AMD and Cisco expand Huwmen's architecture in Saudi Arabia, betting on open processors for sovereign computing
The high-performance computing infrastructure in Saudi Arabia entered a new operational phase with the announcement that AMD, Cisco and Huwmen would begin offering Instinct MI355X AI-targeted processor services to customers inside and outside the Kingdom. The current production organization combines advanced graphics processors with central EPYC processors, Cisco’s Silicon One networking solutions and 800-gigabit optical technologies, enabling Huwmen to provide graphics-processing-as-a-service for training and inference of advanced models.
The plan does not stop at the current capacity, as the three companies are preparing to deploy additional infrastructure up to 250 megawatts by 2027. The next phase will rely on the latest generation of Instinct MI400 processors and open-source ROCm software together with Cisco’s infrastructure, with actual service capacity expected to come online in the second half of 2027, paving the way toward a publicly stated goal of delivering a full gigawatt of AI-dedicated computing power by 2030.
This expansion represents a direct bet on opening market options for model developers and governments that seek local infrastructure independent of closed-system monopolies.
The platform targets government agencies, enterprises, research centres and model developers that require workloads to run locally. AMD markets its architecture on support for open models and free software, granting institutions full sovereignty over data localisation, model customisation and governance, which bolsters the company’s presence in the sovereign AI sector that nations are racing to build to ensure computational independence.
The project arrives as AMD’s data-center business is experiencing record growth, with revenues in the segment up 107 percent year-over-year to $6.7 billion in the second quarter of 2026. The company expects server revenues to grow by more than 70 percent in 2027 and to double total data-center revenue, driven by the spread of Instinct processors whose sales have doubled as they gain traction in national labs, cloud providers and startups.
However, AMD’s path to competing with the giants faces structural challenges, as Nvidia continues to dominate the landscape thanks to its entrenched CUDA software stack and massive hardware base that generated $89 billion in the data-center segment alone in the second quarter of its fiscal 2027. At the same time, Broadcom is applying increasing pressure through dedicated chips and networking technologies, with forecasts that its AI semiconductor revenues will exceed $100 billion in fiscal 2027 and its commitment to provide massive computing capacity to companies such as Meta, Google, Anthropic and OpenAI.
For technical engineering teams and organisations in the Gulf, Egypt and the Arab region, the arrival of MI355X and MI400 processors in regional data centres practically changes the pricing equation and development options.
This shift gives financial institutions and government bodies in the region a direct local alternative that ensures sensitive data remain within geographic borders without relying exclusively on Nvidia platform pricing or suffering from supply-chain constraints. It also compels Arab programming teams to invest in building advanced expertise around ROCm environments and adapting open frameworks, rather than concentrating skills in a single ecosystem, opening the way to lower the cost of training language models and inference at scale within regional infrastructure.