SCS Methodology — How We Score the Saudi Compute Landscape

The Sovereign Compute Score

The Sovereign Compute Score (SCS) is saudicompute.com’s proprietary 0-10 scoring framework for entities in the Saudi compute landscape. It is computed across seven weighted components, normalized to a 0-10 scale, and aggregated into a weighted average. Every entity profile, comparison page, ranking, and sector analysis surfaces the SCS as the primary numeric metric. Entities are mapped to status tiers — Top Tier, Strategic, Competitive, Emerging, Watch — based on their aggregate SCS.

The SCS is the analytical anchor of the platform. It exists for three reasons. First, the Saudi compute landscape is too multi-dimensional for any single metric (capex, capacity, GPU count) to capture entity position adequately. Second, the diversity of entity types — sovereign vehicles, operating companies, silicon suppliers, hyperscalers, regulators, advisory bodies — requires a framework that normalizes across entity classes. Third, the platform’s mission requires a transparent, reproducible, methodology-disclosed score that subscribers can audit and contest.

The Seven Components — Detailed Breakdown

1. Capacity (weight: 18%)

Capacity captures the physical compute footprint of an entity. It combines installed and announced data center capacity (measured in megawatts of IT load) with deployed and contracted accelerator count (measured in GPU-equivalents). The component is normalized so that 2,000 MW of IT load or 600,000 GPU-equivalents map to a 10/10 score; entities with smaller footprints score proportionally lower.

The blended construction reflects that the operationally meaningful metric for an AI-infrastructure entity is the combination of physical kilowatts and the accelerator density inside them. A campus with 1,000 MW of IT load but minimal accelerator deployment is operationally less consequential than a 500 MW campus filled to design density with frontier accelerators. The component aggregation captures both dimensions and weights them roughly equally.

Top scorers on the Capacity component include Humain (1,900 MW of disclosed campus footprint and roughly 600,000 contracted GPU-equivalents), DataVolt’s NEOM program (1,500 MW), Hexagon (480 MW), and the Stargate program where its compute footprint intersects Saudi-resident capacity. Hyperscaler entities score the Capacity component based on their disclosed Saudi-region IT load only; their global capacity is not included.

Edge cases: pure regulatory entities (SDAIA, MCIT, CST) and pure capital entities (PIF, Mubadala) without operational compute footprint score the Capacity component as Not Applicable. The aggregate SCS for those entities is computed across the remaining components, with weights renormalized.

2. Capital (weight: 16%)

Capital captures the financial muscle behind an entity. It is measured as committed capex in USD, normalized so that $100 billion maps to a 10/10 score. Capex is disclosed-and-board-approved spend, including operational expense allocations where they map to compute infrastructure; it does not include announcement-stage commitments that have not cleared internal approval.

Top scorers include PIF (where the relevant capex pool is the AI-attributable subset of the $930 billion AUM, which the methodology approximates conservatively), Mubadala-aligned vehicles ($280 billion AUM), the Stargate program ($500 billion announced), and Humain ($77 billion mandate). Silicon suppliers (NVIDIA, AMD, Qualcomm) score the Capital component based on their Saudi-attributable revenue and capex commitments, not their global capex.

The 16% weight reflects that capital is a necessary but not sufficient input to compute capability. Entities with very large capital pools but limited operational mandates (passive sovereign vehicles) score the Capital component highly but score lower on Capacity, Velocity, and Execution.

3. Velocity (weight: 12%)

Velocity captures execution momentum. It is computed from operational stage and launch recency. The stage classifications are Excluded, Evaluation, Planned, Construction, Operational, Shipping, with the latter two scoring highest. The recency adjustment rewards entities that achieved operational status from 2024 onward and penalizes entities whose operational status is pre-2022.

Top scorers include Humain (launched 2025, operational ramp on aggressive schedule), DataVolt (operational schedule mapped against power-purchase milestones), and the hyperscaler regions that have cut over to general availability since 2024. Excluded entities (those subject to BIS Entity List or AI Diffusion ineligibility) score the Velocity component at the bottom of the scale.

Edge cases: very mature operational entities (Saudi Telecom Company, Aramco) with long-running compute footprint can score moderately on Velocity even with pre-2024 operational dates if their AI-attributable capex and capacity is on a recent launch trajectory.

