The news
On August 31, 2026, Advanced Micro Devices (AMD), Cisco Systems (CSCO) and HUMAIN said Saudi AI infrastructure built on AMD Instinct MI355X graphics processors was live and serving customers. The partners also said they plan to deploy up to 250 megawatts more from 2027.
HUMAIN, which the release describes as a company of the Public Investment Fund (PIF), Saudi Arabia's sovereign wealth fund, is now delivering production AI compute on MI355X GPUs, AMD EPYC processors and Cisco Silicon One-based networking, according to the announcement. Cisco said its N9000 Series platform, built on Silicon One and 800G optics, links the GPUs in a fabric designed for scale and low latency. The systems serve HUMAIN customers in the Kingdom and beyond, the companies said, but the release did not give a megawatt figure or GPU count for the capacity now running.
The next phase calls for up to 250 megawatts of AI infrastructure built on AMD Instinct MI400 Series GPUs, EPYC processors and AMD's ROCm open software, along with Cisco networking and critical infrastructure, through the companies' joint venture. Deployment is planned to begin in 2027, with capacity expected to start coming online in the second half of that year. Citing strong customer demand, the partners said the joint venture remains on track to deploy up to 1 gigawatt of AI infrastructure by 2030. No dollar figures were disclosed.
The venture was announced on November 19, 2025, with AMD, Cisco and HUMAIN as founding investors and AMD and Cisco as exclusive technology partners. That plan described a first phase of 100 megawatts using AMD's MI450 Series GPUs, with operations beginning in 2026. The August announcement instead described MI355X systems now live and MI400 Series chips for the next phase; the companies did not explain the change.
AMD chair and CEO Lisa Su called the launch "an important milestone in our work with HUMAIN." HUMAIN CEO Tareq Amin said demand for sovereign, high-performance AI infrastructure keeps growing. Sovereign AI refers to computing capacity built and controlled within a country. Cisco chair and CEO Chuck Robbins said the region is shifting from AI investment to delivery. Separately, Middle East AI News reported that Amin said 13,000 AMD GPUs would be deployed in 2026.
The numbers
- Next-phase capacity, MI400 Series (deployment from 2027)
- Up to 250 MW
- Joint venture target by 2030
- Up to 1 GW
- First-phase plan announced November 19, 2025 (MI450 Series)
- 100 MW
- AMD-HUMAIN collaboration announced May 13, 2025
- Up to $10 billion for 500 MW over five years
- AMD GPUs to be deployed in 2026, per HUMAIN CEO (Middle East AI News)
- 13,000
Why CEOs should care
For technology buyers, especially companies that need AI computing inside Saudi Arabia or the wider Gulf, HUMAIN now offers AMD-based capacity that is running rather than planned. Before signing, ask which models and frameworks run well on AMD's ROCm software, how much porting work code built for Nvidia's CUDA platform would need, where data is stored and who can access it, and what availability commitments come with the service. Note the hardware timeline too: MI355X systems now, MI400 Series capacity from the second half of 2027.
For CFOs and boards, the announcement shows sovereign AI programs in the Gulf turning into operating capacity for AMD and Cisco, though no revenue or contract values were disclosed. Companies weighing a state-backed provider such as a PIF company should assess regulatory and geopolitical exposure alongside price, and ask how capacity would be allocated if demand outruns the build-out. CISOs should ask how tenants are isolated on the shared Cisco network fabric, who operates the facilities day to day and what audit rights customers get.
The bigger picture
The launch is the latest step in a partnership that began in May 2025. On May 13, 2025, AMD and HUMAIN announced a collaboration under which the parties would invest up to $10 billion to deploy 500 megawatts of AI compute over five years across sites from Saudi Arabia to the United States. In November 2025, Cisco joined AMD and HUMAIN in the joint venture targeting 1 gigawatt by 2030.
For AMD, a running cluster abroad gives it a working reference as it pursues sovereign AI projects in other countries. Cisco gains a large, public deployment of its Silicon One networking inside a GPU cluster, where the network that links accelerators is a major part of system performance and cost.
What happened next
The announcement coincided with the LEAP 2026 technology conference in Riyadh, which ran from August 31 to September 3, 2026, according to tbreak. In a recap published on September 9, AMD said Lisa Su appeared on the LEAP main stage with Amin, and that AMD had partnered with Saudi Arabia's Ministry of Communications and Information Technology and the Digital Cooperation Organization to launch an open developer ecosystem. AMD also described a HUMAIN AI application platform aimed at helping Saudi enterprises test and deploy AI tools.
Watch whether the partners disclose how many megawatts are live, whether MI400 Series capacity starts arriving in the second half of 2027 as planned, and whether HUMAIN names anchor customers.
