Get Your Own Businessabc AI Visibility ID
Boost AI visibility, credibility, and discoverability
Michael Kagan is the Chief Technology Officer (CTO) of NVIDIA, with more than four decades in microprocessor architecture, high-performance computing, data centre networking, cloud infrastructure, and artificial intelligence.
Michael Kagan is the Chief Technology Officer (CTO) of NVIDIA. His career spans more than four decades, covering microprocessor architecture, high-performance computing, data centre networking, cloud infrastructure, and artificial intelligence.
At NVIDIA, Michael oversees technology architecture across accelerated computing platforms, integrating GPUs, CPUs, networking systems, and data processing units (DPUs).
His work includes AI infrastructure, large-scale computing systems, and technologies supporting digital twins and simulation.
Michael, previously, co-founded and was the former CTO of Mellanox Technologies, where he worked on InfiniBand and high-speed Ethernet networking technologies used in supercomputing and cloud data centres.
The networking and semiconductor company was later acquired by NVIDIA for approximately $7 billion.
During the early years of his career, Michael worked at Intel, contributing to the Intel i860XP vector processor and leading the design of the Pentium MMX processor.
Michael holds a bachelor's degree in electrical engineering from the Technion – Israel Institute of Technology.
He also serves on the Board of Governors of the University, and also received the 2026 Technion Alumni Medal.
Michael Kagan was born in Leningrad, Soviet Union (now Saint Petersburg, Russia), in 1957. His parents were physicists, and he developed an interest in engineering and mechanical systems during childhood. Following his secondary education, Kagan initially sought admission to a technical university in the Soviet Union but was unsuccessful. In interviews, he has attributed this rejection to discrimination against Jewish applicants. In 1975, at the age of 18, he immigrated to Israel with his family.
After arriving in Israel, Kagan studied Hebrew and subsequently enrolled at the Technion – Israel Institute of Technology in Haifa. He studied electrical engineering and computer architecture from 1976 to 1980, graduating with a Bachelor of Science degree in Electrical Engineering. During his university years, he supported himself through part-time employment. Following graduation, he served in the Israel Defense Forces before beginning his professional engineering career.
In August 1983, Kagan joined Intel Corporation at its engineering facility in Haifa, Israel, as a design engineer. His first assignment involved an experimental vector processor project known as N3, which was eventually discontinued. He subsequently transferred to Intel's operations in Hillsboro, Oregon, where he worked on floating-point processing technology associated with the Intel 80387 family of mathematical coprocessors.
After returning to Israel, Kagan worked on the architecture of the Intel i860XP, a high-performance vector processor designed for computationally intensive applications. The processor demonstrated substantial computational capabilities, but the product did not achieve sustained commercial adoption. Kagan later identified the limited software ecosystem surrounding the processor as an important lesson in computer architecture, arguing that hardware performance alone was insufficient without supporting software and development tools.
Between 1986 and 1992, Kagan held senior architecture responsibilities at Intel, contributing to processor design and development. From 1992 to 1995, he managed the design of the Pentium MMX processor, which incorporated multimedia-oriented instruction extensions into Intel's mainstream processor architecture. The Pentium MMX became one of Intel's commercially deployed microprocessor products during the expansion of personal computing in the 1990s.
Kagan subsequently assumed broader architectural and management responsibilities within Intel. He served in senior positions associated with processor development and the company's Basic PC product group, eventually becoming Chief Architect. His responsibilities included processor architecture, engineering coordination, and the development of computing technologies intended for commercial markets. He remained at Intel until April 1999, completing approximately 16 years with the company.
In April 1999, Kagan left Intel to co-found Mellanox Technologies alongside Eyal Waldman and seven other co-founders, many of whom had previously worked at Intel. Mellanox was established in Israel with the objective of developing high-performance networking and interconnect technologies for computing and storage systems. Kagan joined as Vice President of Architecture, a position he held until January 2009.
At Mellanox, Kagan contributed to the architecture and development of InfiniBand networking products. InfiniBand was designed to provide high-bandwidth, low-latency communication between computing systems, enabling large numbers of processors and servers to operate together. These technologies became relevant to high-performance computing environments, scientific research institutions, and enterprise computing systems.
One early application of Mellanox networking technology was the Virginia Tech supercomputer project in 2003. The system, known as System X, connected approximately 1,000 Apple computers into a computing cluster using high-performance networking. It became one of the highest-ranked supercomputers at the time, demonstrating the potential of constructing large-scale computing systems from commercially available computing hardware.
