ACWA Power: The Energy Backbone of Saudi AI Infrastructure
Every data center is ultimately a power plant in reverse: instead of generating electricity, it consumes it at scale and converts electrical energy into computation. The most sophisticated AI training cluster is inert without reliable, affordable, high-voltage power infrastructure connected to it. This is why ACWA Power — Saudi Arabia’s largest renewable energy developer and a majority-owned PIF entity — is not a peripheral player in the Saudi AI compute buildout. It is a structural enabler, and its execution determines whether Saudi Arabia’s $77 billion AI investment can deliver at the scale being promised.
ACWA Power was founded in 2004 as a private developer of water and power projects in Saudi Arabia and has grown into one of the largest independent power producers in the emerging markets world. Today it operates a portfolio of more than 70 projects across 13 countries, with total generation capacity of over 41 GW across renewable energy, conventional thermal, and desalination assets. The company went public on the Saudi Exchange (Tadawul) in 2021 in one of the largest IPOs of that year, but the Public Investment Fund retains majority ownership — a structural alignment with the Humain AI buildout that has significant implications for how power supply for AI data centers will be priced and prioritized.
The Power Calculus for AI Compute
The scale of Saudi Arabia’s AI compute ambitions creates power demand that must be contextualized carefully. The stc-Humain joint venture alone targets 1 GW of data center capacity. The SDAIA Hexagon DC operates at 480 MW. Humain’s full buildout — including the 600,000 NVIDIA GPU deployment over three years — will require additional gigawatts. The DataVolt partnership at NEOM targets 1.5 GW. Adding it up, Saudi Arabia’s AI infrastructure pipeline implies 3-5 GW of new, dedicated power demand over the next five years.
For context: 1 GW of continuous power is roughly equivalent to the electricity consumption of a city of 750,000 people. Saudi Arabia is planning to add the equivalent of several medium-sized cities’ worth of power demand, primarily for AI compute, in a compressed timeframe. This is not business-as-usual grid planning — it requires deliberate, coordinated infrastructure investment at a scale that only a sovereign-backed developer like ACWA Power is positioned to execute.
The power economics matter enormously for the competitiveness of Saudi AI compute as a global service. If data center operators in Saudi Arabia can access electricity at $0.02-$0.03/kWh (achievable with large-scale solar plus storage in the Saudi desert), they have a structural cost advantage over equivalently-equipped facilities in Europe ($0.08-$0.15/kWh) or the US ($0.04-$0.07/kWh). Cheap power is not sufficient for AI competitiveness — talent, connectivity, and regulatory environment matter too — but it is a necessary condition, and ACWA Power is the entity responsible for delivering it.
The NEOM DataVolt Partnership
Among ACWA Power’s most significant AI-adjacent projects is the partnership with DataVolt, a specialized data center developer, to build a 1.5 GW data center campus at NEOM — the futuristic city being developed in northwestern Saudi Arabia. This would be among the largest single-site data center deployments in the world, and its power supply would rely on NEOM’s renewable energy infrastructure, including the Neom Green Hydrogen project (which ACWA Power is developing alongside Air Products and NEOM).
The strategic logic of combining NEOM’s renewable energy infrastructure with AI data center capacity is elegant but technically complex. NEOM is being built in a region with exceptional solar irradiance and wind resources — genuinely among the best renewable resource sites in the world. The challenge is intermittency: solar generates power during daylight hours but AI compute runs 24/7. Bridging the gap requires either storage (batteries or pumped hydro), transmission connections to a broader grid, or firm power backup — typically natural gas in Saudi Arabia’s case.
The green hydrogen project at NEOM is intended to provide a form of long-duration energy storage: excess renewable power converts water to hydrogen via electrolysis, which is then stored and converted back to electricity (or used directly) when renewables are insufficient. This is technically viable but economically challenging at current hydrogen production and storage costs. The real-world economics of the 1.5 GW DataVolt-ACWA facility will depend heavily on how the power supply solution evolves over the next two to three years.
PIF Alignment and the Sovereign Stack
ACWA Power’s majority ownership by PIF creates a structural alignment that sets it apart from purely commercial power developers. PIF owns Humain (the AI infrastructure vehicle), has major stakes in NEOM, and is the controlling shareholder of multiple entities involved in the AI buildout. This means ACWA Power is not negotiating at arm’s length with Humain for power purchase agreements — it is, in effect, one arm of the sovereign investment apparatus negotiating with another.
