The Saudi DC pipeline by megawatt

Saudi data center MW under construction is the cleanest leading indicator of compute capacity coming online through 2030. Pre-2024, Saudi DC capacity was concentrated in conventional colocation in the 50-150 MW range across roughly 22 active facilities serving telecom, banking and enterprise workloads. Post-2024, the pipeline shifted dramatically as PIF, Humain and the broader sovereign AI program activated capital and procurement: Hexagon (480 MW, world’s largest sovereign DC), Humain campuses (multiple facilities under construction at 100-200 MW per site across Riyadh and Dammam), DataVolt-NEOM (1.5 GW net-zero AI factory at Oxagon), the xAI joint-venture site (500 MW alongside Humain), and the Center3 expansion adding 1 GW of cumulative capacity through new builds and brownfield additions.

The cumulative pipeline reaches 6.6 GW across announced projects if all deliver on stated timelines — roughly 4x the operational base at end-2025. The ranking below filters announcement-stage from genuinely-under-construction projects to surface which MW commitments will actually convert to operational capacity in the 2026-2030 window. It is one thing to announce a gigawatt; it is another to source the EPC contractor, the substation transformer, the chiller equipment and the rack-level integration on a coherent timeline.

Reading the top of the pipeline

Humain occupies the top position because of cumulative committed capacity across multiple sites rather than a single mega-facility. Humain’s published 2030 target is 1.9 GW of operational capacity across the Riyadh, Dammam and adjacent campus footprint. The phasing puts the first major operational tranches online during 2026, scaling through 2028 toward the full target. Humain’s facilities are designed to host frontier-tier accelerators (GB300 and successors) at liquid-cooled rack densities matching hyperscaler-grade specifications.

DataVolt holds the second slot via the NEOM-Oxagon 1.5 GW commitment. The DataVolt facility is differentiated by its renewable-first design — solar, wind and battery storage feeding a net-zero AI factory through dedicated PPAs. The 2028 operational target for the first major tranche is the structural milestone for global net-zero hyperscale compute. If DataVolt delivers, Saudi Arabia sets the precedent that AI infrastructure at gigawatt scale can be built renewable-first rather than retrofit; if it slips, the renewable-AI thesis takes a credibility hit not just for Saudi Arabia but globally.

Center3 (the stc-affiliated DC operator) holds the third slot via cumulative capacity additions across multiple Saudi cities. Center3 differs from Humain in serving the multi-tenant cloud and enterprise market rather than the sovereign AI workload directly. Hyperscaler regions (AWS, Google Cloud, Oracle) frequently lease capacity from Center3 for their Saudi presence. The 1 GW expansion target consolidates Center3 as the canonical neutral-host operator in the Kingdom.

xAI’s 500 MW joint venture at the Humain campus is a strategic entry rather than an operational entry — it represents the first frontier-AI-lab anchor tenant outside the US and China hosting capacity at this scale. The xAI commitment came alongside the broader November 2025 framework and is timed to deliver during 2026-2027 as Humain’s facilities go operational.

SDAIA’s Hexagon facility (480 MW, Riyadh region, supporting the National Data Lake and government AI workloads) anchors the sovereign-government segment of the ranking. Hexagon is structurally consequential beyond its MW size because it is the substrate for SDAIA’s AI factory partnership with NVIDIA, the deployment site for sovereign LLM workloads (Allam), and the integration point for cross-ministry analytical capability under the Year of AI 2026 framework.

How the pipeline gets built

Constructing data center capacity in Saudi Arabia at gigawatt scale is a multi-workstream coordination problem. Site acquisition runs through Modon (industrial cities) for most Riyadh and Dammam sites, NEOM authority for NEOM-region sites, and PIF-affiliated land arrangements for selected positions. Power supply is contracted with Saudi Electricity Company for grid-connected facilities, ACWA Power for renewable-PPA-backed sites and direct Aramco gas supply for industrial-zone facilities including Dammam. Connectivity flows through stc, Mobily and Zain fiber backbones connected to subsea cable landings via the SmartHub Saudi Arabia program.

Construction itself runs through Saudi EPC firms (Al Khodari, Almabani, Saudi Binladin Group at scale) frequently partnered with international firms for specialized data center engineering (Bechtel, Jacobs, Aecom). Cooling specialty firms (Vertiv, Schneider Electric, Stulz) supply the liquid-cooling and chilled-water infrastructure. The integration layer (rack-level systems, networking, security) is increasingly handled through OEM partnerships (Lenovo, Dell, HPE) coordinated by Humain’s internal engineering team or by Cisco under the Cisco-AMD-Humain JV.

