EPC & Construction

Before the GPUs, There Are Buildings

Every GPU cluster in Saudi Arabia’s $77 billion AI buildout sits inside a building that someone had to design, procure materials for, and physically construct. Before Humain can commission NVIDIA Blackwell racks, before stc Cloud can offer inference capacity to Saudi enterprises, before any of the announced hyperscale compute enters service — there is concrete, steel, cooling infrastructure, and power distribution equipment that must be manufactured, shipped, installed, and tested. The EPC (Engineering, Procurement, Construction) sector is the layer of the Saudi AI buildout that most technology coverage ignores and most financial models undercount.

This is not a minor logistical detail. At the scale Saudi Arabia has announced — Humain’s commitment alone implies somewhere between 50 and 100 large-scale data center facilities depending on average facility size — the construction requirement represents one of the largest coordinated infrastructure programs in the Kingdom’s modern history. The EPC sector’s capacity, speed, and cost structure will determine whether Saudi Arabia’s AI compute ambitions translate from PowerPoint to operational megawatts on the timeline that officials and investors are projecting.

What EPC Means in the Data Center Context

EPC contracting means a single firm or consortium takes responsibility for three interlocked phases of a capital project. Engineering covers design, specifications, and technical drawings — the structural engineering of a building that must support extremely dense server racks, the mechanical engineering of cooling systems capable of handling 40+ kilowatts per rack, the electrical engineering of power distribution from utility substation to server floor. Procurement means purchasing the long-lead equipment: generators, uninterruptible power supply systems, cooling units, switchgear, cabling — items with 12- to 18-month lead times that must be ordered before a foundation is poured. Construction is physical execution: site preparation, civil works, structural steel, building envelope, fit-out, commissioning.

Data center EPC is not the same as building an office tower or a hospital. The mechanical, electrical, and plumbing (MEP) density in a hyperscale data center is extreme — MEP can represent 65–80% of total project cost, compared to 20–30% for a typical commercial building. The power density trends driving AI workloads are pushing this further. A facility designed for 20 kilowatts per rack in 2020 is obsolete for AI training workloads in 2025, where 80–100 kilowatts per rack is the emerging standard and liquid cooling is shifting from optional to mandatory.

The Scale of Construction Saudi Arabia Has Committed To

Humain’s announced $77 billion AI compute commitment, phased over five years, implies a capital construction program of roughly $15–25 billion, depending on what proportion of spend goes to hardware versus facilities. Industry benchmarks suggest that for AI-optimized hyperscale data centers, facility construction (including power infrastructure) represents 20–35% of total lifecycle cost before hardware. Applied to $77 billion, that implies $15–25 billion in physical construction across the program.

For context: Saudi Arabia’s entire data center market was estimated at under $2 billion annually before 2024. The announced buildout implies a 10x or greater step-up in annual data center construction activity, sustained over five years. This requires not only contractor capacity, but skilled labor, specialized equipment, and supply chains that do not currently exist at the required scale inside the Kingdom.

The NEOM project — itself one of the world’s largest infrastructure programs — has absorbed substantial EPC contractor bandwidth in Saudi Arabia since 2017. The AI compute buildout now competes for the same pool of experienced project managers, electrical subcontractors, and construction equipment. This competition for resources is a real constraint that announced timelines rarely acknowledge.

Bechtel: The Global Benchmark Contractor

Bechtel is the world’s largest privately held engineering and construction firm, and it has been active in Saudi Arabia for over 75 years. Its Saudi portfolio spans oil and gas megaprojects (Saudi Aramco’s Ras Tanura refinery expansion, the Jazan refinery complex), infrastructure (Riyadh Metro), and NEOM-adjacent assignments. This deep institutional presence — local offices, established supplier relationships, a trained Saudi workforce — gives Bechtel a structural advantage when bidding on landmark technology infrastructure projects.

Bechtel’s relevance to the AI compute buildout is both direct and indirect. Direct: Bechtel has moved aggressively into data center construction globally, establishing a dedicated data center practice that has delivered hyperscale facilities for major cloud providers in the US, Europe, and Asia. It brings the same discipline — modular design, standardized procurement, rigorous commissioning protocols — that hyperscale operators demand. Indirect: Bechtel’s presence on NEOM and other mega-projects has seeded the Saudi construction sector with project management practices and technical talent that will be drawn into data center work as those projects wind down or plateau.

The challenge for any global EPC firm operating in Saudi Arabia is Saudization — the Kingdom’s requirement that domestic employees constitute a mandated percentage of the workforce across industry sectors. Bechtel’s Saudi operations have decades of experience navigating this, and the company actively runs training and development programs for Saudi engineers. For data center construction, where the technical skill profile is specialized, meeting Saudization targets while maintaining construction velocity is a real management challenge.