This can accelerate execution: when a single sovereign fund controls both the power developer and the compute infrastructure owner, interagency coordination barriers that would slow private-sector dealings are reduced. A data center needing 500 MW of new power capacity in a new location — a process that might take five to seven years in a Western market navigating permitting, grid studies, and competitive procurement — can be compressed to two to three years in Saudi Arabia when the same ultimate owner controls both sides.
The risk is that alignment replaces discipline. Commercial power purchase agreements impose financial accountability on both developer and offtaker. If ACWA Power and Humain operate on terms that are not commercially realistic — underpriced power, lenient performance standards, deferred payments — the capital efficiency of the AI buildout will suffer even if the physical infrastructure gets built.
Saudi Arabia’s Renewable Energy Targets and AI Demand
Saudi Arabia has committed to generating 50% of its electricity from renewable energy by 2030, up from roughly 3% today. This is an extraordinarily aggressive target that requires adding tens of gigawatts of solar and wind capacity in under a decade. ACWA Power is the primary vehicle for that buildout, having won the majority of Saudi Arabia’s large-scale renewable energy tender awards.
The AI compute buildout and the renewable energy transition are, in the Saudi strategic framework, the same story told from different perspectives. The oil wealth generates the capital. The capital funds both the renewable energy build and the AI compute build. The renewable energy provides cheap, clean power for the compute. The compute generates the AI capabilities that help Saudi Arabia diversify its economy away from oil. The loop is intentional and coherent.
In practice, the renewable energy build is running behind the pace needed to power AI compute at full scale by 2030. Large-scale solar projects (the Al Shuaibah project, the Sudair solar farm) are proceeding, but grid infrastructure, storage deployment, and permitting for new sites remain bottlenecks. Saudi Arabia will likely rely on natural gas backup power for AI data centers for much of the 2025-2028 period, transitioning to higher renewable shares as the solar and storage buildout catches up.
This is not a disqualifying limitation — the US and EU data center industries are also overwhelmingly powered by fossil fuels today, despite renewable energy commitments. But it is a reality that AI compute operators and their customers should factor into sustainability claims.
ACWA Power’s Project Portfolio: Select AI-Adjacent Assets
Beyond NEOM, ACWA Power’s portfolio includes several projects with direct AI compute relevance:
The Sudair Solar Energy Project (1.5 GW) near Riyadh, developed with ARAMCO, is one of the largest single-site solar projects in the world. Its location near Riyadh, where multiple data center campuses are being developed (including Hexagon DC and Humain’s initial facilities), makes it a natural power source for near-term AI infrastructure.
The NEOM Green Hydrogen Project, a joint venture with Air Products and NEOM, targets 4 GW of electrolysis capacity powered by 4 GW of wind and solar. If the hydrogen production economics improve to the point where stored hydrogen can reliably provide firm power, it would enable genuinely round-the-clock renewable-powered AI compute at NEOM.
Thermal desalination plants developed by ACWA Power throughout the Gulf also consume and produce energy in ways that interact with the AI compute buildout — wastewater from data center cooling systems, for example, can potentially feed desalination processes in a circular economy configuration.
Execution Risks and Competitive Dynamics
ACWA Power’s execution risks fall into three categories: supply chain, regulatory, and geopolitical.
Supply chain: Large-scale solar projects require solar panels, inverters, and transmission equipment. The global supply chain for these components is heavily dominated by Chinese manufacturers, and US-Saudi AI agreements (notably the NVIDIA deal and Humain’s US partnerships) occur in a geopolitical environment where US officials are scrutinizing supply chain dependencies. If US pressure on Saudi Arabia to reduce Chinese technology dependence extends to energy infrastructure, ACWA Power’s procurement options become more constrained and expensive.
Regulatory: Saudi Arabia’s electricity market is in transition, moving from a fully state-controlled system toward a competitive market. The regulatory framework for independent power producers, power purchase agreement terms, and grid access tariffs is still evolving. Changes in regulatory policy could affect the economics of projects that ACWA Power has committed to at specific return assumptions.
