Humain officially activated its new processing infrastructure focused on artificial intelligence at Arábia Saudita. The venture, which has links with the country’s Fundo of Investimento Público, marks a strategic step in the region’s technological sovereignty. Nesta inaugural stage, the company completed the installation and put into operation a batch consisting of 1,024 accelerators from the Cloud AI 100 line.
The move is part of an ambitious plan that aims to reach a total capacity of 200 megawatts. The goal established by the company foresees that this infrastructure expansion will be completed by the end of the schedule stipulated for 2026, consolidating the country as a relevant hub in the sector.
The choice of Qualcomm processors, originally launched in 2019, reflects a pragmatic approach to the global shortage of cutting-edge semiconductors. The immediate availability of the hardware allowed Humain to accelerate the start of commercial activities without relying on the long waiting lines that affect other manufacturers.
First customer and commercial applications
Adobe was confirmed as the first major customer to use the resources of the new data center. The creative software giant will leverage the infrastructure to run hybrid AI workloads, operating between the edge and the cloud. Essa partnership validates the technical capability of deployed chips to handle complex demands, including generative AI and large language models.
The collaboration demonstrates that, despite not being the newest components on the market, the AI100 Ultra accelerators maintain operational relevance. With 64 AI cores per board and up to 576 MB of on-die SRAM memory, the device offers an attractive cost-benefit ratio for large-scale inference, prioritizing energy efficiency over pure processing brute force.
Supplier diversification and future
To ensure resilience and not depend on a single technology source, Humain has structured agreements with multiple chip manufacturers. Além from Qualcomm, the expansion strategy includes the integration of technologies from NVIDIA and AMD. The planning includes the acquisition of 18 thousand units of the NVIDIA GB300 Grace Blackwell system, strengthening the capacity to train more robust models.
Diversification aims to mitigate supply risks and adapt the data center to different types of workloads. Enquanto Qualcomm chips focus on efficient inference, AMD equipment should make up a 500 MW base focused on high-performance computing, creating a hybrid and flexible ecosystem.
The company’s technology roadmap also foresees hardware upgrades in the coming years. The expectation is to integrate new generations of processors, such as the AI200 and later the AI250, promising significant leaps in memory and efficiency to compete directly with segment leaders.
Operational infrastructure details
The data center operation was designed to offer low latency and high availability, critical factors for real-time applications. Thermal management and energy consumption are constantly monitored to ensure the sustainability of the project in a region with high climatic demands.
The main pillars of the current project include:
- Initial installation of 1,024 Qualcomm Cloud AI 100 accelerators.
- Focus on energy-efficient inference operations.
- Expansion target to 200 MW of total capacity.
- Strategic partnerships signed with Qualcomm, NVIDIA and AMD.
- Meeting demands for generative AI and hybrid processing.
This configuration allows Humain to offer competitive services in the global market, attracting companies looking for viable and immediate alternatives for their artificial intelligence projects.