
Rivers through the steppe
Patagonian rivers such as the Río Chubut cut through layered cliffs and dry grassland: the typical character of the region's high plateaus.

Confidential · Planning estimates
Sovereign AI Infrastructure · Patagonia Green Data Center
Own land · Own power · Defensible moat
El Calafate, Santa Cruz, Argentina · 30,000 ha · Sovereign · Renewable · Independent
The land today
The site lies near El Calafate in the province of Santa Cruz: open steppe, steady westerly winds, cold dry air and vast space. Today it is estancia land. It is exactly this landscape that makes a sovereign, 100%-renewable AI campus possible.

Patagonian rivers such as the Río Chubut cut through layered cliffs and dry grassland: the typical character of the region's high plateaus.

Steady westerly winds sweep the open plateau: the basis for 18 MW of wind in Phase 1, with a capacity factor of ~45%.

Ice fields and glacier lakes define the climate: 7–12 °C year-round, the basis for free ambient cooling.

Own land, today an estancia: room for a 50 MW IT campus, 140 MW wind + solar, 4,000 nodes and an own substation. Dual use: estancia operations continue.

Before · Future vision
Vision: the campus from the air. Data halls in modular 5 MW blocks, solar farm, wind turbines, battery storage and own substation, sized to own demand.
Drag the slider: the same view today and after full build-out. Data halls in modular 5 MW blocks, wind turbines, solar field, battery storage and an own 132 kV substation, sized to our own demand, with no export.


↔ Drag to compare

The problem
$40B+ annual spend on AI infrastructure, yet regulated buyers are locked out of the best models.
$12B
annual AI budget · <5% deployed safely
Cannot send customer data to OpenAI, Anthropic or any Chinese API. Cannot run AI on core workloads. Competitive disadvantage grows monthly.
$8B
annual AI budget · compliance paralysis
Patient data regulated under HIPAA, GDPR and local law. Even running a benchmark on real data is illegal without verified isolation.
$6B
annual AI budget · mostly stalled
Sovereign data requirements. Foreign jurisdiction = disqualification. Existing private cloud is expensive and not AI-native.
$14B+
annual AI budget · fastest-growing
Data sovereignty laws tightening fast. Need verified proof, not contractual promises, that data never leaves the jurisdiction.
The gap: open-weight models (DeepSeek, GLM) are 10–30× cheaper than GPT-4, but no one offers cryptographically verified private serving at enterprise scale on own infrastructure.
The solution
We run the most capable open-weight AI models inside cryptographic envelopes, on our own green infrastructure, in your jurisdiction.
GLM-5.2 & DeepSeek V4 inside hardware TEEs. Every inference cryptographically attested. Customer data never leaves the enclave: provably, not contractually.
→ Sells to regulated buyers others cannot touch
Sealed-box benchmark: prove an open model matches GPT-4 on your real workload, without data leaving the jurisdiction. The only way regulated buyers can legally test.
→ Sales wedge: turns “we'd love to but can't” into signed deals
Dedicated instance per enterprise. Required jurisdiction. Full SLA. Migration tooling. The rented-cloud alternative costs 4–5× more per node.
→ $40–90K/month · 75–85% gross margin
Pricing: $40–90K/node/month sovereign · $5–15K/month confidential serving · $2–8K per parity eval · 75–85% gross margin on own infrastructure vs. 45–65% rented
The Bloovian moat
Each layer is defensible alone; together they are near-impossible to replicate in under 24 months at our cost structure.
Asset
Zero lease, zero eviction risk. Room for a 50 MW IT campus, 140 MW wind + solar, 4,000 nodes and an own substation. Competitors pay $15–22M/yr in DC leases per 10 MW.
Cost
Phase 1: 28 MW wind + solar + 50 MWh BESS sized to 10 MW IT; scales to 50 MW IT. Own 132 kV substation, 100% own use. Structural, not a temporary discount.
OPEX
7–12 °C year-round. PUE < 1.15 without liquid cooling. Competitors spend 20–35% of power on HVAC; our cooling cost is effectively zero.
Regulatory
Santa Cruz, Argentina. No US/EU extraterritorial reach, no CLOUD Act, no forced disclosure. The credible LATAM-sovereign option; valid for EU/UAE data-residency buyers.
Technical
H200 Confidential Compute + Intel TDX / AMD SEV-SNP + our per-request attestation chain. 12–18 months to build right. Others give contracts; we give mathematical proof.
Timing
No comparable sovereign AI DC in South America. SENASA + CAMMESA approvals, grid study, provincial permits: 12–18 months of lead time we already own.
Copycat cost + timeline: land ($3M+) + permits (18 mo.) + 28 MW energy park ($47–63M) + 10 MW DC ($70–90M) + TEE engineering (12 mo.) = $130–160M (excl. GPUs) and 24–30 months before the first node is online.

