"A lot of people don't realize that Nvidia is the largest network provider in the world today," Nvidia Senior Director of HPC and AI Hyperscale Infrastructure Solutions Dion Harris told "Globes" at the Computex computing conference in Taiwan last week.
He explained, "This stemmed from processes that Nvidia identified, primarily the recognition that to expand and develop AI and maintain its pace of progress, systems that can grow quickly are needed - that's why we acquired Mellanox."
Nvidia's communications division, which is managed from Israel and is based on Mellanox's operations, last month reached an annual revenue rate of $40 billion, with an annual growth rate exceeding 200%. "To advance in the field of AI, the network is becoming critical," observes Harris. "Due to this, we built the complete solution that we offer to companies today."
The days of the "agent revolution"
Harris is Nvidia's man working with the cloud giants and AI companies. His job is to understand the needs of each company, help them design the architecture of their data centers and identify future developments that will keep Nvidia relevant with clients like Meta and OpenAI, and in competition with AMD and Intel.
Last week, Nvidia launched "Vera," a core processor designed for home computers and AI servers, directly challenging Intel, which launched its advanced Xeon 6 chip at the same time. Following the announcement, many in the industry wondered where the strategic partnership between Intel and Nvidia, which was announced only last September, had ended. Harris dismissed the claims: "Intel is still part of our strategy, and we continue to support the x86 architecture. A customer who wants to combine our GPU with an Intel processor will be able to do so, just as we enable it with ARM technology."
Harris explains the rationale behind the development: "The new workloads in data centers, resulting from the ‘agent’ revolution, require a new type of processor. The reason we created the new processor is to address a new market. If it weren’t for this need, we would continue to rely on existing Intel or ARM technologies."
The data centers of the cloud giants need increasing processing power. They are no longer just about training models but also running a variety of applications simultaneously. This phenomenon, known as ‘heterogeneous data centers,’ requires the use of chips from different vendors. The entry of players like Intel, AMD, and SambaNova into the picture raises the question of whether this is a share that has been taken away from Nvidia.
Harris says, "In the era of AI agents, systems produce content from scratch. In this reality, the order of priorities changes and efficiency in token production becomes critical. Such efficiency cannot be achieved using a single type of processor or graphics card. It requires complex communication between processors and software and hardware layers, deployed at different sites and sometimes coming from external suppliers. To meet these requirements, a combination of graphics accelerators is required alongside rapid expansion capacity and a network infrastructure that will connect all data centers. The aim is to ensure efficient data processing from the individual server rack to connectivity between remote servers."
Published by Globes, Israel business news - en.globes.co.il - on June 11, 2026.
© Copyright of Globes Publisher Itonut (1983) Ltd., 2026.