Saudi Binladin Group: The Domestic Giant

Saudi Binladin Group (SBG) is the Kingdom’s largest domestic contractor by revenue and, arguably, by legacy. Founded in 1931, SBG has built some of Saudi Arabia’s most recognizable infrastructure: the expansion of the Grand Mosque in Mecca and Medina (an ongoing decades-long assignment), King Abdulaziz International Airport, major road and rail networks, and commercial developments across the Kingdom.

SBG’s relevance to the AI buildout is significant precisely because it is domestic. It has existing relationships with Saudi government entities, established supply chains for construction materials sourced locally or through known import channels, and a workforce already operating under Saudi labor regulations. For government-directed projects — and most of Humain’s initial buildout will involve government-aligned entities — domestic contractors with established government relationships carry implicit advantages in the procurement process.

SBG went through a severe restructuring between 2016 and 2020, following the financial pressure of the NEOM contract disputes and the broader slowdown in Saudi government capital expenditure during the 2014–2016 oil price decline. That restructuring reduced its workforce from a reported 200,000+ employees at peak to a leaner operation. The recovery under Vision 2030 spending has restored much of its capacity, but SBG’s current operational throughput is a question that any serious analysis of Saudi AI construction timelines must address.

Alfanar: Construction Meets Energy

Alfanar occupies an unusual dual position in the Saudi AI buildout: it is both a major EPC contractor and an energy company. Founded in 1976 as an electrical equipment manufacturer, Alfanar has expanded into construction services and renewable energy development. For data center projects specifically — which require not just building construction but substantial power infrastructure — Alfanar’s ability to provide both is strategically valuable.

Alfanar Construction has delivered industrial and commercial projects across the Gulf, with particular depth in electrical and mechanical systems. Its energy division has developed utility-scale solar and wind projects. As Saudi data centers increasingly seek to align with Vision 2030’s renewable energy targets — whether for genuine sustainability reasons or to satisfy international customers’ carbon reporting requirements — contractors that can deliver both the facility and its power supply under a single EPC structure will carry a cost and coordination advantage.

Turner Construction and Jacobs Engineering: International Specialists

Turner Construction, a subsidiary of CBRE Group, is one of the leading data center general contractors in the United States, with a data center portfolio that includes hyperscale campuses for AWS, Google, and Microsoft. Turner has a growing international presence and the data center construction expertise that few domestic Saudi contractors can match on specialized technical execution. Its role in Saudi AI infrastructure is likely to be as a technical partner or joint venture lead on the most complex facilities, where international hyperscale design standards must be met to satisfy operator requirements.

Jacobs Engineering brings a different profile: it is a program management and design engineering firm with less emphasis on direct construction execution. Jacobs’ value in the Saudi AI context is in front-end engineering and design (FEED), master planning for multi-building data center campuses, and the kind of complex systems integration that comes with connecting a large facility to the Saudi grid. Jacobs has longstanding Saudi government relationships through its infrastructure advisory work.

Building in Extreme Heat: The Cooling Infrastructure Challenge

Saudi Arabia’s climate is one of the most hostile environments on Earth for data center construction and operation. Summer ambient temperatures in Riyadh routinely exceed 45°C (113°F), and in coastal locations like Jeddah, high humidity compounds the thermal load. Data centers generate enormous amounts of heat — at 100 MW of IT load, a facility is dissipating 100+ MW of heat that must be removed from the building continuously.

In temperate climates, data centers can use free cooling — leveraging cool outdoor air to reduce mechanical cooling loads for significant portions of the year. This is simply not available in Saudi Arabia. Facilities in the Kingdom must run full mechanical cooling year-round, which means higher capital costs for chillers and cooling towers, higher operating costs for electricity, and greater complexity in the mechanical plant.

Liquid cooling is becoming the answer for the highest-density AI workloads. Direct liquid cooling — running coolant directly to the chip package — bypasses air cooling entirely for the compute load, dramatically reducing the heat that must be removed by facility systems. Rear-door heat exchangers and immersion cooling tanks are increasingly being specified for AI-optimized facilities. These technologies add capital cost and construction complexity, require specialized installation skills, and introduce new maintenance requirements. For Saudi data centers targeting the performance levels that Humain’s AI ambitions require, liquid cooling is not a future option — it is a present necessity.

