Middle East AI Infrastructure Spending Reaches $8.4B as Adoption Scales
AI infrastructure investment across the Middle East is accelerating as businesses move from early experimentation towards large-scale, long-term deployment.
Global spending on AI infrastructure reached $318 billion, with the GCC accounting for $8.4 billion, according to IDC. The global figure was almost double the $153 billion recorded in 2024, highlighting the rapid expansion of infrastructure supporting artificial intelligence.
Spending is expected to continue rising in 2026 as organisations increase investment in computing capacity, cloud infrastructure and systems required to support AI applications.
The shift also reflects a broader move from AI experimentation towards deployment at scale, bringing new security and governance challenges for enterprises.
Abu Dhabi-based Ansen Technology, an AI-native cybersecurity company, is focused on addressing some of these risks as organisations across the region expand their use of AI.
Arabian Business spoke to Zheng Li, Co-founder, GVP and Chief Product Officer at Ansen Technology, about the changing AI landscape and the security implications of wider enterprise deployment.
Li said one of the most significant changes is the transition from AI systems generating information to systems capable of taking action.
“Traditional cybersecurity is largely concerned with whether a user, file, device or process is malicious. With agentic AI, organisations must also determine whether an AI system is authorised to perform a particular action, in a particular context, at a particular time,” Li said.
He added that an AI agent can have legitimate access while still making an inappropriate decision because of incomplete information, incorrect interpretation or excessive permissions.
This means security frameworks need to evolve beyond protecting AI systems themselves to include governance over the actions AI systems are permitted to take.
Li also said the potential for AI to make an incorrect decision could become a significant enterprise risk as adoption expands.
“The risk is not necessarily that an attacker compromises an AI system. It may be that the system operates as designed but makes the wrong decision,” he said.
As AI becomes embedded in critical business workflows, organisations will need to assess not only whether systems can be compromised, but also whether their decisions remain within defined and authorised boundaries.
Li also highlighted identity and access management as an increasingly important issue as AI agents begin initiating actions on behalf of users and organisations.
Traditional access controls focus on questions such as who a user is and what they can access. Agentic AI environments require additional considerations, including what an agent is attempting to do, why it is taking an action and whether that action is permitted in a specific context.
He said organisations should adopt more granular and context-aware permissions, apply least-privilege principles to AI agents, continuously assess their actions and maintain detailed audit trails.