A National Strategy With Real Infrastructure Consequences
Artificial intelligence (AI) has become a national priority in Saudi Arabia. Essentially, AI can be described as the capability of computer systems to accomplish tasks that normally require the use of human reason. Such tasks include pattern recognition from large data sets, decision-making, language generation, and process optimization. Speaking of the infrastructure aspect, the significance of this term lies in the appearance of a completely different kind of compute workloads, which are much more energy-intensive compared to the typical ones.
The National strategy by the Kingdom of Saudi Arabia for Data and Artificial Intelligence, introduced in 2020 and being operationalized under the Saudi Data and Artificial Intelligence Authority (SDAIA), seeks to position Saudi Arabia among the top countries in terms of AI technologies in the year 2030. This target is an essential component of Vision 2030, which is the reform initiative of the kingdom started in the year 2016 with the intention of making the country less dependent on oil and diversifying its economy. Digital infrastructure is central to this transition process, and AI is central to digital infrastructure.
What seems to be overlooked in the wider debate around Vision 2030 is just how particular the demands on infrastructure will be. It is not just a case of expanding existing data center capacity in the traditional way. The workloads involved in national AI programs – including training large language models, running high throughput inferencing systems, and processing satellite imagery in bulk – demand dedicated infrastructure designed specifically to accommodate their demands. Upgrading general infrastructure to meet AI demands is costly, time-consuming and often ineffective.
What the National AI Strategy Actually Calls For
The NSDAI is built around several interconnected objectives: attracting AI investment into the kingdom, developing local AI talent, creating a regulatory environment for responsible AI deployment, and building the physical infrastructure that makes all of it possible. Each of those objectives depends on the others, but the physical infrastructure question is the one that sets the pace.
The strategy also has a strong sovereignty dimension. Saudi Arabia wants AI models trained on Saudi data, running on Saudi infrastructure, serving Saudi citizens and institutions. Sovereign AI infrastructure, meaning compute resources physically located within the kingdom and operating under Saudi jurisdiction rather than on foreign-owned cloud platforms, is a stated requirement, not a preference. That rules out treating offshore hyperscale cloud capacity as the primary compute layer. It demands physical assets in-country, controlled by entities that operate under Saudi law.
This is where the infrastructure question becomes concrete. Sovereign AI compute at a national scale requires GPU-dense facilities capable of handling the thermal and power loads that AI accelerators produce. It requires cooling systems designed for those loads from day one. And it needs a deployment model that can match the pace at which the strategy seeks to operate.
The Gap Between Ambition and Current Capacity
In the past few years, Saudi Arabia has made considerable investments in its data center ecosystem. There have been announcements from hyperscalers for new data centers, expansion by colocation firms, and allocations of funds through government-sponsored programs for digital infrastructure. This is the progress the Kingdom is expecting. The question is whether the existing and planned capacity is configured for AI workloads specifically.
Most traditional data center designs were based on rack densities of 5 to 15 kilowatts. A rack of AI servers powered by NVIDIA H100 or H200 GPUs, which are the AI accelerators commonly used for large-scale AI training today, demands 50 to 100 kilowatts per rack, possibly even higher. These cooling systems, power distribution units, and structural designs of a typical facility cannot provide such a density without undergoing a major retrofit. A retrofit process is costly and time-consuming.
There is an additional thermal challenge related to the climatic conditions of Saudi Arabia. Temperatures across Saudi Arabia reach beyond 45 degrees Celsius in the summer months. Air-based cooling approaches that function adequately in temperate climates become dramatically less efficient in these conditions. Facilities relying on conventional cooling architectures face higher energy consumption, lower reliability, and greater operational cost.
One way to measure the efficiency cost of poor thermal management is Power Usage Effectiveness, or PUE. PUE is the ratio of total energy consumed by a data center to the energy that actually reaches the compute hardware. A PUE of 1.0 is theoretically perfect: every watt in goes to compute. The global data center average sits at around 1.58. AI workloads, given their continuous high-density power draw, make a low PUE a financial necessity rather than a performance metric. Every tenth of a point above 1.2 translates directly into higher operating costs at exactly the scale where those costs compound.
