺 SBT Industries

    SBT Industries was honored to host a panel during AI Infra Summit 2026 featuring the insights of three respected semiconductor industry leaders: Sid Sheth, CEO of d-Matrix; Anton McGonnell, VP of Product at Sambanova; and Naveen Verma, CEO of EnCharge AI. In this wide-ranging discussion moderated by SBT's President, Justin Kinsey, the group discusses the realities of building full-stack solutions to meet the growing demands of modern datacenters and hyperscalers, looking ahead to the changing landscape of LLMs for inference, and how these leaders are building best-in-class teams. Here are th…

    Read the verified transcript → Watch on YouTube ↗

  • #AIInfraSummit26: d-Matrix's 3D Stacked XPU & NVIDIA Partnership

    September 20, 2026 · 📺 NextGenInfra

    Check out videos from the #AIInfraSummit 2026 in Santa Clara https://ngi.fyi/AIinfra26yt Sid Sheth, Founder & CEO of d-Matrix, discusses his company's evolution in the inference computing market and reveals a major strategic partnership that transforms how d-Matrix brings its technology to market. Sheth explains how d-Matrix addresses the growing demand for ultra-low latency AI computing through silicon, subsystems, and software solutions. He shares insights into the company's aggressive expansion strategy, including two key acquisitions in its first year, and details the shift from chip-cent…

    Watch on YouTube ↗

  • d-Matrix Plugs Into Nvidia’s AI Ecosystem

    September 10, 2026 · 📺 Bloomberg Tech

    Nvidia is embracing the rise of custom AI chips as demand shifts toward specialized systems built for inference. d-Matrix CEO Sid Sheth discusses the startup’s new partnership with Nvidia, which will integrate its inference-focused XPUs with Nvidia GPUs using NVLink Fusion, and why inference is emerging as the next major battleground in AI compute. He joins Ed Ludlow on "Bloomberg Tech." -------- Like this video? Subscribe to Bloomberg Technology on YouTube:    / @bloombergtech     Watch the latest full episodes of "Bloomberg Technology" with Caroline Hyde and Ed Ludlow here:    • 'Bloom…

    Watch on YouTube ↗

  • Why People Are Paying 10x More for AI -and What That Means for the Chip Market | Sid Sheth, d-Matrix

    August 17, 2026 · 📺 Eye on AI

    Everyone is talking about the AI chip market like it's one thing. D-Matrix's CEO says it's splitting in two, and the half that matters most right now is the one NVIDIA isn't built for. In this episode, Craig Smith sits down with Sid Sheth - Founder, President & CEO at d-Matrix Corp - to unpack one of the least understood dynamics in the AI infrastructure story: the emergence of what Sid calls the "premium token economy", a new tier of AI inference where users pay ten times more per million tokens in exchange for instant, real-time interactivity. The conversation explains why general-purpose…

    Read the verified transcript → Watch on YouTube ↗

  • The New Era Of AI Inference Memory | Sid Sheth, CEO D-Matrix

    August 11, 2026 · 📺 Under The Banyan Tree With Sam Awrabi

    Sid from D-Matrix joined Under The Banyan Tree podcast to break down the future of AI inference and memory-centric computing. This episode explores how stacking memory and using a chiplet architecture can solve the massive AI memory bottleneck. Sid also shares his entrepreneurial journey, explains key memory types like SRAM and DRAM, and reveals how his team is building highly efficient hardware optimized for the next wave of AI compute. 00:00 Hook & Intro 01:40 What is d-Matrix? 07:15 A Family of Entrepreneurs 13:01 Leaving a Comfortable Job for a Startup 21:24 AI Memory Explained: SRAM, DRA…

    Read the verified transcript → Watch on YouTube ↗

  • Fast, Efficient Inference at Scale with Heterogeneous Hardware | Gimlet Labs x d-Matrix | RAISE 2026

    July 23, 2026 · 📺 RAISE Summit

    Zain Asgar, Co-Founder & CEO of Gimlet Labs, and Sid Sheth, CEO of d-Matrix, on fast, efficient inference at scale across heterogeneous hardware — and why inference efficiency is the new battleground. Recorded live on the Ada Lovelace Stage at the Carrousel du Louvre, Paris, on July 8, 2026. RAISE Summit is Europe's leading AI gathering, bringing together 9,000+ founders, executives, investors, researchers and policymakers. Learn more: https://www.raisesummit.com Subscribe for every session from RAISE Summit 2026.

    Read the verified transcript → Watch on YouTube ↗

  • Fireside Chat with Sid Sheth, Founder & CEO of d-Matrix | RAISE Summit 2026

    July 22, 2026 · 📺 RAISE Summit

    Sid Sheth, Founder & CEO of d-Matrix, joins Tony Kim of BlackRock on why inference efficiency is becoming the defining competitive battleground in AI hardware. Recorded live on the Grace Hopper Stage at the Carrousel du Louvre, Paris, on July 8, 2026. RAISE Summit is Europe's leading AI gathering, bringing together 9,000+ founders, executives, investors, researchers and policymakers. Learn more: https://www.raisesummit.com Subscribe for every session from RAISE Summit 2026.

