LEAP 2026 reveals the direction of Saudi Arabia’s next AI strategy

The opening days of LEAP 2026 have turned Riyadh’s Exhibition and Convention Center in Malham into a hub for major technology deals, with around $15 billion in investments, partnerships and computing infrastructure agreements announced at the event.

Held under the theme “Into New Worlds,” the fifth edition of the conference reflects Saudi Arabia’s growing ambition to move beyond adopting technology towards building, developing and exporting it. Much of the early dealmaking has focused on AI infrastructure, sovereign cloud and physical AI and logistics.

Here’s a look at some of the biggest announcements and cross-border partnerships making headlines at the event.

MAJOR HUMAIN COLLABORATIONS

Several of the major announcements at LEAP 2026 have involved HUMAIN, the Public Investment Fund-owned AI company, as it expands its partnerships across cloud infrastructure, AI computing and personal technology.

AWS

Amazon Web Services (AWS) made a major announcement at the conference, stating that its first cloud infrastructure region in Saudi Arabia is still expected to open in December 2026. The region, first mentioned in March 2024, involves a planned investment of more than $5.3 billion and will expand AWS’s global infrastructure presence to 40 regions.

AWS is also collaborating with HUMAIN to make up to 50 megawatts (MW) of capacity available in Saudi Arabia’s first AI Zone by 2028; the facility is expected to include AWS Trainium chips and NVIDIA AI infrastructure to support AI training and inference workloads.

AMD AND CISCO

AMD, Cisco and HUMAIN have also announced that AI infrastructure, based on AMD Instinct MI355X GPUs and Cisco networking technology, is now operational in Saudi Arabia and is being used by HUMAIN customers.

The deployment uses AMD GPUs and EPYC CPUs, along with Cisco’s Silicon One networking architecture, enabling HUMAIN to provide GPU-as-a-service for applications such as model training and inference.

The companies also intend to greatly expand the infrastructure. The joint venture is expected to deploy up to 250MW of AI infrastructure, using AMD’s next-generation Instinct MI400 Series GPUs, EPYC CPUs, and ROCm software, as well as Cisco networking technology.

Deployment is set to start in 2027, with the first capacity expected to come on line during the second half of that year. The larger project is still on course to have up to 1 gigawatt of AI infrastructure in place by 2030, the companies say.

MICROSOFT

Microsoft has expanded its collaboration with HUMAIN by combining HUMAIN ONE with Microsoft 365, which includes Copilot and Microsoft’s IQ capabilities, to create an AI productivity service for one million users in the Middle East and Africa.

HUMAIN ONE, which is hosted on Azure, will offer a single interface for enterprise applications, data and workflows, and may be extended to include HR, finance, procurement and IT.

The partnership is based on a plan to make HUMAIN’s ALLAM models available via Microsoft Foundry and to provide Arabic-language agents through Microsoft 365 Copilot, with support from HUMAIN AI specialists and Microsoft engineers.

The partnership is being expanded to include devices, with HUMAIN launching an AI PC at LEAP, designed to handle AI workloads both locally and via cloud services. The PC runs both Microsoft Windows and HUMAIN’s AI features and the first generation is due to become available to enterprise customers on September 20, 2026, with a goal of selling one million devices in the Middle East and Africa by 2030.

Microsoft also announced at LEAP 2026 that the new cloud data center region in Saudi Arabia East will be made available to customers in November 2026.

The facility, situated in the Eastern Region, will include three availability zones, enabling both government and private-sector organizations to store their data in Saudi Arabia while benefiting from lower latency when accessing Azure and AI services, along with improved security, resilience, and business continuity.

stc

The wider AI buildout is also increasing demand for the connectivity and data-center infrastructure needed to support it. At LEAP, stc group said its partnership with HUMAIN is focused on three areas: data-center infrastructure, connectivity and go-to-market. 

Through its data-center subsidiary center3, stc is developing AI-ready data-center capacity starting at 250MW, while expanding connectivity through subsea, terrestrial and satellite networks to link Saudi Arabia’s AI infrastructure with global markets. stc Group CEO Olayan Alwetaid said the Kingdom is undergoing “a rapid transformation in the field of artificial intelligence.” 

QUALCOMM

HUMAIN is also expanding into consumer and enterprise computing. At LEAP, Qualcomm Technologies and HUMAIN launched Horizon Ultra, an AI PC powered by Qualcomm’s Snapdragon X2 Elite platform. Developed over more than 12 months, the device is designed to run AI workloads locally while connecting to cloud infrastructure when additional computing capacity is required.

The first version will run on Microsoft Windows, while a subsequent version planned for 2027 will introduce HUMAIN OS. The device uses an 18-core Qualcomm Oryon CPU alongside Qualcomm’s Adreno GPU and Hexagon NPU, allowing different computing tasks to be distributed across dedicated processing components.

The approach aims to bring AI processing closer to users and devices, while allowing certain workloads and sensitive data to remain on-device.

NEW HP INNOVATIONS

HP is expanding its technology and innovation activities in Saudi Arabia, announcing a new startup incubator at LEAP 2026.

HP Inc. said HP Garage 2.0 will serve as an incubator for startups developing solutions aligned with Saudi Arabia’s Vision 2030 priorities. Inspired by the original HP Garage in Palo Alto, where Bill Hewlett and Dave Packard founded the company, the initiative will bring together startups, educators, industry partners and HP experts around areas including AI and the future of work.

The company said the program will focus on education, skills development and collaboration, with the aim of helping startups move ideas from early-stage concepts toward practical applications.

The announcement comes alongside updates on HP’s AI R&D Center of Excellence (CoE) in Dhahran, which was first announced at LEAP 2025. The center is now developing AI applications for Saudi organizations, with work spanning enterprise AI, Arabic-language development, applied research and local talent development.

One of its projects, HP Agent One, is an enterprise agentic AI solution that manages IT service requests and incidents, from understanding Arabic-language requests and routing tickets to resolving issues and escalating cases to employees. It can integrate with existing systems and run on-premises, allowing organizations to keep sensitive data within their infrastructure.

The Dhahran center is also developing FlowMaster, which combines optical character recognition with multimodal AI to analyze documents and automate administrative processes, including compliance checks and the identification of inconsistencies.

HP said the center is also working with academic institutions on training and knowledge transfer, as well as on research into Arabic large language models, including the Najdi dialect. Saudi datasets are being used for local applications, including a platform designed to predict student dropout risk.

THINK ANNOUNCES THINK GRID

At the conference, AI infrastructure company Think unveiled Think Grid, a dedicated AI compute service for enterprises, governments and AI-focused organizations, alongside its broader Think Fabric infrastructure platform.

Think Grid provides dedicated, high-performance AI compute through a monthly subscription, removing the upfront capital investment and operational burden typically associated with deploying dedicated AI infrastructure. Hosted initially in a Tier III facility in Riyadh, the service allows Saudi customers to run AI workloads while keeping their data within the Kingdom.

At launch, each Think Grid SuperNode combines four NVIDIA PRO 6000 Blackwell accelerators with 384GB of total VRAM, alongside Think’s proprietary Intelligent LLM Management (ILM) software, networking, storage, cooling and support. ILM is designed to improve GPU utilization by allowing multiple AI models to share available compute rather than leaving individual accelerators underused.

Think said Grid is already available to government and enterprise customers in Saudi Arabia and is being rolled out globally. Its Riyadh deployment gives local organizations access to dedicated AI infrastructure while Think manages the hardware, cooling, maintenance and underlying operations.