Geopolitical: ACWA Power operates in 13 countries, including several conflict-affected markets. Project delays or losses in its international portfolio don’t directly affect Saudi AI infrastructure, but they affect the company’s financial position and management attention.
Water and Cooling: The Hidden Infrastructure Dependency
Power is the dominant infrastructure input for AI data centers, but water is close behind. AI GPU clusters generate enormous heat that must be dissipated, and the most efficient cooling systems use water — either direct liquid cooling applied to server components or large-scale cooling towers that evaporate water to reject heat to the atmosphere. A 1 GW data center campus can require millions of gallons of water per day for cooling, depending on the cooling technology deployed.
Saudi Arabia is a water-scarce nation. Its aquifers are ancient and depleting; its surface water resources are minimal. Municipal water supply relies heavily on desalination — energy-intensive conversion of Red Sea and Arabian Gulf seawater — supplemented by highly treated wastewater for non-potable uses. Using desalinated water for data center cooling at gigawatt scale would represent a significant additional strain on already constrained water infrastructure.
This is where ACWA Power’s desalination portfolio becomes directly relevant to the AI compute buildout in a way that is not immediately obvious. ACWA Power is Saudi Arabia’s largest private desalination plant operator, operating multiple large-scale thermal and reverse osmosis desalination plants that supply municipalities and industrial users. A data center campus co-located with an ACWA-operated desalination and wastewater treatment complex could access treated water for cooling, with the wastewater from the cooling process recycled back into the desalination plant’s input stream — a circular water economy configuration that reduces net freshwater consumption.
This kind of integrated energy-water-compute infrastructure design is technically feasible and economically attractive if planned from the outset. NEOM’s development framework, which incorporates water recycling, renewable energy, and industrial co-location as design principles, is the most likely venue for such integration. ACWA Power’s presence in both the NEOM power supply (through the green hydrogen project) and its broader desalination portfolio makes it the natural operator for this integrated infrastructure approach.
Regulatory Coordination and Saudi Grid Stability
ACWA Power’s ability to deliver power to AI data centers at scale depends not only on building generation assets but on the Saudi electricity grid’s ability to absorb and transmit that power reliably. Large data centers are not typical grid loads: they draw power continuously at very high levels, with relatively low tolerance for voltage fluctuations or brief outages that industrial manufacturing plants can often ride through with appropriate UPS systems.
AI training workloads are particularly sensitive: a power interruption during a multi-week training run can corrupt model checkpoints, wasting enormous amounts of compute time and effectively making the interruption far more costly than the energy cost of the outage itself. Cloud service providers that have signed agreements with Humain or other Saudi data center operators will have demanding power reliability requirements — five nines uptime (99.999%) or better — that require both redundant power generation and grid infrastructure capable of maintaining stable frequency and voltage.
Saudi Arabia’s grid is operated by Saudi Electricity Company (SEC), a separate entity from ACWA Power. The coordination between ACWA Power (as generator), SEC (as grid operator), and data center operators (as major new industrial loads) is a three-way coordination challenge that requires regulatory oversight from the Electricity & Cogeneration Regulatory Authority (ECRA). MCIT, SDAIA, and Humain are not energy regulators; their ability to deliver on data center commitments depends on energy sector coordination that happens in a regulatory domain they do not control.
The Bottom Line for AI Compute Investors
For investors and operators evaluating Saudi AI compute infrastructure, ACWA Power is a critical due diligence subject. The viability of multi-gigawatt AI campus proposals in Saudi Arabia depends on ACWA Power’s ability to deliver committed power on schedule, at the specified tariff, with the reliability (uptime) that AI compute workloads require.
The company has a strong track record in Saudi Arabia on large-scale renewable energy projects. Its PIF alignment provides political cover and procurement priority. Its balance sheet (post-IPO) gives it access to capital markets for project finance. And its technical expertise in large-scale power project development is genuine.
The open questions are about pace and integration: can ACWA Power deliver power infrastructure fast enough to keep up with Saudi Arabia’s AI compute ambitions, which are themselves extraordinarily aggressive? And can it do so while meeting the reliability and carbon intensity requirements that international AI compute customers increasingly expect?
The answers to those questions will shape not just ACWA Power’s future, but the credibility of Saudi Arabia’s claim to be a global AI compute hub rather than a collection of ambitious announcements.