Mellanox subsequently expanded beyond supercomputing into enterprise computing and cloud infrastructure. Its networking technologies were adopted in database systems, storage environments, and large-scale data centres. The company also developed products that incorporated capabilities associated with InfiniBand into standard Ethernet networking, allowing cloud operators to benefit from high-performance communication technologies while retaining Ethernet-based infrastructure.
During this period, Kagan was involved in developing networking architectures intended to support distributed computing, where computational resources could be shared across large numbers of connected machines. The company's products included network adapters, switches, interconnect systems, and technologies for accelerating communication between processors and storage resources.
In January 2009, Kagan became Chief Technology Officer and Vice President of Architecture at Mellanox. In this role, he was responsible for the company's technology direction and architectural development. His work covered high-speed networking, data centre interconnects, network acceleration, and infrastructure technologies supporting cloud computing.
As cloud services and accelerated computing expanded, Mellanox began collaborating with NVIDIA on high-performance computing systems. NVIDIA's GPUs provided parallel computational capabilities, while Mellanox networking products enabled communication between large numbers of processors. The two companies worked together on supercomputing and accelerated computing infrastructure for approximately a decade before their merger.
In March 2019, NVIDIA announced an agreement to acquire Mellanox Technologies for approximately $6.9 billion. The transaction was completed in April 2020. Following the acquisition, Kagan joined NVIDIA and was appointed Chief Technology Officer in May 2020.
His appointment expanded his responsibilities from networking architecture to the integration of NVIDIA's computing technologies across hardware, software, and systems. At NVIDIA, his work has focused on accelerated computing infrastructure, including the interaction between graphics processing units (GPUs), central processing units (CPUs), data processing units (DPUs), networking technologies, and software platforms.
A central component of Kagan's architectural approach is treating the entire data centre as a single computing system rather than a collection of independent servers. This approach involves connecting large numbers of processors through high-performance networking and coordinating their operation through software. It is particularly relevant to artificial intelligence workloads, where training and running large models can require computation distributed across thousands of GPUs.
Kagan has discussed the importance of two complementary approaches to expanding computational capacity: scale-up and scale-out. Scale-up involves connecting multiple processors into larger computing systems through technologies such as NVIDIA NVLink, while scale-out involves connecting multiple computing systems through high-performance networks. Both approaches require management of communication bandwidth, latency, and the variation in communication delays across connected components.
He has also worked on NVIDIA's BlueField data processing units, which provide dedicated processing capabilities for networking, storage, infrastructure management, and security functions. By separating infrastructure processing from application processing, DPUs can reduce workloads on CPUs and GPUs while providing additional isolation between computing environments.
Other technologies associated with NVIDIA's computing infrastructure include InfiniBand, Spectrum-X Ethernet networking, NVLink, and NVIDIA's CUDA software platform. Kagan has emphasized the relationship between processor performance, networking efficiency, and software compatibility in determining the overall performance of accelerated computing systems.
In a March 2022 article published on NVIDIA's official blog, Kagan discussed the development of digital twins for data centres using NVIDIA Omniverse. These virtual representations allow engineers to evaluate system configurations, software changes, resource utilization, energy consumption, and potential equipment failures before implementing changes in physical infrastructure. He also described applications of digital twin technologies in industrial environments, autonomous systems, climate modelling, and smart cities.
In a December 2023 interview with EE Times Europe, Kagan discussed NVIDIA's approach to building accelerated computing platforms for artificial intelligence. He described the transition from conventional processor development toward computing architectures designed to support large-scale AI models, cloud services, and high-performance applications. The interview also addressed NVIDIA's Grace Hopper architecture, which integrates CPU and GPU technologies through high-bandwidth interconnects.
Kagan has discussed the computational requirements of generative AI, particularly the differences between model training and inference. He has explained that inference workloads involve repeated computation during the generation of text, images, and other outputs, and that reasoning models introduce additional computational requirements. His technical discussions have also addressed the distinction between prefill and decoding operations, which place different demands on processing capacity and memory bandwidth.
In subsequent interviews, including a conversation recorded at Sequoia Capital's Europe100 event, Kagan examined the engineering challenges associated with AI clusters containing 100,000 or potentially one million GPUs. These challenges include network congestion, communication latency, hardware reliability, workload scheduling, energy consumption, and cooling requirements. He has explained that as computing systems increase in scale, failures of individual components become unavoidable, requiring hardware and software architectures capable of maintaining operations despite equipment faults.
