AMD and Cisco launch Humane computing cluster in Saudi Arabia: Instinct processors enter operational service toward a gigawatt target

AMD Instinct MI 355 X graphics processors and Cisco networking equipment have entered direct operational service in the Kingdom of Saudi Arabia, beginning to process actual AI workloads for Humane’s customers inside and outside the Kingdom. This compute cluster, which combines Instinct cards, EPYC processors and networks built on Cisco Silicon One chips, represents a tangible shift from the stage of announcements and investment promises to the stage of generating compute capacity and selling it to organisations in the region.

Tariq Amin, chief executive officer of Humane, said that operating the cluster demonstrates the company’s ability to build advanced AI infrastructure, run it and deliver it at scale. For its part, Lisa Su, chair and chief executive of AMD, said that bringing these systems to work in the Kingdom constitutes a key milestone in the partnership to build an open, high-performance compute platform that serves customers today and will expand measurably in the coming years, a point also confirmed by Chuck Robbins, chief executive of Cisco, by linking network capabilities to the planned expansion.

The three companies plan to launch a second phase with capacity up to 250 megawatts by 2027, paving the way to reach a full-gigawatt target by 2030.This expansion relies on AMD’s open-source ROCm software suite, and the compute capacity is offered to customers as a cloud service that rents processing units to cover model training and inference tasks, supported by Cisco’s Nexus 9000 network platform equipped with 800-gigabit optics, moving Humane into a regional compute vendor and provider role rather than merely importing hardware.

To avoid servers remaining without staff able to harness them, AMD, in collaboration with the Saudi Ministry of Communications and Information Technology and the Digital Cooperation Organization, launched an open developer programme during the LEAP 2026 event. The programme aims to train local engineers on the ROCm suite, building a domestic software community that can run these advanced capacities and develop applications on top of them without relying entirely on bringing in external experts.

The announcement coincided with Amazon Web Services, via its chief executive Matt Garman, confirming that its first cloud region in Saudi Arabia is on track for launch before the end of 2026. It was accompanied by the recording of technology investments exceeding $15 billion on the first day of the LEAP 2026 conference, with an additional $2.5 billion on the second day and $857 million on the third day, all focused on expanding data centres and localising infrastructure.

Owning compute locally frees financial institutions and government sectors in the Gulf from data-transfer restrictions and regulatory compliance requirements for cross-border cloud services.This shift means that technology managers and development teams in the Gulf and the region can reserve local processing and training resources with lower latency and more flexible cost, as well as break reliance on global service providers outside the borders. The practical bet in the next phase remains tied to Humane’s ability to price inference and training services efficiently enough to compete with parallel regional bets, foremost among them the UAE’s G42, in a genuine compute race that redraws the map of technical sovereignty in the Middle East.