Jonathan Ross

Jonathan Ross holds an absolutely elite, unprecedented background in advanced semiconductor design and massive-scale AI hardware architecture. Having spent years at Google aggressively designing the highly secretive Tensor Processing Unit (TPU), this incredibly specialized, deeply technical background in optimizing massive matrix-multiplication operations provided the exact engineering blueprint required to invent Groq's revolutionary Language Processing Unit.

Jonathan Ross is the co-founder and Chief Executive Officer of Groq, Inc. While at Google in 2013, Ross incubated and developed the original Google TPU as a '20% project', creating the specialized accelerator chip that powered AlphaGo's historic victory over Lee Sedol and became Google's primary AI compute engine. Recognizing that future AI systems would be bottle-necked by non-deterministic GPU memory architectures during real-time inference, Ross departed Google in 2016 to found Groq. As CEO, Ross invented the Language Processing Unit (LPU), raised over $1.0 billion in institutional capital from BlackRock and Cisco, scaled GroqCloud to over 500,000 developers, and built Groq into a $2.8 billion AI silicon leader generating over $100 million in annual revenue. Ross's technical legacy is immortalized through his invention of the Google TPU, which transformed machine learning acceleration across Google Search, Translate, and DeepMind's AlphaGo. At Groq, Ross directs overall corporate strategy, semiconductor roadmaps, and global cloud infrastructure alliances. A vocal advocate for software-defined silicon and open foundation models, Ross has positioned Groq as the essential high-speed execution layer powering real-time conversational intelligence worldwide. At Google, Ross observed firsthand that computing architectures were reaching a critical inflection point where memory communication was outstripping arithmetic processing capacity. His design for the TPU proved that domain-specific architectures could transform computing. At Groq, Ross took this insight to its ultimate conclusion, establishing a software-first engineering discipline that eliminated hardware speculation, resulting in the most predictable, low-latency AI inference engine ever built.