4. Sovereignty (weight: 13%)

Sovereignty captures alignment with the sovereign-AI thesis. Entities are classified into four ownership categories: state-owned (e.g., SDAIA, KACST), sovereign-funded (e.g., Humain, NEOM), majority-state-owned (e.g., Aramco, Saudi Electricity Company), and private/public (most foreign counterparties). State-owned and sovereign-funded entities score highest; private and public companies score lower.

The component is not a value judgment; it is a structural classification. Hyperscaler subsidiaries operating Saudi-resident regions score the Sovereignty component as private/public, even where they operate inside the Cloud SEZ; the sovereignty capture for those workloads is at the Cloud SEZ administrative layer, not at the operating-entity level.

Top scorers on the Sovereignty component are SDAIA, the National Information Center, Humain, NEOM, KAUST, KACST, KFUPM, and the broader sovereign-funded portfolio. Mid-tier scorers are the majority-state-owned operating entities (Aramco, SABIC, SEC, ACWA Power). Foreign hyperscalers and silicon suppliers score the Sovereignty component lower.

5. Silicon Access (weight: 16%)

Silicon Access captures access to frontier compute. The component is computed differently for three entity classes. For Saudi consumers (entities deploying compute), Blackwell-tier access scores 10, Hopper-tier scores 7.5, and entities without frontier silicon access score lower. For silicon suppliers, Saudi-export-active status scores 9.5 and restricted status scores 3. For hyperscalers, US-aligned entities operating Saudi-resident regions score 8 and other configurations score lower.

The Silicon Access component is the SCS dimension most directly exposed to BIS export-control posture. Any tightening of AI Diffusion Tier-2 access would propagate into the Silicon Access scores of every Saudi consumer in the universe; any expansion (e.g., movement toward Tier-1 status) would propagate upward. The 16% weight reflects that frontier silicon access is among the most consequential capability inputs.

6. Geopolitical Resilience (weight: 13%)

Geopolitical Resilience captures durability under shifting policy regimes. The component anchors on post-November 2025 US export-control alignment status. Saudi-domiciled US-aligned entities score highest. Excluded entities (e.g., Huawei in its Saudi-relevant configurations) score lowest. Mid-tier scorers are entities whose alignment posture is mixed or whose exposure to alignment shifts is structurally higher.

The component is the SCS dimension most exposed to the US-Saudi diplomatic relationship. The November 2025 reset is the current baseline; future resets would propagate into the component scores.

7. Execution (weight: 12%)

Execution captures the announced-versus-delivered ratio. It is a heuristic from operational stage and capacity presence. Operational entities with deployed capacity score highest; planned and evaluation-only entities score lower; entities with very large announcement-stage commitments and limited operational delivery score the Execution component lower than the headline announcements would imply.

Execution complements Velocity. Velocity captures recency and trajectory; Execution captures the absolute level of delivery against announced commitments.

Aggregation and Tier Mapping

The aggregate SCS is the weighted average of the seven component scores. Component scores that are unavailable for an entity (e.g., Capacity for a pure regulatory entity) are excluded; weights are renormalized over present components. The aggregate is mapped to one of five status tiers: Top Tier (8.5+), Strategic (7.0-8.4), Competitive (5.5-6.9), Emerging (3.5-5.4), Watch (under 3.5).

Update Cadence and Reproducibility

SCS scores are recomputed weekly based on the prevailing entity universe and underlying data. Status tier transitions are flagged automatically and surface in the Sovereign Compute Terminal. The computation is fully scripted and reproducible. The underlying entity universe is published at /players/. The component weights and thresholds are fixed in this methodology. The aggregate scores are deterministic given the entity inputs.

If you want to inspect or contest a score, the per-entity profile pages surface the seven component scores. The component score for any entity is recomputable given the input attributes (capacity, capex, ownership, stage, US alignment).

What the SCS Does Not Capture

The SCS is a structured capability index, not a complete description of an entity. It does not capture quality of execution (Humain may execute faster or slower than its score implies; SCS captures structural position, not operational pace). It does not capture strategic intent (entity goals are summarized in the descriptive layer, not the score). It does not capture market dynamics (regional competition, customer demand, pricing pressure). It does not capture risk-adjusted return (the score is a capability index, not an investment recommendation).

Use the SCS as the entry point. Read the entity profile, the comparisons, the deal flow, and the policy framework for the full picture.

