Inside Story of Building the World’s Largest AI Inference Chip | Cerebras CEO & Co-Founder Andrew Feldman
with Andrew Feldman — CEO
Andrew Feldman, Co-Founder & CEO of Cerebras Systems, reveals the inside story of building the world's largest AI chip. Learn how their wafer-scale architecture challenges NVIDIA by delivering unprecedented speed for AI inference.
5 Things You'll Learn from This Episode
- Understand why a new compute workload, like AI, creates massive opportunities for new computer architectures to emerge and challenge incumbents.
- Learn how Cerebras's wafer-scale chip overcomes GPU memory bottlenecks by placing massive amounts of fast SRAM memory directly on-chip.
- Discover why AI inference is exploding and becoming the dominant cost and performance challenge, shifting focus away from just training.
- Analyze the strategy of selling a complete, integrated system rather than just a component chip to capture more value and solve customer problems.
- Recognize that in deep tech, aiming to be 10-20x faster is a better strategy than being 30-40% cheaper when competing with giants.
About the Episode
Andrew Feldman shares the origin story of Cerebras Systems, from their 2015 thesis that a new AI workload demanded a new chip architecture to building the largest chip ever made. He breaks down the technical advantages of their wafer-scale design for AI inference, the business strategy required to compete with giants like NVIDIA, and why the explosion in AI usage makes inference, not training, the new frontier.
Timestamps
- 0:00 — Introduction: The $3 Trillion Race for AI Compute
- 3:45 — The Thesis for Starting a Chip Company in 2015
- 6:21 — When a New Workload Creates New Opportunity
- 9:05 — Building the World's Largest Chip
- 14:15 — How Wafer-Scale Architecture Solves the Memory Bottleneck
- 18:01 — Why AI Inference is Exploding
- 21:20 — Competing with Giants like NVIDIA and AMD
- 26:48 — The Myth of Compute as a Commodity
- 33:05 — Cerebras's Product Strategy: Systems, Cloud, and On-Prem
- 40:45 — The Rise of Specialized 'Neo-Cloud' Providers
- 46:30 — Why Cerebras Sells Systems, Not Just Chips
- 52:00 — How the Rise of Inference Weakens NVIDIA's Cuda Moat
- 57:15 — The Strategic Partnership with G42
- 59:50 — Discovering the Power of Inference
About Andrew Feldman
Andrew Feldman is the Co-Founder and CEO of Cerebras Systems, a company dedicated to accelerating artificial intelligence compute. A serial entrepreneur, Andrew previously founded SeaMicro, a pioneer in low-power server technology, which was successfully acquired by AMD.
Co-Founder & CEO at Cerebras Systems
About the Host
Nataraj Sindam is the creator of The Startup Project, a podcast featuring founders, investors, and operators building the future.
→ Website
Subscribe to Startup Project for more founder conversations!
→ Spotify
#StartupProject #AndrewFeldman #Cerebras #AIChips #Semiconductors #NVIDIA #AIInference #DeepTech #VentureCapital #WaferScale #AICompute #Hardware