Open-weight inference, engineered at the site.
A structural discount to hyperscaler marginal cost —
anchored by contract, captured by operation.
A 1 MW IT load cluster draws power from generation that, on stranded sites, costs a fraction of market-rate data-centre electricity — and the gap compounds over every hour the facility operates. At site-level power below $0.03/kWh, the operating cost of a megawatt of compute is a fraction of what a market-rate competitor pays, and the gap does not close with scale, silicon generations, or model releases. Most of the industry optimises for tokens-per-watt because watts are scarce. Cathedral manufactures watt-abundance from stranded energy, and competes on tokens-per-dollar instead. Real. Structural. Permanent.
Cheap power alone does not guarantee a competitive token price. The throughput lever is not exotic silicon; it is refurbished A100 80GB GPUs acquired from hyperscaler refresh cycles — depreciated-but-capable data-centre silicon moving off cloud balance sheets at a meaningful discount to new-GPU cost — run under vLLM, TensorRT-LLM, and SGLang with quantisation and continuous batching. That stack serves the mid-size open-weight band (roughly 8B–120B) at a cost per token that clears the economics. Refurbished fleets from refresh cycles and neocloud wind-downs are a structural supply channel, not a one-off opportunity — and the channel deepens with every new GPU generation the industry adopts.
The first two variables produce a cost floor. The third determines whether that floor is captured as revenue. A pure merchant posture at commodity rates underperforms at site scale; a pure capacity commitment to a single counterparty is bankable but forecloses the upside. The operative structure is hybrid: a meaningful share of each site committed to an anchor counterparty — sovereign or frontier-lab — on take-or-pay terms for dedicated open-weight serving, with the balance run as merchant open-weight token supply at clearing rates. The anchor makes the site bankable at commissioning; the merchant overlay captures the spread between Cathedral's cost floor and the clearing price. This is the structure the work converges on.
The served model class is the premium open-weight band — Llama 70B-class, Qwen, gpt-oss, and the leading open families — closing the quality gap to proprietary frontier APIs for the majority of agentic execution workloads. At target throughput and site-level power cost, the hybrid structure produces a structural discount to hyperscaler marginal cost for equivalent workloads. The captured return accrues to Cathedral as operator.
The AI inference market is a routing stack. Per-call model selection is already standard practice: high-stakes orchestration runs on proprietary frontier models; cost-sensitive agentic sub-tasks, batch classification, document synthesis, and retrieval route to open-weight models where quality is sufficient and cost is the constraint.
Cathedral's commercial reference point is the open-weight inference clouds serving that cost-routed layer — Together AI, Fireworks, Replicate class — pricing open-weight models on market-rate electricity and new-GPU fleets. Cathedral's sites target structurally lower power costs on fleets acquired meaningfully below new-silicon replacement cost — a compounded cost advantage. Applied to the open-weight executor tier, that advantage produces a token cost floor at the site level.
Stranded power + refurbished GPUs + open-weight models = a token cost floor that is structural, not cyclical. That is the asset.
The capex tells the same story from the other direction. New AI data-centre capacity is being built at tens of millions of dollars per megawatt. Cathedral's entire site — power generation, shell, and compute included — costs roughly what the industry pays for the empty building.
The proprietary frontier APIs priced many multiples above Cathedral's cost are not the competitive set. They are the demand driver — the pressure that forces frontier labs and enterprise AI builders to route cost-tolerant workloads to cheaper infrastructure, creating and sustaining the market Cathedral's sites are built to serve.
The electricity advantage does not erode with chip generations or model releases. A new chip available to Cathedral is equally available to any competitor — and the refurbished-fleet supply channel Cathedral is built on grows as the industry refreshes into each new generation. The stranded-power condition — local demand absent, transmission to market uneconomical — is not solved by a model release or a faster chip. It is a function of geography, physics, and the economics of power transmission. It accrues to whoever holds the site.
The first chapter of AI was about training. Enormous GPU clusters, hundreds of megawatts, billions of dollars, months of computation — all to produce a model weight file. That chapter is closing. The frontier models now exist. They are trained. Many are open. The capital and the engineering attention of the global AI industry is rotating, decisively, toward inference — the act of running those models at production scale, billions of times per day.
By 2025, inference already accounted for the large majority of all AI compute cycles. By 2026, inference spending is projected to represent two-thirds of total AI infrastructure investment. Within that aggregate, the fastest-growing slice by token volume is the agentic executor tier — cost-sensitive, latency-tolerant, open-weight-model-dominant — which already consumes the majority of tokens routed through multi-agent systems. The inference market is segmenting, and Cathedral is built for the largest, fastest-growing, most price-sensitive segment.