Kagan has also discussed the increasing energy requirements of AI infrastructure and the technical implications of connecting computing resources across geographically separated data centres. His comments have addressed the physical constraints imposed by communication distances, electricity availability, heat dissipation, and the need for efficient cooling technologies.
Alongside his work in computing infrastructure, Kagan has addressed NVIDIA's activities in simulation and scientific computing. These include NVIDIA Omniverse, autonomous vehicle simulation, and Earth-2, an initiative to develop digital representations and simulation capabilities for weather and climate research. He has also discussed potential applications of AI in computational biology, medical research, drug discovery, and the identification of physical relationships from scientific data.
In a November 2025 interview with Calcalist, Kagan discussed NVIDIA's development as an AI infrastructure company, the integration of Mellanox technologies into its computing systems, and the company's expansion in Israel. He also addressed his working relationship with NVIDIA founder and CEO Jensen Huang, the importance of engineering talent, and the company's approach to organisational development.
In the same interview, Kagan discussed NVIDIA's partnership with Intel, describing the potential integration of conventional general-purpose computing with accelerated computing technologies. He also expressed the view that artificial intelligence could become a widely used technological utility, comparable in its accessibility to electricity and other forms of infrastructure.
Beyond his executive responsibilities, Kagan has served as a member of the Board of Governors of the Technion – Israel Institute of Technology since May 2017. His association with the institution includes activities connecting engineering education, research, and the technology industry.
As of October 2026, Kagan continues to serve as NVIDIA's Chief Technology Officer, with responsibilities spanning computing architecture, AI infrastructure, accelerated networking, and the development of technologies supporting large-scale artificial intelligence systems.
Michael Kagan's stated vision focuses on developing computing infrastructure capable of supporting increasingly complex artificial intelligence applications. He advocates integrating processors, high-performance networks, and software into unified computing systems, with an emphasis on scalability, efficiency, and reliability. He views AI as a general-purpose technology that could become accessible across industries in a manner comparable to electricity. His longer-term interests include AI-assisted scientific discovery, computational biology, digital twins, climate modelling, and the potential for artificial intelligence to identify physical relationships and scientific principles not yet understood by researchers.
In 2026, Michael Kagan was named a recipient of the Technion Alumni Medal, an institutional distinction awarded by the Technion – Israel Institute of Technology to graduates for achievements in science, engineering, technology, medicine, architecture, industry, and public service. The Technion announced on 3 June 2026 that Kagan and Boaz Levy, Chairman of Israel Aerospace Industries, had been selected to receive the medal at a ceremony scheduled for 8 June 2026. Kagan was recognised for his contributions to high-performance computing and communication technologies, as well as his continuing support for the Technion and its relationship with the technology industry. The available institutional announcement establishes the award decision and planned presentation, although it does not independently confirm completion of the ceremony.
Kagan holds the professional membership grade of Senior Member of the Institute of Electrical and Electronics Engineers (IEEE). He joined IEEE in 1997 and has described the organisation as a source of technical information and professional exchange. IEEE Senior Member status is a professional membership distinction associated with experience and demonstrated professional performance in IEEE-designated fields. His membership grade was documented in an IEEE Spectrum profile published in 2022 and included in the publication's March 2023 print edition.
Since May 2017, Kagan has served on the Board of Governors of the Technion – Israel Institute of Technology. This is an institutional governance appointment rather than a competitive award. His participation reflects an ongoing formal association with the university where he completed his undergraduate education. The Technion also identified his contributions to strengthening links between academia and industry in its 2026 Alumni Medal announcement.
Kagan's professional career and technical contributions have been documented in industry publications, including IEEE Spectrum and EE Times Europe. IEEE Spectrum published a profile examining his work in high-performance computing, Mellanox networking, NVIDIA BlueField DPUs, and digital twins. In December 2023, EE Times Europe published an interview addressing his engineering career, processor architecture, and NVIDIA's accelerated computing strategy. In November 2025, Calcalist published an interview examining his role in NVIDIA's technology development and the integration of Mellanox. These publications constitute professional media coverage rather than formal awards.
Kagan has also participated in technical and industry events. He was listed as a speaker at the Flash Memory Summit in Santa Clara, California, in August 2019. In June 2022, the Technion's Architectures and Circuits Research Center announced a workshop featuring Kagan on entrepreneurship in semiconductors. He has additionally participated in industry conversations on the development of AI computing infrastructure, including a discussion at Sequoia Capital's Europe100 event.
