DataVolt AI Factory — NEOM Oxagon: The 1.5 GW Net-Zero Ambition

The DataVolt AI Factory at NEOM’s Oxagon industrial complex is the most ambitious single AI infrastructure project in Saudi Arabia’s pipeline and, in power specification, among the most ambitious data center projects announced globally. At 1.5 gigawatts of planned capacity targeting 2028 delivery, the DataVolt facility is simultaneously Saudi Arabia’s largest planned AI compute project and the most rigorous test of NEOM’s energy infrastructure ambitions. Its success or failure will determine whether Saudi Arabia’s AI infrastructure claims extend to net-zero sustainability — a competitive differentiator that becomes more commercially valuable as global enterprise customers face increasing carbon accounting obligations.

To understand the DataVolt project, the NEOM Oxagon context is essential. Oxagon is one of NEOM’s multiple planned urban developments — an industrial city anchored to the Red Sea coast in the Tabuk region of northwestern Saudi Arabia, with portions of the complex extending over the water as floating platforms. The Oxagon concept envisions advanced manufacturing, clean technology, and digital infrastructure co-located with net-zero energy generation from the Tabuk region’s abundant solar and wind resources. In the hierarchy of NEOM’s ambitious urban development projects, Oxagon represents the industrial and infrastructure complement to The Line’s residential ambition and Sindalah’s luxury tourism offering.

Scale in Context: What 1.5 GW Actually Means

The 1.5 GW capacity specification requires careful interpretation because data center capacity figures can reflect different things: IT load capacity (the power consumed by compute equipment), total facility power (including cooling, lighting, and other facility systems), or connected load capacity (the maximum power the facility is designed to handle under peak conditions). Assuming the industry-standard convention that 1.5 GW refers to IT load, the AI workload implications are substantial.

At modern AI training densities — 80-100 kW per rack for NVIDIA GB300 or comparable hardware in liquid-cooled configurations — 1.5 GW of IT load supports approximately 15,000-18,750 high-density AI training racks. At 72 GPU chips per GB300 system (one per rack), this implies capacity for 1.08 million to 1.35 million Blackwell-generation GPU chips. By way of comparison, OpenAI’s reported training cluster for GPT-4 involved approximately 25,000 A100 GPUs; the DataVolt facility could host fifty to fifty-five equivalent clusters simultaneously.

The scale comparisons within Saudi Arabia’s own facility pipeline are equally illustrative:

  • DataVolt NEOM (1.5 GW) vs HUMAIN Riyadh Phase 1 (200 MW): 7.5x larger
  • DataVolt NEOM (1.5 GW) vs Hexagon DC (480 MW): 3.1x larger
  • DataVolt NEOM (1.5 GW) vs xAI Campus (500 MW planned): 3x larger
  • DataVolt NEOM (1.5 GW) vs entire current Center3 campus (100 MW): 15x larger

These comparisons establish DataVolt as occupying a different order of magnitude from the other Saudi facilities — not just the largest, but categorically different in scale. This is not incremental infrastructure expansion; it is a step-change in Saudi Arabia’s total AI compute capacity that would, if operational, make Saudi Arabia one of the world’s most significant AI compute locations by a considerable margin.

DataVolt as EPC Contractor: Construction at Unprecedented Scale

DataVolt’s role as EPC (Engineering, Procurement, and Construction) contractor for the Oxagon facility means DataVolt bears the project delivery risk — designing the facility, procuring the equipment, managing construction, and delivering a commissioned facility to the client. This is a different model from conventional data center development where the owner/operator handles development and uses construction management services; DataVolt is the project delivery specialist responsible for turning NEOM’s specifications into operational infrastructure.

DataVolt’s track record in large-scale data center EPC work is the execution credential that justifies this role. Building a 1.5 GW facility is not simply scaling up a smaller project: it involves power infrastructure at utility scale (multiple dedicated substations, potentially requiring new transmission infrastructure), civil engineering for a facility that may span millions of square feet, and construction logistics at a remote coastal site in northwestern Saudi Arabia that lacks the mature contractor ecosystem of Riyadh or Dammam. The construction challenge is as much about supply chain management, labor logistics, and project coordination as it is about technical design.

The 2028 delivery target gives DataVolt approximately three years from project announcement to operational status for a 1.5 GW facility. Even for experienced large-scale data center EPC contractors, this is an aggressive timeline. Phased delivery is the most operationally rational approach: early phases of 200-300 MW coming online in 2026-2027, with subsequent phases completing through 2028. This phased approach also allows the project to demonstrate operational credibility before the full investment is deployed — reducing financial risk for NEOM and DataVolt if early phases underperform against commercial expectations.

