A senior data center infrastructure executive from Amazon Web Services joins xAI to lead machine learning infrastructure operations.

Saurabh Kumar, principal engineer for Amazon Web Services' data center networking strategy, has left the $132 billion cloud giant after 9 years to lead machine learning infrastructure for Elon Musk's startup xAI. "After 9 years leading datacenter network strategy at AWS, I have decided to take on a new challenge," Kumar announced in a LinkedIn post that garnered over 120 comments. "It was a privilege to work alongside the exceptional talent at AWS and shepherd its transformation from a consumer of network technology to an industry-leader, driving bleeding-edge products across multiple networking domains," he said.

Kumar's career spans more than a decade across leading technology companies. He began at optical networking solutions and semiconductor company Infinera, where he held top engineering and development roles until the company's 2025 acquisition by Nokia for $2.3 billion. He then joined Google for approximately one year in a network strategy and architecture role for Google Fiber before moving to AWS in 2017 as principal engineer for the Data Center Network Engineering unit.

At xAI, based in Palo Alto, California, Kumar will focus on "building teams, processes and interconnect products from GPUs to Backbone" for the company's ML infrastructure. xAI, owner of the popular Grok AI chatbot rival to OpenAI's ChatGPT, has been aggressively expanding its data center footprint. The company currently operates two data centers in the U.S. and recently announced the purchase of a third building called Macrohard to expand infrastructure. "I look forward to delivering the network infrastructure for massive-scale compute gigafactories powering the innovative products at xAI," Kumar stated. CEO Elon Musk confirmed the expansion on X, writing "xAI has bought a third building called Macrohard," and later noted the company is developing a new AI software company under that same name. Musk previously unveiled that xAI is aiming to boost training capacity to nearly 2 gigawatts of compute power. Last week, xAI introduced Grok Business and Grok Enterprise for organizations, with Grok Business pricing set at $30 per seat per month and Grok Enterprise featuring custom pricing.

Kumar's move reflects broader industry shifts in infrastructure demands. In a LinkedIn post last week, he wrote, "After a decade of speeds-and-feeds, we find ourselves at a critical juncture for datacenter networks—emerging modalities like CPC, NPO, CPO, THz radio; technologies like ring resonators and TFLN, OCS-based reconfigurability, and the mandate to identify and champion the most power-efficient approaches to stay in lock-step with GPU cadence while accelerating buildouts and improving build quality." Meanwhile, AWS—which generated $33 billion in revenue during the third quarter of 2025—continues its own massive infrastructure investments, building dozens of new data centers globally to expand customer reach.