The Construction Timeline Math

Industry data on hyperscale data center construction timelines points to 24–36 months from groundbreak to first commercial operations for a large campus, and 36–48 months to full campus buildout at scale. This baseline assumes experienced contractors, established supply chains, and sites with existing power and water infrastructure. In Saudi Arabia, several of these assumptions are challenged:

Power infrastructure: Saudi Arabia’s electricity grid has expanded rapidly under Vision 2030, but connecting a 100+ MW data center campus to transmission-level power requires grid upgrades that can take 18–24 months in their own right. The Saudi Electricity Company (SEC) lead times for large industrial connections are a real constraint on data center development velocity.

Water availability: Most large data centers use cooling towers that evaporate water to reject heat. In a country where water is scarce, this is both a cost issue (water costs money in Saudi Arabia) and a regulatory issue (environmental approvals for large water consumers take time). Dry cooling and adiabatic cooling alternatives are available but add capital cost.

Equipment procurement: The long-lead items for data center construction — large generators, transformers, switchgear, precision cooling units — are globally constrained. The surge in data center construction globally from 2023 onward has created 12–24 month lead times for some equipment categories. Saudi projects compete for the same global supply as US and European hyperscale programs.

Skilled labor: Hyperscale data center construction requires specialized subcontractors for raised floor systems, precision power distribution, building management systems, and fire suppression. This skills base exists in the US, Europe, and Singapore. It does not exist in Saudi Arabia at scale. Projects will need to import significant specialized labor, which has its own logistics, housing, and regulatory requirements under Saudi labor law.

Applying realistic timeline assumptions to Humain’s announced 18,000 GB300 Phase 1 delivery and its longer 600,000 GPU target over three years creates tension. GPU hardware can ship in months once export licenses are granted. Facilities to house that hardware at scale take two to four years to build. The gap between announced hardware commitments and operational compute capacity is, in large part, a construction gap.

Saudization and the Workforce Question

Saudi Arabia’s Vision 2030 includes ambitious targets for increasing Saudi citizen participation in the private sector workforce — a policy known as Saudization or Nitaqat. For construction, the target percentages are lower than for knowledge industries, but they still shape hiring and training costs for EPC contractors.

The more significant workforce issue for data center construction is the technical skill gap. Saudi Arabia is investing heavily in STEM education and technical training through institutions like KAUST, various technical colleges, and industry training programs. But the specific skills required for high-performance data center construction — computational fluid dynamics modeling for airflow optimization, electrical engineering for 33kV power distribution, control systems integration for building management — take years to develop in a workforce.

International EPC firms operating in Saudi Arabia are navigating a dual mandate: meet Saudization targets while maintaining the technical quality that hyperscale operators require. The pragmatic solution is tiered — Saudi nationals in project management, logistics, quality assurance, and entry-level roles, with specialized expatriate technical labor in roles where no equivalent Saudi skill base yet exists. Over the five-year buildout horizon, this balance should shift as the training investment bears fruit.

What EPC Reveals About the Announced-vs-Operational Gap

The EPC layer is where the gap between Saudi AI announcements and operational reality is most visible. When Humain announced $77 billion in AI compute commitments in May 2025, the announcement was a statement of intent and a financial commitment. It was not a statement that facilities were under construction, that power connections were contracted, or that EPC firms had signed contracts.

The sequence from announcement to operational capacity runs roughly: strategic announcement → site selection and land acquisition → utility negotiations → EPC procurement → design and permitting → civil construction → MEP installation → equipment installation → commissioning → operations. Each step takes months. Several steps depend on predecessor steps. For a $77 billion program to be operational on a 3–5 year timeline, EPC contracts at scale needed to be signed by late 2025 at the latest for the first facilities.

Tracking EPC contract awards — through Saudi government procurement databases, company announcements, and project registries — is therefore one of the most reliable leading indicators of when announced Saudi AI capacity will actually enter service. A project with signed EPC contracts and power agreements is structurally closer to operational than one where only a headline MOU exists.

The EPC sector also reveals where Saudi Arabia’s AI buildout is genuinely different from a marketing campaign: you cannot fake a poured foundation. When SBG or Bechtel breaks ground on a data center campus and begins pulling electrical permits, the construction is real. The sector’s physical evidence is a more reliable signal than press releases — and tracking it is a core function of a serious Saudi AI intelligence platform.

---

Tracked entities in this sector include Bechtel, Saudi Binladin Group, Alfanar, Turner Construction, Jacobs Engineering, and major Saudi civil contractors active in technology infrastructure. Sector coverage focuses on contract awards, groundbreaking events, construction progress, and the commissioning milestones that mark the transition from announced to operational compute capacity.