Net-Zero Energy Architecture: Green Hydrogen as the Firm Power Solution

NEOM’s net-zero energy commitment for Oxagon is the project’s most distinctive feature and its most technically challenging claim. The Tabuk region has world-class solar irradiance — among the highest in the world, reliably exceeding 2,000 kWh/m²/year — and consistent wind resources from the Gulf of Aqaba corridor. These resources can generate power at low marginal cost, but they cannot alone power a reliable AI compute facility: solar stops at night, wind is variable, and AI training jobs require uninterrupted power measured in days and weeks.

Green hydrogen solves the reliability problem by converting excess renewable generation into storable energy. The NEOM Green Hydrogen Company joint venture — combining NEOM’s project rights, ACWA Power’s renewable energy development expertise, and Air Products’ industrial gas technology — is developing large-scale electrolysis at Oxagon to produce green hydrogen from solar and wind power. This hydrogen can be stored and reconverted to electricity through fuel cells or hydrogen-capable gas turbines, providing firm, dispatchable, zero-carbon power for periods when solar and wind generation is insufficient.

For the DataVolt AI factory specifically, the reliability requirement is stringent. A transformer training run on 100,000 GPU chips running for three weeks cannot tolerate power interruptions — even a 30-second power event that triggers UPS switchover will typically cause a training job to fail and require restart from the most recent checkpoint, potentially wasting days of compute time. The green hydrogen firm power architecture must achieve the same power quality as conventional utility power, not merely the same average generation volume. NEOM’s energy engineers are designing the Oxagon power architecture with grid-quality reliability as a non-negotiable requirement alongside the net-zero carbon objective.

Competitive Differentiation vs UAE: The Carbon Premium

NEOM’s net-zero positioning for the DataVolt facility represents a calculated competitive differentiation strategy against UAE’s data center ecosystem. The UAE — particularly Abu Dhabi with G42’s Microsoft-partnered campus and Mubadala’s data center assets, and Dubai with its large commercial colocation market — has been the dominant regional AI and cloud infrastructure hub. UAE data centers operate primarily on fossil-fuel-generated grid power, with renewable energy procurement programs that offset some carbon but do not constitute net-zero AI compute.

As global enterprise buyers face tightening carbon accounting requirements, the ability to certify AI compute workloads as genuinely net-zero becomes commercially valuable and increasingly necessary. European enterprises reporting under the Corporate Sustainability Reporting Directive (CSRD), multinationals committed to Science Based Targets (SBT) net-zero by 2030, and sovereign wealth fund portfolio companies with ESG mandates all face requirements that make the carbon intensity of AI compute relevant to infrastructure procurement decisions.

DataVolt’s 1.5 GW net-zero facility at NEOM is the asset that enables Saudi Arabia to offer what UAE cannot: AI compute at scale with verifiable carbon neutrality. The addressable market for premium-priced net-zero AI compute is growing as carbon accounting frameworks tighten, and Saudi Arabia’s renewable energy economics in the Tabuk region — where net-zero power can be produced at competitive cost due to exceptional solar and wind resources — mean the net-zero premium may be smaller than in locations where renewable energy carries a large cost premium over fossil alternatives.

2028 Timeline and Saudi Arabia’s AI Infrastructure Maturity

The 2028 delivery target positions the DataVolt facility as a second-wave Saudi AI infrastructure asset — coming online after the first-wave HUMAIN, Hexagon, and Center3 facilities have established Saudi Arabia’s AI compute credentials. By 2028, Saudi Arabia will have demonstrated operational competence in large-scale AI factory management, developed the technical workforce to operate compute at scale, and built the commercial relationships with hyperscalers and AI companies that generate demand for additional capacity.

From this perspective, DataVolt-NEOM’s timing may be optimal: it avoids the execution risk of being the first large-scale facility in an underdeveloped market, while capturing demand that the first-wave facilities cannot satisfy at their scale. If Saudi Arabia’s first-wave facilities demonstrate the operational credibility and commercial viability of Saudi AI compute, the demand pipeline for DataVolt’s 1.5 GW will be better-defined in 2028 than it is in 2025. Saudi Arabia’s AI infrastructure ambition targets gigawatt-scale capacity; DataVolt adds a single 1.5 GW facility that alone represents the majority of the first-wave capacity. See Infrastructure for the complete Saudi facility capacity timeline.

Green Hydrogen Economics: The Energy Storage Calculation

The economics of green hydrogen as AI data center power supply are worth examining quantitatively to assess DataVolt-NEOM’s financial viability. Green hydrogen production via electrolysis requires approximately 50-55 kWh of electricity per kilogram of hydrogen produced. Converting hydrogen back to electricity via fuel cell or gas turbine adds further efficiency losses — fuel cells achieve 50-60% electrical efficiency, turbines achieve 40-50%. The round-trip energy efficiency of hydrogen storage is therefore approximately 25-35%: for every 100 kWh of solar or wind energy converted to hydrogen and back to electricity, 25-35 kWh of usable electricity is recovered.

This round-trip efficiency penalty means green hydrogen electricity costs significantly more than direct solar or wind electricity at the point of generation. For DataVolt-NEOM’s AI compute facility, this premium is justified by the reliability value: solar and wind alone cannot provide continuous power, and the alternative to green hydrogen firm power is diesel backup generation (expensive, polluting, and operationally complex at 1.5 GW scale) or grid connection to fossil fuel power infrastructure (negating the net-zero claim). Green hydrogen’s higher cost per kWh is the price of reliability in a net-zero power architecture.

The NEOM Green Hydrogen Company’s scale — designed as one of the world’s largest green hydrogen production facilities — creates cost reduction potential through manufacturing scale effects. Electrolysis equipment costs have been declining at approximately 10-15% per year as the green hydrogen industry scales, and Oxagon’s large-scale installation will benefit from the lowest available capital costs for electrolysis equipment at the time of construction. The economics of green hydrogen power for DataVolt-NEOM in 2028 will be materially better than the economics at 2025 cost levels, which is one reason the 2028 delivery target is relevant beyond construction timeline considerations.

Construction Logistics: Remote Site Challenges

The DataVolt-NEOM project faces construction logistics challenges that distinguish it from every other facility in Saudi Arabia’s AI infrastructure pipeline. The Tabuk region — while not inaccessible — lacks the mature construction contractor ecosystem, electrical equipment supply chains, and specialized labor pools that Riyadh and the Eastern Province offer. Every major piece of equipment — power transformers, cooling systems, server hardware, networking infrastructure — must be transported to the site via road or sea from suppliers and ports that are hundreds of kilometers away.

NEOM has invested significantly in transportation infrastructure for its construction programs, including the expansion of Sharma Airport near NEOM’s Tabuk project area, port development on the Red Sea coast, and road construction connecting NEOM sites to the regional highway network. DataVolt’s EPC execution plan will leverage this NEOM transportation infrastructure, but the remoteness premium is real: construction costs at Tabuk are higher than at comparable Riyadh or Dammam sites, and the construction timeline is more sensitive to supply chain disruptions because there is less local buffer stock to compensate for delivery delays.

The labor logistics challenge is equally significant. Saudi Arabia does not have a large pool of specialized data center construction labor available at Tabuk — electricians experienced in high-voltage data center systems, mechanical engineers for cooling infrastructure commissioning, and IT infrastructure technicians for technology installation all need to be recruited globally or transported to site from Riyadh and other Saudi cities. NEOM has experience managing this logistics challenge from its construction programs across multiple Tabuk-region sites, and DataVolt will benefit from this institutional knowledge, but the labor logistics complexity adds cost and schedule risk relative to established Saudi data center construction markets.

The Saudi-UAE Net-Zero AI Competition

DataVolt-NEOM’s net-zero positioning is best understood within the context of a broader competition between Saudi Arabia and the UAE for global AI infrastructure investment and tenant relationships. The UAE, particularly through Abu Dhabi’s G42 and Mubadala infrastructure programs, has been the Gulf’s dominant AI infrastructure hub, attracting Microsoft’s $3.2 billion regional commitment and establishing Falcon LLM at TII as the Arab world’s leading open-source language model.

Saudi Arabia’s response to UAE’s AI infrastructure head start has multiple dimensions, of which DataVolt-NEOM’s net-zero AI compute is one of the most differentiated. UAE’s data centers operate primarily on the UAE grid, which draws from a mix of natural gas generation, nuclear (Barakah plant), and solar (Mohammed bin Rashid Al Maktoum Solar Park). Even with nuclear and solar contributions, UAE grid power is not net-zero, and UAE data center operators seeking net-zero certification must rely on renewable energy certificates rather than genuine on-site or dedicated renewable generation.

DataVolt-NEOM’s direct-connected renewable generation — solar, wind, and green hydrogen firm power at the Oxagon site — enables a genuinely net-zero carbon certification that UAE’s grid-connected facilities cannot match on current terms. As European and American enterprise customers face tightening Scope 3 emissions reporting requirements, this certification quality difference will become commercially significant. DataVolt-NEOM is positioned to capture the premium market segment of sustainability-committed AI compute customers — a segment that will grow as carbon accounting becomes mandatory rather than voluntary for major multinational corporations.