    Read the verified transcript → Watch on YouTube ↗

  • Sudeep Bhoja & Sid Sheth, d Matrix | RAISE Summit 2026

    July 16, 2026 · 📺 pessalou

    How do you deliver faster AI inference without replacing your GPU infrastructure? At the RAISE Summit 2026, d-Matrix CEO Sid Sheth and CTO Sudeep Bhoja join theCUBE to discuss the next phase of AI infrastructure and why heterogeneous inference is becoming essential for production AI. In this interview, they explain how d-Matrix Corsair™ inference accelerators work alongside NVIDIA GPUs to accelerate token generation, reduce latency, and improve the economics of large-scale AI inference. The discussion covers speculative decoding, rack-scale deployment, memory bandwidth, fast token generation…

    Watch on YouTube ↗

  • Sid Sheth, Founder & CEO, d-Matrix, shares his experience at TiEcon 2026

    June 01, 2026 · 📺 TiE Silicon Valley

    At the Media Lounge during TiEcon 2026, Sid Sheth shared his perspective on AI, the future of computing, and the opportunities shaping the next wave of technology. 🎤 Speaker: Sid Sheth – Founder & CEO | d-Matrix #TiEcon2026 #MediaLounge #TiESiliconValley Visit https://tiesv.ai/ Follow us on   / tiesv   Follow us on   / tiesiliconvalley   Like us on   / tiesiliconvalleychapter   Tweet us at https://x.com/tiesv

    Watch on YouTube ↗

  • How A Semiconductor Veteran Is Betting On AI

    May 22, 2026 · 📺 Supercompanies with Greg Shove

    AI costs are dropping fast. So why isn't your bill going down? d-Matrix CEO Sid Sheth has the answer. Sheth started building AI infrastructure in 2019, years before ChatGPT made the rest of the world pay attention. On this episode of Supercompanies, he joins host Greg Shove to explain why he bet early on AI inference, and why he made his own co-founder the company's "AI czar." Sheth breaks down a counterintuitive economic reality: even as the cost of AI keeps dropping, the price of a "unit of intelligence" is staying flat, because the models keep getting smarter just as fast as they get c…

    Read the verified transcript → Watch on YouTube ↗

  • Tech Podcast: D-Matrix CEO Sid Sheth on AI Test Time Scaling | AI With Sally

    March 05, 2026 · 📺 EE Times

    In this podcast conversation, Sally chats with D-Matrix CEO Sid Sheth about test time scaling and what it means for the world of AI hardware. every time. 🎙 Listen to more EE Times On Air podcasts on power, embedded, AI, and more: https://www.eetimes.com/podcasts/?utm... 🎥 Don't miss an episode and subscribe to our YouTube channel:    / @eetimes747   Stay up to date with the latest news in the electronics industry ➡️ EE Times: https://www.eetimes.com/?utm_source=y... ➡️ Power Electronics News: https://www.powerelectronicsnews.com/... ➡️ Embedded: https://www.embedded.com/?u…

    Read the verified transcript → Watch on YouTube ↗

Verified interview transcripts

Notable quotes

“We live in a world where resources are finite, right? So we have limited amount of energy. You know there's only so many data centers that we can build and capital is scarce, right? So how do we solve this problem? By not brute forcing it, right? How do we build a solution that can do more with less? And that is what we've been focusing on at Dmatrix. We call it the world of infinite inference with no latency. We wan”

“We started 2019, we came from a background building really good silicon, and the world was very different back then. The concept of an ecosystem didn't even exist because people didn't know what the ecosystem was. Inference didn't exist in 2019, we spent so much time just educating people on what inference is. The training ecosystem did exist, right, and mostly Nvidia at the time, but everything changed in the last f”

“There were really two watershed moments in the last four years around AI inferencing specifically. Training has been around for a while, starting 2015, and that was a world in which GPU compute dominated a lot. Then we had the ChatGPT watershed moment end of 2022, and we really for the first time saw the transition from pre-training training to inference. People first started worrying about what does it take to deplo”

“The inferencing space specifically is kind of bifurcating, because I think everyone looked at inference as one singular entity, but there is lots of nuance when it comes to inference, it's not a one-size-fits-all. Depending on where you are doing the inferencing, what your target markets are, what applications you're going after, you can't really build a single chip that serves all of inferencing's needs. GPUs, just ”

“It really comes down to the margin structure at an OpenAI. They are running businesses, right? At the end of the day, they're not building models to release into the open community; they're actually building an application software business on top of that, and there is a huge premium. So when they talk about token pricing, there is a margin built into that. But somebody like a Moonshot or a Zhipu doesn't need to, the”

“I think this is just volume economics at work. This is what it looks like when you have a new market with infinite demand. I always say that the inferencing market has infinite demand because there is no cap on the amount of intelligence that we as humans would like to get access to. So let's assume that the size of this market, there is no cap on it, and okay, if you are serving a market of that size, then what kind”

The summary and quote tags on this profile are produced with AI assistance from verified, first-person interview transcripts, then checked by our team to confirm the speaker's identity and the accuracy of every quote. See how we verify →