Energy plant
Phase 1 energy CAPEX $47–63M · ~86 GWh/yr = DC demand (10 MW IT × PUE 1.15) · no export, no surplus sale · ×5 to 50 MW IT
Phase 1: 3 × 6 MW wind turbines and 10 MW solar PV on own land.
Solar PV
10 MW
$6–8M
~17,000 panels · ~15 GWh/yr · ~20 ha own land · 9–12 months build
Wind
18 MW
$23–29M
3 × 6 MW turbines · ~71 GWh/yr · CF ~45% · 15–18 months build
BESS
50 MWh
$10–14M
LFP · 4 h at 12.5 MW · firms wind/solar to the 24/7 DC load
Infrastructure
132 kV
$8–12M
Substation sized for 50 MW · grid backup tie · SCADA · EPC

| Phase | Energy plant | DC nodes | Energy CAPEX | Node OPEX/mo | Own use / yr |
|---|---|---|---|---|---|
| Phase 1 | 28 MW + 50 MWh | 800 (10 MW IT) | $47–63M | $3–5K | ~86 GWh |
| Phase 2 | 56 MW + 100 MWh | 1,600 (20 MW IT) | $87–114M cumul. | $3–5K | ~171 GWh |
| Phase 3 | 98 MW + 175 MWh | 2,800 (35 MW IT) | $146–190M cumul. | $3–5K | ~300 GWh |
| Full build | 140 MW + 250 MWh | 4,000 (50 MW IT) | $205–266M cumul. | $3–5K | ~428 GWh |
Own energy is the margin engine behind 75–85% gross margin: it makes the product possible, it is not the product.
| Cost driver | Own green DC (Bloovian) | Rented cloud (AWS/Azure) | Rented GPU (CoreWeave) |
|---|---|---|---|
| Power cost | $0.045–0.06/kWh | $0.10–0.15/kWh | $0.10–0.18/kWh |
| Cooling overhead | ~0% (free ambient) | 20–35% of power budget | 20–35% of power budget |
| PUE | < 1.15 | 1.4–1.6 typical | 1.3–1.5 typical |
| DC lease cost | $0 (own land) | $150–400/sqft/yr | Embedded in rack rate |
| Node OPEX / month | $3–5K | $18–25K | $15–22K |
| Gross margin at $50K/node | ~82% | ~52% | ~58% |
| Energy supply | 100% own, sized to demand | Grid mix | Grid mix |
Result: own power + free cooling bring the all-in energy cost per node to ~$500–800/month vs $1.5–2.5K at a standard DC.

Inside the data center
Not a collection of open-source components: the integrated stack creates the value.
Data halls 800 GPU nodes (~6,400 GPUs) on 10 MW IT in Phase 1; 4,000 nodes (~32,000 GPUs) at full build. 12.5 kW IT per 8-GPU node.

Modular 5 MW halls on own land, with large air-intake louvres for free ambient cooling. PUE < 1.15.

H200 SXM in Confidential Compute mode · TEE-enabled tensor parallel · NVLink isolation. Every request cryptographically attested.

7–12 °C year-round. PUE < 1.15 without liquid cooling, no water use, $0 HVAC.

Air-gapped management plane and customer-verifiable attestation logs: customer data never leaves the enclave.

SCADA-controlled hybrid plant: wind + solar + BESS firm the 24/7 DC load, with no export.
GLM-5.2 (744B MoE, 40B active, MIT) · DeepSeek V4 · model-agnostic serving
Open weight = no API dependency, 28× cheaper than GPT-4, licence allows commercial serving; switch models in weeks.
Per-request cryptographic proof · remote attestation chain · customer-verifiable logs
Competitors offer contracts. We offer math. For regulated buyers that is the requirement.
H200 SXM in Confidential Compute mode · TEE-enabled tensor parallel · NVLink isolation
CC-mode H200 ≠ standard H200. Specialist setup, 18-month head start, node OPEX $3–5K/mo.
Own substation · own fibre · own cooling · air-gapped mgmt plane · zero DC dependency
CoreWeave, Lambda, Together all rent DCs. We own ours: no eviction, no lease hike, no landlord SLA.
28→140 MW hybrid plant (wind + solar + BESS) · SCADA · sized to own load, no export
$0.045–0.06/kWh vs $0.10–0.15 for any DC in the EU/US. A permanent structural advantage.
Rented = 45–65% gross margin; own infra = 75–85%. At 10 MW (800 nodes) that is ~$50M/yr. Renting also removes sovereignty and ESG.
Power costs $0.10–0.20/kWh in Europe vs $0.045–0.06 own firmed generation here. Structural, not cyclical. Land cost is $0.
Attestation and infrastructure are model-agnostic. Value is in enclave + jurisdiction + energy cost; we switch models in weeks.
Competitive landscape
↔ Scroll sideways to see all columns
| CoreWeave | Nebius | Lambda Labs | Together AI | Crusoe Energy | Phala (TEE) | Bloovian | |
|---|---|---|---|---|---|---|---|
| GPU infrastructure | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Open-weight models | Partial | Partial | Partial | ✓ | — | ✓ | ✓ |
| Hardware TEE / attestation | — | — | — | — | — | ✓ (limited) | ✓ Full stack |
| Sovereign jurisdiction | — | EU only | — | — | — | — | ✓ LATAM/EU/UAE |
| Own renewable energy | — | — | — | — | Partial | — | ✓ 100% own-site |
| Own land / zero lease | — | — | — | — | Partial | — | ✓ 30,000 ha |
| LATAM / South America | — | — | — | — | — | — | ✓ Only player |
| Est. revenue / funding | $1B+ ARR | $200M+ | $60M+ | $300M+ | $80M+ | ~$3M rev | Phase 1: 10 MW |
| Node OPEX / month | $18–25K | $15–20K | $12–18K | $10–15K | $14–20K | $10–18K | $3–5K |
| Gross margin (own infra) | 45–55% | 50–60% | 45–55% | 50–60% | 55–65% | 40–55% | 75–85% |
White space: no competitor offers cryptographically attested, sovereign, 100%-renewable AI inference with own land and own power in LATAM, or anywhere at our cost structure.
Business model
Revenue = price × customers × utilisation · 75–85% gross margin on own infrastructure
$40–90K / node / month
Regulated enterprises: banks, healthcare, legal, defence, government
→ 80 nodes × $50K × 80% util. = $38.4M ARR at Phase 1
GM 75–85% · 4–8 nodes per customer · 12–24 month contracts
$0.008–0.025 / 1K tokens + $5–15K/mo SLA
Dev platforms, agents, compliance-sensitive SaaS migrating from closed models
→ 720 nodes (API + reserved) ≈ $126.7M ARR at Phase 1
GM 65–75% · OpenRouter listing → instant discovery
$2–8K per run or $15–25K/mo retainer
Enterprises evaluating open-model migration, incl. regulated buyers
→ 33 evals/month × $5K ≈ $2.0M ARR
GM 80–90% · funnel: eval converts to sovereign deployment
| $M | Phase 1 · 10 MW | Phase 2 · 20 MW | Phase 3 · 35 MW | Full · 50 MW |
|---|---|---|---|---|
| Reserved capacity (60%) | 80.6 | 161.3 | 282.2 | 403.2 |
| Confidential API (30%) | 46.1 | 92.2 | 161.3 | 230.4 |
| Sovereign (10%) | 38.4 | 76.8 | 134.4 | 192.0 |
| Parity evaluations | 2.0 | 3.0 | 5.0 | 8.0 |
| Total ARR | $167.1M | $333.3M | $582.9M | $833.6M |
Derivation: run-rate at phase completion; 12.5 kW IT per 8-GPU node. Sovereign = $50K × 80% util.; API ≈ $20K × 80%; reserved (2–3 yr contracts) ≈ $17.5K × 80%. No power-export revenue: energy is 100% own use.
Financial model