Cycle time from greenfield to operational Tier IV-equivalent capacity runs 24-36 months for hyperscale projects under Saudi conditions. The major projects (Hexagon, Humain campuses, NEOM-DataVolt) operate against tighter timelines through executive prioritization and concentrated resources. Even with prioritization, slippage of 3-9 months on individual milestones is common; cumulative slippage on the 6.6 GW pipeline is likely to push some announced 2028 capacity into 2029.

What the ranking captures and misses

The ranking captures publicly disclosed Saudi data center projects with credible MW figures and operational targets. The ranking is biased toward AI-relevant capacity (Tier IV-equivalent, liquid-cooled, frontier-accelerator-host-able) rather than legacy colocation or telecom-grade facilities. A 50 MW telecom-grade DC running enterprise workloads is operationally significant but excluded from this ranking; a 50 MW AI-grade DC hosting frontier accelerators is included.

The ranking misses undisclosed enterprise-internal capacity inside Aramco, SABIC, Saudi banks and government agencies. The aggregated undisclosed capacity is meaningful (probably several hundred MW in cumulative enterprise-internal deployment) but its absence from the public ranking does not change the structural picture. The ranking also undercounts the smaller multi-tenant facilities operated by regional operators (BCX, eHosting DataFort regional positions) which serve enterprise workloads at smaller scale.

What changes the ranking

Three forcing functions reshape the ranking through the next 12 months. First, Hexagon’s full operational status — currently in ramp through 2026, fully operational expected during the year, which moves SDAIA from announcement-stage to top-of-ranking. Second, the Humain campus tranche-by-tranche operational milestones — each tranche moving online shifts cumulative operational MW by 50-200 MW per event. Third, the DataVolt-NEOM first operational tranche — slipped from initial 2027 targets into the 2028 window, but if the first 200-400 MW comes online during the year it materially changes the renewable-AI ranking and the broader DC pipeline picture.

Outside the top tier, additional capacity comes from hyperscaler region launches (Microsoft Saudi region Q4 2026, Oracle Riyadh expansion, Google Cloud Dammam ramp), the broader Center3 expansion, and emerging projects from new entrants (Gulf Data Hub via the KKR investment, regional integrators like Mobily-affiliated entities scaling DC capacity to capture spillover demand from the Humain ecosystem).

The methodology disclosure

The ranking weights MW capacity by four factors: operational status (operational > construction > announced), AI-grade specification (liquid-cooled and frontier-accelerator-capable carries higher weight than legacy spec), strategic anchor tenants (sovereign-aligned > commercial multi-tenant > pure colo), and durability of capital backing (PIF / hyperscaler-anchored > regional-developer-led). The result is a composite ranking rather than a pure MW-size league table.

Two recurring data-quality issues affect the methodology. Announced MW figures often span phased delivery and the headline number conflates fully-built capacity with multi-year buildout. Where possible the ranking distinguishes between operational and committed-but-not-yet-built MW. Hyperscaler regions’ MW figures are not always publicly disclosed; the ranking uses construction permits and equipment-import data where available alongside announced figures.

The energy-supply binding constraint

Every gigawatt of announced Saudi DC capacity assumes coordinated power delivery on the same timeline. Slippage in power infrastructure caps the achievable compute buildout regardless of GPU procurement velocity. Saudi Arabia’s structural electricity-cost advantage (industrial tariffs in the $20-50/MWh range versus $80-150/MWh in US data center hubs) is meaningless if grid connection delivery slips. The ACWA Power renewable buildout is the long-cycle answer; Saudi Electricity Company grid expansion is the short-cycle answer; direct Aramco gas supply for industrial-zone facilities provides the bridge.

The 2030 cumulative power demand for the announced 6.6 GW pipeline is roughly 50-60 TWh annually at typical AI workload utilization rates. That is non-trivial fraction of total Saudi electricity demand and requires coordinated grid planning that is publicly visible in Saudi Electricity Company’s capex disclosures and ACWA Power’s project pipeline. Watch the alignment of announced DC capacity with announced power capacity additions — material divergence between the two would signal that the pipeline is over-scheduled relative to the power supply and that DC operational milestones will slip.