Massive GPU clusters building foundation models. Dominated by the newest Nvidia silicon. Requires ultra-high bandwidth interconnects, precision-critical arithmetic, enormous power budgets. The largest frontier labs continue to train, but open-weight releases have fundamentally altered the economics — frontier-class capability is now freely deployable by anyone. The capital intensity of training does not reduce the inference opportunity; it amplifies it. Every new open-weight release becomes a new revenue-generating workload for Cathedral.
Real-time chat, voice assistants, coding copilots — applications where a human is waiting for a response and sub-200ms time-to-first-token matters. Latency is the product. Expensive real estate, premium connectivity, proximity to population centres. A competitive market with established players. Requires proximity to end users. Not Cathedral's primary target.
Autonomous agents, multi-step reasoning, batch synthesis, background processing — workloads where no human waits for the next token, and the primary constraint is cost per million tokens at sustained throughput. Crucially, this is not a purely open-weight market. Frontier labs themselves route a growing share of their cost-sensitive, latency-tolerant workloads through open-weight models on cheaper infrastructure, blending proprietary and open-weight inference across a single agentic pipeline depending on the task. The orchestrator plans; the executor serves. Cathedral's infrastructure is built for this executor tier of the stack — wherever open-weight models are the right tool and token price is the governing constraint. This segment is growing faster than consumer-facing inference by an order of magnitude and is structurally uncaptured by existing high-cost infrastructure.
Autonomous agent hierarchies are replacing the human-at-keyboard model of AI interaction. An orchestrator agent decomposes a goal and delegates to sub-agents — each of which calls specialist models for reasoning, coding, retrieval, verification, and summarisation. A single user intent can propagate to hundreds or thousands of model invocations before the goal is resolved. Agentic pipelines are not monolithic: high-stakes orchestration may use a proprietary frontier model, but the bulk of sub-agent calls — document parsing, code generation, classification, summarisation, retrieval synthesis — are well-served by open-weight models running at a fraction of the cost. The intelligence hierarchy in an agentic system is a cost-routing problem as much as a capability problem.
IDC projects a thousandfold increase in token loads by 2027. Enterprise AI inference bills are growing at triple-digit rates quarter-on-quarter. The pressure to find cheaper inference is the defining operational constraint of the agentic era — and it compounds as pipeline complexity grows. Ultra-low latency is not a requirement for the executor tier, which is precisely what makes stranded-power geography viable. Cathedral's target jurisdictions all sit within round-trip latency tolerances of the EU, US, and Middle East demand centres they serve. A sub-agent synthesising a research brief, processing a document batch, or running a background classification task is indifferent to whether the model responds in 40ms or 140ms. Token cost at sustained throughput is the governing variable.
The agentic executor tier is one demand curve. The second is sovereign. A nation that wants AI under its own jurisdiction faces a constraint no budget removes.
The leading closed models exist only as a service from foreign data centres, so every prompt, document, and output leaves the country by design. Genuine sovereignty therefore depends on open-weight models served in-country — the only configuration that keeps data, workload, and control inside national borders. This is the same infrastructure Cathedral builds for the cost-routed executor tier, pointed at a different demand: a national platform that needs frontier-class open-weight capability hosted on its own soil, under its own data law, operated as a finished service rather than a rack of hardware.
The training era was built on new GPUs. The inference era is being subsidised by them. Every hyperscaler refresh into the newest silicon creates a structural supply of depreciated A100 fleets moving off cloud balance sheets at a meaningful discount to new-GPU cost basis. These GPUs remain highly capable for the mid-size open-weight workloads Cathedral is built to serve; the inefficiency was in pricing them at cloud retail, not in the silicon itself.
Cathedral acquires this fleet through hyperscaler refresh channels, neocloud wind-downs, end-of-lease returns, and direct broker relationships — then runs it under vLLM, TensorRT-LLM, and SGLang with quantisation and continuous batching to clear the open-weight serving economics. No inference-native ASIC is required. The combination of refurbished silicon, mature open-source serving stacks, and site-level power below $0.03/kWh is what produces the structural cost floor.
The fleet is pure A100 80GB. The 80GB memory envelope is the critical spec — it is what supports modern open-weight serving across the mid-size band, and what makes the refurbished supply channel meaningful rather than incremental. An 8× node serves the 8B–120B class cleanly at INT8 and 4-bit weight quantisation; the workloads Cathedral targets sit squarely inside that envelope. These are the units the commercial model is built on.
The throughput advantage does not come from new silicon; it comes from the open-source serving stack that has matured through 2024–2026. Continuous batching materially lifts effective throughput over batch-one serving. Quantisation preserves quality on the target open-weight models while expanding tokens-per-dollar. Prefill caching lowers the marginal cost of long-context agentic workloads. This stack is as much a part of the cost moat as the power contract — and Cathedral's position depends on the open-source ecosystem keeping mid-size open weights first-class on this class of silicon, which it has every incentive to do.
Why not the newest GPUs: the cost basis is wrong for this workload class. The newest silicon is built for frontier training and inference on frontier-scale models; the executor tier Cathedral serves does not need it and cannot amortise it. The economics favour volume of capable-and-cheap silicon over the performance ceiling of new-and-expensive silicon.
Why not inference-native ASICs: the software ecosystem is not yet mature at Cathedral's scale. The open-source serving stacks are tightly bound to Nvidia CUDA; porting to alternative silicon is an engineering tax that the cost-per-token advantage does not yet cover at realistic deployment sizes. The ASIC case will clear as those ecosystems mature, and Cathedral's PropCo/OpCo structure accommodates hardware-class evolution at the compute-overlay level without re-underwriting the site. For now, the refurbished-GPU arbitrage is the decisive variable.
Cathedral is a vertically integrated power, real-estate, and inference operator. Each site is a physical asset — long-tenure power contract, purpose-built modular shell, refurbished A100 compute fleet running open-weight models — held in a per-site SPV under a PropCo / OpCo structure, with capital matched layer-by-layer to the useful life of each component. Power and shell are infrastructure-grade and sit in PropCo; the compute fleet and commercial contracts sit in OpCo, where Cathedral operates the site and captures the operating margin. On flared-gas sites, the build also carries a methane-mitigation overlay — on-site combustion of gas that would otherwise be vented or flared.
Each site sits in a dedicated per-site SPV. Physical assets — power contracts, land, shell, cooling — are ring-fenced in PropCo. Operating activity — the GPU fleet, anchor and merchant contracts, the inference serving stack, commercial relationships — sits in OpCo. Failure at any single site is non-recourse to the holdco. Capital stacks and commercial structures can vary site by site without disturbing the portfolio.
The foundation asset and the moat. Long-dated offtake on associated gas, stranded hydro, or structurally underpriced grid. Contracted duration defines the outer horizon of every downstream decision. Financed against contract quality with infrastructure-fund and insurance-company debt. Capital matched to the useful life of the power asset, not to chip generations. This is the reason Cathedral exists.
Modular, rapidly-deployable, purpose-built for air-cooled refurbished A100 SXM fleets at high rack density. Standardised design compresses lead time from site origination to first MW energised. Rebuild cadence tied to cooling and power topology, not GPU generation. Anchor take-or-pay contracts unlock CRE-style debt at institutional coupons once site-level credit is established.
Cathedral as operator: refurbished A100 80GB fleet acquired from hyperscaler refresh cycles, running open-weight models under vLLM, TensorRT-LLM, and SGLang. Revenue split between an anchor take-or-pay slice and a merchant token-serving overlay — the hybrid posture detailed in Section 009. Selection of the specific hybrid mix is made per site against anchor credit, merchant demand absorption, and risk-adjusted return. The operating margin belongs to Cathedral.
Cathedral captures this advantage as operator — anchor + merchant, per site, per contract.
Cathedral is led by a founder with a background in project finance, energy, and complex cross-border commercial law, supported by a bench assembled for exactly this problem — stranded-energy infrastructure at the intersection of power, capital, and inference. Named team and advisors are available on request. The capabilities on the bench:
Upstream energy origination across the Gulf, with direct relationships to national oil companies and sovereign-adjacent institutional capital. The bridge between a stranded-gas asset and a financeable site.
A repeat-exit software founder and operator, with deep experience in enterprise platforms and data-residency — the in-country inference angle that decides sovereign engagements (the credibility behind Section 004).
CFO-grade structuring and a track record in emerging-market infrastructure joint ventures, including African energy and port assets. The discipline that turns a thesis into a closeable structure.
Published expertise in methane mitigation through stranded-energy compute, with a network across climate and development-finance capital — the pools aligned with the flare-gas overlay.
How sovereign and state institutions actually behave over multi-year horizons — succession, durability, follow-through — and the framing that makes the thesis legible to national-security-aligned capital.
Serving-stack engineering for open-weight inference at scale — the throughput layer that turns refurbished silicon and cheap power into a defensible token cost. Seat being finalised.
Cathedral deploys wherever electricity is structurally underpriced — where generation capacity exists but local demand is insufficient to absorb it, transmission to higher-demand markets is uneconomical, or associated gas is flared at the wellhead for lack of monetisation. That may mean grid power in a jurisdiction with structurally low industrial tariffs, associated gas at a remote production site with no pipeline access, or a purpose-built economic zone with allocated industrial capacity. Across the Gulf, that condition increasingly coincides with sovereign-scale AI demand and capital — national compute built on abundant low-cost energy. Grid-connected or off-grid, the selection criteria are identical: kWh price and supply reliability. Flaring data shown is third-party (World Bank GGFR).
A national digital-infrastructure agenda and an in-force data-residency framework create sovereign demand for in-country inference. Two distinct site channels then operate in parallel for power. The first is purpose-built digital-infrastructure capacity inside Oman's free zones — long-dated corporate tax holidays, full foreign ownership, and grid power at low industrial tariffs under cost-reflective pricing for large users. The second channel is on-pad and near-pad deployment alongside the major upstream operators active in Oman's interior production blocks, whose remote sites generate associated gas with no economic pipeline route to market. Oman flared on the order of 1.0–1.2 bcm of associated gas across 111 flare sites in 2022 (World Bank GGFR) — equivalent to several hundred MW of continuous power potential at standard generator efficiency, with further upside from gas currently reinjected at non-EOR-critical fields. National Zero Routine Flaring commitments by 2030 create active commercial pull for offtakers. These are bilaterally negotiable stranded-power opportunities at gas-cost economics, sized to the field. Subsea connectivity from the coast; USD-pegged currency since 1986 — zero FX risk.
Saudi Arabia is building sovereign AI compute at national scale, backed by sovereign capital and an explicit state strategy — a demand pull few jurisdictions can match, paired with the energy to serve it. Power is delivered through purpose-built economic zones and energy parks at low industrial tariffs under full foreign-ownership regimes. Beyond the national gas grid, remote and unconventional gas — including the largest unconventional gas development in the region — sits far from existing demand at gas-cost economics. In-country sovereign demand, allocated economic-zone power, and energy abundance define the opportunity. Riyal pegged to USD — zero FX risk.
The UAE hosts the region's most developed sovereign-AI ecosystem — national compute champions, sovereign technology capital, and a US-aligned posture on advanced compute. Power is abundant and low-cost, drawn from associated and non-associated gas, and delivered through mature industrial free zones with full foreign ownership, strong digital infrastructure, and dense subsea connectivity. The jurisdiction pairs deep sovereign-AI demand with energy abundance and the regulatory and connectivity baseline for in-country inference. Dirham pegged to USD — zero FX risk.
Qatar sits atop the North Field — the world's largest non-associated gas field — and is the leading global LNG exporter, giving it unmatched low-cost gas. Sovereign capital and a growing national digital-infrastructure agenda are drawing data-centre and compute investment into purpose-built economic and free zones with full foreign ownership. Gas at gas-cost economics, sovereign backing, and expanding connectivity make it a natural behind-the-meter and economic-zone compute jurisdiction. Riyal pegged to USD — zero FX risk.
Algeria holds roughly 25 GW of installed gas-fired generation against a peak demand near 19 GW — a nominal surplus, with a meaningful share realistically available as industrial offtake outside peak periods. Separately, Algeria flared on the order of 8.6 bcm of associated gas in 2022 (World Bank) — equivalent to several thousand MW of continuous power potential across remote Saharan flare sites. The national oil company is under hard regulatory pressure to eliminate routine flaring by 2030. Domestic industrial electricity is heavily subsidised. Mediterranean subsea cables provide connectivity.
Cathedral is an operator at every site. The company owns the hardware, runs the compute, and captures the operating margin. The commercial question at each site is how to combine a bankable anchor counterparty with merchant token supply to produce returns that clear both a debt test and an equity-return test.
The operative structure, validated in the site-level financial model, is hybrid anchor + merchant: a meaningful share of each site's capacity committed to an anchor counterparty on take-or-pay terms, with the balance run as merchant open-weight token supply. The anchor makes the site bankable at commissioning; the merchant overlay captures the spread between Cathedral's cost floor and the open-weight clearing price. Two alternative postures exist as fallback and as evolution; neither is the target.
The served model class across all postures is the premium open-weight band — Llama 70B-class, Qwen, gpt-oss, and the leading open families — where quality meets cost and proprietary-API pricing creates structural routing pressure from the agentic executor tier. The commodity floor is not the strategy.