Versioning

This is Version 1.0 of the SCS methodology, effective April 2026. Significant methodology changes will increment the version number and are documented in a changelog. Entity scores are recomputed on every methodology change so that historical comparability is preserved.

Worked Example: Humain

To illustrate the SCS computation in concrete terms, consider Humain. On the Capacity component, Humain’s roughly 1,900 MW of disclosed campus footprint and roughly 600,000 contracted GPU-equivalents map to a near-maximal score, in the 9.0+ range, against the 2,000 MW / 600,000 GPU normalization anchor. On the Capital component, the $77 billion mandate maps to a score in the 7.5 to 8.0 range against the $100 billion = 10/10 normalization. On the Velocity component, Humain’s 2025 launch and aggressive operational ramp earn a near-maximal score. On the Sovereignty component, Humain’s wholly-owned PIF status earns a maximal score. On the Silicon Access component, Humain’s Blackwell-tier access via the May 2025 NVIDIA framework earns a maximal score. On the Geopolitical Resilience component, Humain’s Saudi-domiciled US-aligned status post-November 2025 earns a maximal score. On the Execution component, Humain’s status as a 2025-launched entity with operational delivery still ramping earns a moderate-to-high score that will scale upward as the operational picture matures.

The weighted aggregate places Humain firmly in the Top Tier of the SCS framework, with the specific aggregate being the largest single SCS in the universe at the time of writing. The score is not surprising; Humain is the consolidated operating company of the Saudi compute thesis, and the SCS framework is engineered to surface exactly that kind of structural position.

Worked Example: SDAIA

SDAIA presents a different SCS profile because it is a regulatory and strategic body rather than an operating compute entity. The Capacity component is Not Applicable for SDAIA — it does not operate compute infrastructure directly — and is excluded from the aggregation, with the remaining six weights renormalized over present components. The Capital component reflects the AI-attributable subset of the broader Saudi government commitment, scoring in a moderate range. The Velocity component reflects the institutional pace of policy-and-strategy publication, which has been continuous since 2019. The Sovereignty component is at the maximal end of the scale: SDAIA is a state authority by definition. The Silicon Access component is Not Applicable for a regulatory body and is excluded. The Geopolitical Resilience component is at the high end given SDAIA’s central role in the institutional architecture. The Execution component reflects the institutional output cadence of decrees, strategies, and operational programs.

The SDAIA aggregate places it in the Strategic tier, reflecting its high-impact policy role even without operating compute infrastructure. The example illustrates how the SCS framework normalizes across entity classes by selectively excluding components and renormalizing weights, so that the aggregate is comparable across operating companies, capital providers, regulatory bodies, and silicon suppliers.

Edge Cases and Treatments

Several edge cases recur across the entity universe. Foreign hyperscalers (AWS, Azure, Google Cloud, OCI) are scored on their Saudi-resident posture rather than their global footprint; the global capacity, capex, and capability are not in scope. Silicon suppliers are scored on their Saudi-export-active status rather than their global revenue. Regulatory bodies score Capacity and Silicon Access as Not Applicable, with renormalization across remaining components. Entities subject to BIS Entity List or AI Diffusion ineligibility (e.g., Huawei in its Saudi-relevant configurations) score Velocity and Geopolitical Resilience at the bottom of the scale and do not score the Silicon Access component. Joint ventures (e.g., DataVolt at NEOM) are scored on the JV’s operational profile, not on the individual partner profiles, with cross-references to the parent entities documented at the entity-profile level.

Auditing and Contesting Scores

The SCS framework is open by design. Subscribers and public readers can audit any score by inspecting the per-entity profile page, which surfaces the seven component scores and the underlying input attributes. Where the input attributes differ from a reader’s understanding, the contestation pathway is to submit the dispute through the platform’s contact layer, attaching the source material that supports the alternative inputs. The platform evaluates the submission against the underlying source corpus and integrates corrections into the next refresh cycle where they meet the editorial standard.

The contestation pathway has been used in the platform’s operating history primarily for input-attribute corrections (e.g., a counterparty’s correction to a disclosed capacity number) rather than for component-weight or threshold disputes. The framework’s weights and thresholds are fixed at the methodology level; component-level disputes are routed to the methodology-versioning workstream rather than to per-entity score adjustments.

For deeper reading:

• Terminal — the analytical surface where the SCS is consumed
• About — platform mission and editorial standards
• Humain section — the highest-SCS entity in the universe
• Players index — the full entity universe at the published cadence