The geographic clustering pattern

Saudi DC capacity clusters geographically around Riyadh, Dammam and NEOM-Oxagon for distinct reasons. Riyadh hosts the bulk of government-aligned and sovereign workloads (Hexagon, the SDAIA infrastructure, Humain Riyadh campus) because of proximity to ministerial decision-making and to the SDAIA / Humain executive teams. Dammam hosts industrial-zone capacity (Humain Dammam campus, Google Cloud’s $10B Dammam hub) because of access to Aramco gas supply, proximity to the petrochemical industrial cluster and connectivity to subsea cable landings. NEOM-Oxagon hosts the renewable-first capacity (DataVolt 1.5 GW) because of dedicated renewable buildout and the broader NEOM strategic positioning.

Secondary positions (Jeddah, smaller regional sites) host enterprise and telecom capacity at smaller scale. The geographic concentration is rational from an infrastructure-coordination perspective but produces concentration risk for Saudi compute resilience — a regional disruption affecting Riyadh or Dammam would impact a disproportionate share of national capacity. Watch the geographic diversification through the decade as additional sites at smaller scale absorb spillover demand.

The pipeline-versus-operational gap

The single most important analytical distinction in reading this ranking is between announced pipeline and actually-operational capacity. Announced 6.6 GW is a ceiling on 2030 operational capacity. Realistic operational capacity at end-2030 is likely in the 4-5 GW range after typical construction slippage, capital reallocation and pipeline rationalization. A 25-35% slippage between announced and operational is the typical pattern for hyperscale DC pipelines globally and there is no strong reason to expect Saudi Arabia to outperform that benchmark.

The slippage matters because downstream commitments — GPU procurement, hyperscaler region launches, sovereign AI program milestones — assume the announced pipeline. If actual operational capacity at end-2027 is materially below the announced trajectory, downstream commitments must be re-sequenced. The leading indicator is permit-issuance and EPC-contract-signing cadence; the lagging indicator is actual MW-online disclosed in operational reporting. The 2026 mid-year update will be the cleanest read on whether the pipeline is tracking or slipping.

Anchor tenants and the multi-tenant question

Saudi DC capacity is increasingly differentiated by its anchor-tenant structure. Hexagon is a single-tenant sovereign facility supporting SDAIA workloads exclusively. Humain campuses serve a hybrid model: the sovereign workload tier alongside commercial-fleet tenants accessing capacity through Humain’s commercial offerings. Center3 is a multi-tenant neutral host serving hyperscalers and enterprise customers. DataVolt-NEOM positions to serve both sovereign and commercial workloads with the renewable differentiation as the value proposition.

The anchor-tenant structure determines who controls the operational decisions, who absorbs capex risk, and who captures the operating margin. Single-tenant sovereign facilities concentrate decision authority and align long-term capital with strategic objectives but lose the multi-tenant economics. Multi-tenant facilities optimize utilization across diverse workloads but expose operators to commercial demand cyclicality. Hybrid facilities like Humain campuses attempt to capture both benefits but require more sophisticated capacity allocation across tenant tiers.

The international comparison

Saudi Arabia’s announced 6.6 GW DC pipeline through 2030 puts the Kingdom in the upper tier of sovereign DC programs globally. The US hyperscaler aggregate adds tens of gigawatts annually and remains the dominant bloc. The UAE program adds comparable capacity to Saudi Arabia at slightly different timing. China’s domestic pipeline adds tens of gigawatts but the AI-grade fraction is smaller given export-control constraints on the latest accelerators. India’s emerging program is smaller than Saudi Arabia’s by a factor of several. European national programs are individually smaller though the EU aggregate is comparable to Saudi Arabia at the GW level.

Saudi Arabia’s distinguishing characteristic is the concentration of capacity in a small number of mega-facilities (Hexagon at 480 MW, DataVolt at 1.5 GW, the Humain campus aggregate above 1.5 GW) versus the more distributed pattern in US and European programs. The concentration produces operational efficiency but increases the consequence of any single facility’s slippage.

For the GPU-cluster side of the picture, see the MENA GPU cluster ranking. For the broader regional capital allocation, see the MENA data center investments ranking. For the renewable-specific subset, see the Saudi renewable AI projects ranking.

For deeper reading: