The Power Cost Equation That Defines Saudi AI Competitiveness
When Humain — the Saudi sovereign AI company backed by PIF — announced plans for a 500-megawatt AI data centre campus, the most consequential number in the entire project was not the GPU count or the rack density. It was the cost of electricity. Power economics determine whether Saudi Arabia can credibly position itself as a global AI infrastructure hub or whether its compute ambitions remain confined to serving domestic demand. The answer, when the electricity pricing is examined in detail, is more nuanced than either Saudi promotional materials or sceptical outside commentary typically acknowledge.
Saudi Arabia’s electricity pricing for industrial and commercial consumers is governed by the Saudi Electricity Company (SEC), which operates as a regulated monopoly distribution utility. The pricing structure uses tiered tariffs that vary by consumption volume, connection voltage, and customer classification. Large industrial consumers — including data centres classified under the commercial-industrial tariff schedule — access electricity at rates that are among the most competitive in the world for grid-supplied power, though the full cost picture requires accounting for cooling infrastructure and the reality of desert operating conditions.
SEC Industrial Tariff Structure for Data Centres
The Saudi Electricity Company’s tariff framework for large commercial and industrial consumers establishes rates that data centre operators access at the high-voltage connection tier. For facilities connecting at 115 kV or above — the appropriate connection voltage for a data centre drawing 100+ MW — the effective energy charge falls in the range of $0.038 to $0.057 per kilowatt-hour, depending on consumption volume and any applicable demand response agreements with SEC.
This rate structure reflects Saudi Arabia’s position as one of the world’s largest producers of hydrocarbons, with domestic natural gas prices set by government policy at levels that make gas-fired power generation extraordinarily cost-competitive by international standards. The Kingdom’s gas pricing for power generation has historically been subsidised at levels that translated into electricity production costs well below international norms, and while subsidy rationalisation has occurred since 2016, industrial power prices remain significantly below European or even US commercial rates.
For a data centre drawing 100 MW continuously — a facility at the lower end of the hyperscale range — the annual electricity cost at $0.048/kWh blended rate amounts to approximately $42 million per year. At the 500 MW scale of Humain’s planned campus, total electricity expenditure at this rate would run to approximately $210 million annually before accounting for any renewable PPA arrangements or demand response credits. This is the baseline against which alternative structures are being evaluated.
The demand charge component of SEC’s industrial tariff — a fixed monthly charge per kW of peak demand — adds approximately $5–8 per kW per month for high-voltage industrial connections. For a 500 MW facility operating at high utilisation, this adds $30–48 million annually to the electricity cost structure, bringing the total effective cost of grid power to approximately $0.053–0.063/kWh all-in before cooling is factored in.
Benchmarking Against Global Data Centre Power Markets
Saudi Arabia’s grid power pricing sits at a meaningful discount to every major data centre market outside a handful of countries with exceptional renewable resources. In the UAE, data centre operators in Abu Dhabi and Dubai access electricity at rates of $0.05 to $0.08 per kWh for large commercial consumers, with the higher end of that range applying to facilities that have not secured direct renewable agreements. The Gulf comparison is instructive: despite similar hydrocarbon wealth, UAE electricity pricing for data centres is 20–40% higher than Saudi rates for equivalent facilities.
US data centre power pricing varies enormously by geography — from roughly $0.035/kWh in the Pacific Northwest (where hydropower keeps rates low) to $0.12/kWh or higher in California and the Northeast. The US national average for commercial electricity is approximately $0.11/kWh, making the typical US data centre’s power cost roughly double the Saudi baseline. Northern Virginia, which hosts the highest concentration of hyperscale data centre capacity in the world, sees prices of $0.06–0.08/kWh for large consumers, partially offset by the scale efficiencies of mature infrastructure ecosystems.
European data centre markets face the most challenging power economics. Germany’s industrial electricity prices range from $0.12 to $0.18/kWh, reflecting the energy transition costs embedded in the German electricity price. The Netherlands, one of Europe’s largest data centre hubs, sees rates of $0.10–0.15/kWh for commercial data centre loads. Scandinavia offers the best European pricing — Sweden and Finland provide rates of $0.05–0.08/kWh for large consumers with access to substantial hydropower — making it the primary European market where Saudi pricing is competitive on power cost alone rather than clearly superior.
The benchmarks that matter most competitively are not Europe or the US, however. The relevant comparison is against the cheapest power markets globally, because these are the locations that ultra-cost-sensitive AI training workloads will migrate to if Saudi Arabia charges premium prices. Iceland, with geothermal and hydropower resources providing electricity at approximately $0.03–0.04/kWh for large industrial consumers, represents the global floor for grid-sourced renewable power in a stable jurisdiction. Paraguay (Itaipu hydropower: ~$0.02/kWh) and certain Canadian provinces (Quebec hydro: ~$0.04/kWh) are in similar territory. Saudi Arabia’s grid pricing cannot match these outliers, but its renewable PPA potential changes the calculation.
ACWA Power and the Renewable PPA Opportunity
The most significant factor in Saudi Arabia’s long-term electricity cost competitiveness for data centres is not the SEC grid tariff but the availability of long-term power purchase agreements from solar generation assets at costs that are among the lowest in the world. Saudi Arabia’s solar irradiance — averaging 2,200+ kWh per square metre annually across the central and western regions — is among the highest of any large land mass on earth, and ACWA Power has demonstrated in competitive tenders that this resource can be converted to electricity at costs that are globally without precedent.
ACWA Power’s Sudair solar project, commissioned in 2023 with a capacity of 1,500 MW, was contracted at a levelised cost of electricity of approximately $0.0104/kWh — the lowest confirmed solar PPA price ever recorded for a utility-scale project at the time of signing. Even accounting for the need to add battery storage or grid backup for firm power delivery to a data centre with 99.9%+ uptime requirements, the total cost of firm renewable power at scale in Saudi Arabia can be structured in the range of $0.025–0.035/kWh through a carefully structured PPA that combines a base solar contract with modest grid backup insurance.
For a data centre operator willing to commit to a 15–20 year PPA with ACWA Power or a comparable developer, the economics of Saudi Arabia improve dramatically relative to the SEC grid tariff baseline. At $0.028/kWh for firm renewable power delivered under a long-term PPA, a 500 MW facility’s annual electricity cost falls to approximately $122 million — a saving of roughly $88 million annually versus pure grid supply at the blended tariff rate. Over a 15-year contract period (net present value at 8% discount rate), that saving represents approximately $750 million in present value — a number large enough to drive location decisions for major data centre investments.
ACWA Power’s PPA structure for large data centre campuses typically includes minimum offtake commitments, firming mechanisms that use grid backup or battery storage to guarantee availability, and pricing structures that include capacity charges alongside energy charges. The bankability of these agreements — ACWA Power’s investment-grade credit rating and Saudi government sovereign backing — makes them acceptable as infrastructure for project finance structures.
NEOM’s 26 GW Renewable Target and Oxagon Implications
NEOM, the $500 billion giga-project on Saudi Arabia’s northwest Red Sea coast, has announced ambitions for 26 gigawatts of renewable energy generation — approximately 70% of Saudi Arabia’s current total installed generating capacity — to power its planned cities and industrial zones. The power infrastructure at NEOM is intended to be 100% renewable from inception, with a combination of solar, wind (the NEOM region has exceptional wind resources compared to central Saudi Arabia), and green hydrogen production serving as both the energy source and an export commodity.
Oxagon, NEOM’s planned floating industrial city on the Red Sea, is conceived as a zero-carbon industrial platform where energy-intensive manufacturing and data processing colocate with renewable generation at unprecedented scale. For AI compute workloads, Oxagon’s positioning is distinctive: it offers potential access to some of the cheapest renewable power in the world at the point of generation, combined with direct subsea cable connectivity to Europe and Asia through Red Sea routing.
The practical implication for data centres at NEOM is that power pricing could be structured at near-generation-cost levels for anchor tenants willing to collocate at Oxagon’s planned data centre zones — potentially $0.020–0.025/kWh for direct-connection renewable power, competitive with the cheapest alternatives available anywhere. The caveat is execution risk: NEOM’s timelines have been revised repeatedly, its scale has been partially reduced from original announcements, and the 26 GW renewable target remains a planning aspiration rather than a funded construction programme.
The Desert Cooling Cost Adder
Saudi Arabia’s electricity pricing advantage for data centres must be qualified by the cooling cost reality of operating in an extreme desert climate. Riyadh’s summer ambient temperatures regularly exceed 45°C, with dry-bulb temperatures above 40°C persisting for four to five months annually. Jeddah, on the Red Sea coast, adds high humidity to heat, creating a wet-bulb temperature environment that is arguably more challenging for data centre cooling than Riyadh’s dry heat.
Traditional air cooling approaches — the dominant cooling technology in temperate-climate data centres — become progressively less effective and energy-intensive as ambient temperatures rise. A data centre achieving a Power Usage Effectiveness (PUE) of 1.15 in a 15°C ambient environment (northern Europe) might achieve only 1.35–1.45 PUE in Riyadh’s summer conditions using equivalent cooling infrastructure. The additional electricity consumed by cooling equipment — compressor chillers running at higher lift, cooling tower systems with high evaporation rates — adds an effective $0.010–0.020/kWh to the total cost of delivering power to IT equipment.
Advanced cooling technologies mitigate but do not eliminate this penalty. Direct liquid cooling (DLC), which delivers chilled water directly to server CPU and GPU thermal interfaces, can maintain IT equipment at safe operating temperatures with greatly reduced dependence on room-level air cooling, bringing PUE in desert conditions down toward 1.20–1.25. Immersion cooling, where servers are submerged in dielectric fluid, achieves PUE values approaching 1.03–1.05 regardless of ambient conditions but requires significant capital investment in facility infrastructure and limits hardware flexibility.
Humain’s 500 MW campus planning incorporates DLC as the primary cooling technology for AI GPU clusters, with immersion cooling for the highest-density inference nodes. This pushes the effective cooling overhead to approximately $0.012/kWh added to the base electricity cost — significant, but well within the margin of advantage that Saudi grid pricing maintains over European and US alternatives.
Total Effective Power Cost for Humain’s 500 MW Campus
Synthesising the grid tariff, renewable PPA potential, and cooling cost adder produces a range of total effective power costs for Humain’s planned 500 MW AI data centre campus. The scenarios reflect different assumptions about the mix of grid power and renewable PPA supply, and the cooling technology deployed.
Under a grid-heavy scenario — 80% SEC grid power at $0.050/kWh blended, 20% renewable PPA at $0.030/kWh, advanced air cooling with PUE of 1.30 — the total effective cost of power delivered to IT equipment is approximately $0.068/kWh. At 500 MW of IT load operating at 85% average utilisation, this implies annual power expenditure of approximately $253 million.
Under an optimised renewable PPA scenario — 60% long-term ACWA Power solar PPA at $0.028/kWh, 40% grid backup at $0.050/kWh, DLC cooling at PUE of 1.22 — the total effective cost falls to approximately $0.051/kWh, implying annual power expenditure of approximately $190 million. This is the economically rational endpoint for a large campus with long time horizon.
For reference, a comparable 500 MW AI campus in Northern Virginia would pay approximately $0.070–0.080/kWh total effective cost. In the UK, $0.13–0.16/kWh. In Germany, $0.15–0.20/kWh. The Saudi advantage, even in the conservative scenario, is material and structurally durable for the foreseeable planning horizon.
Power as a Percentage of Total Data Centre OPEX
In the context of AI data centre economics, understanding power cost as a fraction of total operating expenditure clarifies how much the electricity pricing advantage actually matters to the total cost of ownership. For GPU-dense AI training and inference facilities, power typically represents 35–40% of total annual operating expenditure when hardware amortisation, staffing, network, maintenance, and facility costs are included.
At a well-run AI data centre, the rough OPEX decomposition looks like: power 35–40%, hardware depreciation/amortisation 30–35% (for owned hardware — more if leased), network and connectivity 8–12%, staffing and management 8–12%, maintenance and facilities 6–10%. Power’s share is higher than in general-purpose data centres because AI GPU clusters have significantly higher power density per rack than conventional server infrastructure, and because GPU utilisation rates in AI training workflows are very high, minimising the fraction of time that hardware is idle and drawing low power.
This 35–40% share means that a 25% reduction in electricity cost — achievable by moving from a US or European market to Saudi Arabia’s optimised renewable PPA scenario — reduces total data centre OPEX by 8.75–10 percentage points. At the scale of Humain’s 500 MW campus, with total annual OPEX in the range of $500–600 million (including hardware amortisation), a 9% OPEX reduction represents $45–55 million in annual savings. Over a 10-year planning horizon, this compounds to a present-value advantage of $300–400 million.
Outlook: Pricing Stability and Vision 2030 Renewable Targets
Saudi Arabia’s Vision 2030 energy targets include reaching 50% renewable generation by 2030, up from approximately 2% in 2020. This trajectory — if achieved — has important implications for data centre power cost stability. As renewable generation displaces gas-fired generation in the SEC system, the marginal cost of grid power falls and becomes less sensitive to global gas price fluctuations. Data centres that are currently exposed to indirect gas price risk through SEC’s tariff structure would benefit from a progressively renewables-dominated grid.
The 50% renewable target implies approximately 58 GW of renewable capacity additions by 2030, dominated by solar but including significant onshore wind development in the NEOM and Tabuk regions. The PPA market for large offtakers — including data centre operators — will deepen as more generation capacity becomes available from a broader set of developers beyond ACWA Power. Increased competition among renewable developers bidding for large industrial offtake should maintain downward pressure on PPA prices even as the lowest-cost sites are developed first.
For data centre operators making location decisions with 10–15 year horizons, Saudi Arabia’s power pricing outlook — stable grid tariffs with the potential for declining renewable PPA costs over the planning period — compares favourably to markets where electricity prices are subject to significant policy uncertainty, carbon pricing exposure, or infrastructure capacity constraints that could push prices upward as AI power demand grows.