There is also a timing issue. Traditional data center construction in the region typically takes 18 months to several years from design to commissioning. The AI strategy is not waiting. Research institutions, government agencies, and private sector operators are working to deploy AI systems on timelines measured in months. The infrastructure delivery model has to match that pace.
What AI-Ready Infrastructure Actually Looks Like
Meeting the infrastructure requirements of the national AI strategy means building for the specific demands of GPU compute, not adapting general-purpose capacity after the fact.
High power density is the starting point. Each compute pod needs to support 50 to 100 kilowatts of draw, with the mechanical and electrical systems to match. Power distribution must be engineered for that load from the design stage, with redundancy built in rather than added as an afterthought.
Liquid cooling, at such densities, goes from being a luxury solution to being a necessity. Liquid cooling to the chip involves direct-to-chip liquid cooling to the components of the GPU generating heat, as opposed to carrying away heat from the rack via air circulation. It is this technique that allows the use of high-density GPUs in Saudi Arabia’s climate. A well-designed liquid cooling system brings PUE figures to 1.2 or below. Rear-door heat exchangers and in-row cooling units are further options that can be configured based on workload profile and site conditions.
Deployability is important to the plan because of the time frame involved. The modular AI data center, or the factory-fabricated facility, can be made functional on deployment after its delivery as long there is a power source. This is due to the fact that everything from the thermal management system to power, fire protection, and monitoring systems have been tested before the arrival of the facility on site. There is no lengthy on-site construction phase, no sequential commissioning of separate systems, and no exposure to the construction delays that have affected large data center projects across the region.
Scalability without re-engineering is a further requirement. As models grow larger and inference volumes increase, compute requirements will expand. Infrastructure that can scale horizontally, adding pods as demand grows without forcing a rebuild of existing capacity, keeps options open in a way that a single large fixed facility does not.
Where PodTech Sits in This Picture
From PodTech’s perspective, the Saudi AI strategy represents a deployment challenge that the AI Factory in a Box was specifically built to address. The product is a prefabricated modular AI data center unit, factory-built and commissioned before delivery, delivering 50 to 100 kilowatts of compute density per pod with integrated liquid cooling and pre-commissioned power systems.
What matters most for the Saudi context is the combination of thermal engineering and speed to deployment. The AI Factory in a Box is built around direct-to-chip liquid cooling and enclosed hot/cold aisle containment, engineered for the thermal loads that GPU accelerators produce. In a climate where conventional cooling approaches lose efficiency progressively as outdoor temperatures climb, this is a meaningful operational advantage.
For enterprise teams, government agencies, research institutions, and colocation operators working to meet the demands of the national AI strategy, the infrastructure model that makes practical sense is one that delivers AI-grade compute density quickly, operates efficiently in the Saudi climate, and can be expanded as requirements grow.
The Infrastructure Work Has to Match the Strategy
The national AI vision of Saudi Arabia is sincere and adequately funded, but time-bound. The infrastructure that will be necessary to fulfill the goals set out in that vision is unique in its requirements: high power density, liquid cooling, fast deployment capability, and scalability without reinvention.
The kingdom has the capital and the political will to make the strategy real. Whether or not it meets its deadlines depends on the ability of the physical infrastructure to match the pace. Knowing what needs to be done to provide the compute infrastructure capable of handling AI workloads is the first step towards deploying it.
Frequently Asked Questions
The Infrastructure Work Has to Match the Strategy
PodTech Data Center is a UAE-based modular data center manufacturer serving enterprise, government, and telecom operators across the GCC and Africa. Our AI Factory in a Box delivers high-density GPU compute infrastructure with integrated liquid cooling, rapid deployment capability, and a PUE as low as 1.2. For deployments in Saudi Arabia and across the region, contact our team at info@podtechdatacenter.com or visit podtechdatacenter.com.