# Norman Jouppi Norman Paul Jouppi is an American electrical engineer and computer scientist, currently VP and Engineering Fellow at Google, known for leading the development of Tensor Processing Units (TPUs) and pioneering memory hierarchy designs. Norman Paul Jouppi is an American electrical engineer and computer scientist, currently serving as Vice President and Engineering Fellow at Google. He is best known as the technical lead for Google's Tensor Processing Units (TPUs) since their inception in 2013, which have become central to [artificial-intelligence](https://www.wikiprompt.org/wiki/artificial-intelligence) infrastructure. His career spans decades of contributions to computer architecture, including pioneering work on memory hierarchies and heterogeneous processing. Jouppi's research and engineering efforts have shaped both academic understanding and industrial practice in high-performance computing. He holds over 100 US patents and has received numerous awards, including the Eckert-Mauchly Award and the IEEE Seymour Cray Computer Engineering Award. ## Early Career and Education Jouppi earned his master's degree in electrical engineering from Northwestern University in 1980 and a PhD from Stanford University in 1984. During his doctoral studies, he was one of the computer architects at the MIPS Stanford University Project, an early RISC initiative led by John L. Hennessy. This work contributed to the foundational design of reduced instruction set computing, which later influenced many commercial processors. After completing his PhD, Jouppi joined Digital Equipment Corporation's Western Research Laboratory in 1984. He remained there through the company's transitions, working at Compaq and then Hewlett-Packard by 2002. At HP, he led the Advanced Architecture Lab at HP Labs in Palo Alto from 2006 to 2008, followed by the Exascale Computing Lab from 2008 to 2010 and the Intelligent Infrastructure Lab from 2010 to 2011. ## Contributions to Computer Architecture Jouppi pioneered several key developments in memory hierarchies. He introduced the concept of victim buffers, which reduce cache miss penalties, and prefetching stream buffers that anticipate data needs. He also advanced multi-level exclusive caching, a technique that avoids data duplication across cache levels. These innovations improved the performance of [neural-network](https://www.wikiprompt.org/wiki/neural-network) training and inference systems that rely heavily on memory bandwidth. He developed the CACTI simulator, a widely used tool for modeling cache time, area, and power. This simulator became standard in academic and industrial settings for evaluating memory designs. Jouppi also explored heterogeneous architectures, including single-ISA heterogeneous systems where cores with different capabilities share the same instruction set, enabling more efficient power and performance trade-offs. Throughout his career, he served as principal architect of four microprocessors and contributed to graphics accelerators. His work also extended to telepresence technology and the application of nanophotonics in computing, areas that promise faster data transmission and reduced energy consumption. ## Google and Tensor Processing Units In 2011, Jouppi joined Google as a computer engineer. In 2013, he became the technical lead for the Tensor Processing Unit project, which aimed to accelerate [machine-learning](https://www.wikiprompt.org/wiki/machine-learning) workloads. The first TPU was deployed in 2015 and used in Google's data centers for tasks like ranking search results and powering [deep-learning](https://www.wikiprompt.org/wiki/deep-learning) models. Subsequent generations of TPUs have been integrated into [google-cloud](https://www.wikiprompt.org/wiki/google-cloud) services, enabling external developers to train and run large models. The TPU architecture is optimized for matrix operations common in [transformer](https://www.wikiprompt.org/wiki/transformer) models, which underpin modern [large-language-model](https://www.wikiprompt.org/wiki/large-language-model) systems. Jouppi's leadership helped scale these chips from research prototypes to production systems that support billions of users. His work has been instrumental in making [generative-ai](https://www.wikiprompt.org/wiki/generative-ai) applications practical at scale. ## Awards and Recognition Jouppi received the Eckert-Mauchly Award in 2015 for contributions to the design and analysis of high-performance processors and computer storage systems. He was named a Hewlett Packard Fellow in 2002, an IEEE Fellow in 2003, and an ACM Fellow in 2007. The ACM awarded him the Alan D. Berenbaum Distinguished Service Award in 2013. In 2014, he received the Harry H. Goode Memorial Award and was elected to the National Academy of Engineering. He became a Fellow of the AAAS in 2019 and received the IEEE Seymour Cray Computer Engineering Award in 2024. From 2007 to 2011, he chaired ACM's SIGARCH, the special interest group on computer architecture. ## Academic and Professional Service Jouppi served as a consulting assistant or associate professor at Stanford University from 1984 to 1996. He has been on the editorial boards of Communications of the ACM and IEEE Computer Architecture Letters. His mentorship and service have influenced a generation of computer architects. His work continues to shape the intersection of hardware and [machine-learning](https://www.wikiprompt.org/wiki/machine-learning), with implications for [amd](https://www.wikiprompt.org/wiki/amd), [intel](https://www.wikiprompt.org/wiki/intel), and other chipmakers. His contributions to [cerebras](https://www.wikiprompt.org/wiki/cerebras) and [groq](https://www.wikiprompt.org/wiki/groq)-style accelerators are indirect, but his foundational research informs their designs. As of 2024, Jouppi remains active at Google, where his TPU leadership continues to drive advances in [artificial-intelligence](https://www.wikiprompt.org/wiki/artificial-intelligence) hardware. --- Source: https://www.wikiprompt.org/wiki/norman-jouppi License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/) Last updated: 2026-09-05T13:27:31.20523+00:00