AMD’s Zaid Ghattas explores why scaling enterprise AI in the UAE demands integrated infrastructure, measurable business outcomes and governance from the outset
E ven in the most advanced AI markets, one thing is becoming clear: adoption is relatively easy; scaling is more difficult. This is evident across the UAE, where national AI ambitions are accelerating rapidly, from the UAE AI Strategy 2031 to landmark investments in sovereign AI infrastructure. As organisations move toward agentic systems, the real shift is no longer about experimentation, but about making AI a core part of how the business operates.
The companies that succeed will be those that embed AI into strategy, infrastructure and decision-making, turning it from a tool into a fundamental driver of growth.
AMD estimates underscore this shift. In 2026, more than 60% of global AI compute capacity will be used for inference rather than training. Yet, despite this momentum, enterprise infrastructure is struggling to keep pace.
The Scale Challenge: What IDC Says Is Shaping Enterprise AI
A recent IDC whitepaper commissioned by AMD, showed adoption is broad, but scale is still developing. A striking 83% of enterprises currently run fewer than ten AI use cases simultaneously, indicating many are early in their transformation. Just over two in ten (21.7%) conduct full ROI analysis, and 22.2% verify that their AI investment alignment with strategic objectives. For organisations operating in the UAE, where government and regulators are increasingly focused on measurable AI outcomes, strengthening these basics is not optional — it is essential.
Compounding the gap is a widespread lack of architectural clarity. More than 75% of organisations lack clarity on agentic AI use cases, creating a timely opportunity to set standards, design guardrails and align stakeholders. Security and governance are areas of focus – firms report use of access controls, anonymisation, and lifecycle policies, and many employ adversarial testing and prompt manipulation drills to harden models.
Cost discipline is maturing as well: about two‑thirds of organisations begin GenAI initiatives with comprehensive cost assessments that span infrastructure, licensing, labour, and scalability. Together, these shifts point to an ecosystem that is learning quickly and building the right muscles for sustainable scale.
Compute Redefined: Why the CPU and System Architecture Matter
As AI workload demands escalate, enterprise IT leaders are coming face-to-face with three critical limitations: power and cooling density limits, escalating token costs and complex governance.
Scaling agentic AI requires recognising that AI infrastructure is fundamentally a data center-scale system problem and that there is no one-size-fits-all solutions for AI compute. While GPUs remain essential for heavy mathematical reasoning, they are only part of the story. Agentic workflows rely heavily on agent sandboxes – environments executing logic, querying vector databases, running orchestration frameworks and interfacing with enterprise applications. This shift has placed high-density server CPUs back at the core of the data center, demanding servers optimised for maximum core density, low latency and high memory bandwidth to feed continuous inference streams. The future of enterprise AI is going to be driven by high-performance CPUs and GPUs working together on a combined system, rather than treating them as separate silos.
At the same time, hardware standardisation and efficiency is becoming top of mind. The IDC research confirms that a majority (77%) of enterprises prioritise aligned hardware and software to simplify operations, reduce friction, and enable workload portability and performance predictability.
The UAE Context: Governance, Sovereignty and Sustainability
The UAE’s AI ambitions come with their own set of infrastructure considerations. Data sovereignty is increasingly central to the national agenda, with a growing emphasis on keeping sensitive data within borders and building local AI capabilities that are not dependent on a single vendor or architecture. Rather than acting as a constraint, these priorities are driving organisations to adopt stronger governance frameworks and more deliberate infrastructure strategies from the outset.
To meet both sovereign requirements and commercial imperatives, enterprise leaders in the region are turning to open platforms. An open ecosystem spanning architectures, networking, and software helps prevent vendor lock-in and gives organisations the flexibility to tailor their infrastructure to specific regional and compliance requirements. This approach also aligns to the UAE’s Net Zero 2050 goals, open, efficient, and standardised systems consume less energy and generate less waste than fragmented, proprietary alternatives.
What Enterprises Can Do Now: Building AI That Performs, Scales, and Complies
AI doesn’t scale on experimentation alone — it scales on focus. Organisations that move beyond pilots do three things differently:
- They make AI accountable to the business by tying every initiative to measurable outcomes and disciplined ROI.
- They treat infrastructure as strategy by building consistent, high‑performance environments across cloud, data center, and edge to support emerging agentic systems
- They embed governance from day one, from data controls to model testing, turning it into an enabler of trust, speed, and reliable, production‑ready AI.
This opinion piece is authored by Zaid Ghattas, General Manager at AMD for the METAP region.
Source: Tahawul Tech