AMD warns enterprises to overhaul AI build and scaling strategies

AMD’s senior manager for the METAP region, Zaid Ghattas, says enterprises must rethink AI construction and scaling, citing an IDC whitepaper commissioned by AMD link. The study finds 83% of firms run fewer than ten AI use cases simultaneously, and only 21.7% conduct full ROI analysis, indicating early-stage adoption.

More than three-quarters of organisations lack clarity on agentic AI use cases, prompting calls for standards, guardrails and stakeholder alignment. Security measures such as access controls, anonymisation and adversarial testing are increasingly reported, while about two-thirds begin GenAI projects with comprehensive cost assessments covering infrastructure, licensing, labour and scalability.

The report highlights a shift toward high-density server CPUs to support continuous inference streams, arguing that future AI workloads will rely on tightly coupled CPU-GPU systems rather than siloed GPUs. Seventy-seven percent of enterprises prioritize aligned hardware and software to simplify operations and improve workload portability.

Our read: AMD’s data underscores a gap between AI ambition and operational maturity, especially in the UAE where data sovereignty and sustainability amplify the need for open, standardized infrastructure and rigorous governance.