Build-out 10 → 50 MW IT by Year 5 · EBITDA positive from first nodes · ~70% of CAPEX debt/GPU-financed · FCF positive in Year 5
| P&L / FCF, base scenario | Year 2 · Phase 1 10 MW · 800 nodes | Year 3 · Phase 2 20 MW · 1,600 nodes | Year 4 · Phase 3 35 MW · 2,800 nodes | Year 5 · Full 50 MW · 4,000 nodes |
|---|---|---|---|---|
| Revenue (fiscal year) | $125.8M | $250.7M | $459.1M | $709.8M |
| – Node OPEX (power, cooling, maint.) | –$28.8M | –$57.6M | –$105.6M | –$163.2M |
| – Energy plant OPEX (2%/yr) | –$1.1M | –$2.0M | –$3.4M | –$4.7M |
| – People (all-in, incl. CFO/BD) | –$12.0M | –$20.0M | –$30.0M | –$40.0M |
| – GTM, legal, audit, overhead | –$6.5M | –$11.5M | –$19.9M | –$29.9M |
| EBITDA (margin) | $77.4M (62%) | $159.6M (64%) | $300.3M (65%) | $472.0M (66%) |
| – CAPEX depreciation | –$75.3M | –$150.1M | –$262.2M | –$374.4M |
| EBIT (margin) | $2.1M (2%) | $9.5M (4%) | $38.0M (8%) | $97.6M (14%) |
| – Energy park + DC facility CAPEX | –$67.5M | –$125.0M | –$187.5M | –$187.5M |
| – IT / GPU CAPEX (cash out) | –$168.0M | –$336.0M | –$504.0M | –$504.0M |
| + Debt / GPU financing (70%, net of debt service) | +$121.5M | +$238.3M | +$324.1M | +$222.3M |
| Free cash flow (net) | ~–$37M | ~–$63M | ~–$67M | ~+$3M |
| Cumulative FCF (incl. Yr 1) | ~–$106M | ~–$170M | ~–$237M | ~–$234M |
Basis: revenue on average nodes online; blended $17.2K effective/node/mo. Yr 1 = build year (~$21M revenue, ~–$70M FCF, in cumulative). OPEX $4K/node/mo. Depreciation: energy 20 yr, DC 15 yr, GPUs 5 yr. 70% of CAPEX debt-financed at 8%. Token upside excluded. No taxes.
Valued on year-end run-rate · equity value = enterprise value minus net debt from the financing plan
| Valuation metric | Year 2 10 MW · IPO readiness | Year 3 20 MW · listing / exit | Year 5 50 MW · full scale |
|---|---|---|---|
| ARR run-rate (4–6×) | $0.67–1.0B | $1.3–2.0B | $3.3–5.0B |
| EBITDA run-rate (10–15×) | $1.1–1.6B | $2.2–3.3B | $5.6–8.4B |
| Less net debt | –$0.3B | –$0.6B | –$1.3B |
| Target equity value | ~$0.5–1.0B | $1–2B | $2.5–5B+ |
Multiples: 4–6× ARR and 10–15× EBITDA (listed AI-infrastructure peers). Net debt from the 70% debt/GPU-financing plan. Planning estimates, not a forecast.
Traction & pipeline
Phase 0 target: 2 design partners + anchor contracts for ≥30% of the 10 MW Phase 1 before any energy/DC CAPEX
Design partner
LOI in negotiation
AI compliance team identified. Needs parity eval + sovereign deployment. Pre-approved budget $45K/month.
Design partner
Demo scheduled
HIPAA + GDPR + local law. No current AI solution. Shortlisted as one of 3 vendors.
Pipeline
Tender qualification
Explicit sovereign-AI mandate from the ministry. Own jurisdiction required by tender spec. Budget confirmed.
Pipeline
Commercial proposal
Migrating from OpenAI. Data residency requirement. $8K/month parity eval → $60K/month sovereign.
| Stage | Target (Phase 1) | Conversion | ARR per customer |
|---|---|---|---|
| Parity evaluation | 20 / quarter | → 40% to sovereign | $50–100K/yr |
| Design partner → pilot | 2 in Phase 0 | → 80% renew to full | $540K–$1.08M/yr |
| Sovereign deployment | 10–20 customers, Phase 1 | → 24-month lock-in | $480K–$1.08M/yr |
| Platform API | 10–20, Phase 2 | → volume ramp | $150–400K/yr |
Gate: energy/DC CAPEX starts only when anchor contracts cover ≥30% of Phase 1 (~250 nodes). If not reached, total spend is ~$4–6M: a capped option to test the thesis.
Roadmap
Every phase is gated · energy CAPEX starts only after commercial proof · the IPO is optional acceleration, not a prerequisite
Phase 0
$4–6M
✓ Gate: Anchors ≥30% of 10 MW · financing term sheets · no CAPEX before this gate
Phase 1
$0.45–0.57B
✓ Gate: EBITDA positive · ≥$150M ARR run-rate · 10 MW on own green power
Phase 2
$0.45–0.5B
✓ Gate: ~$330M ARR · 20 MW live · audited financials · S-1 in progress
Phase 3+
$1.4B (IPO + debt)
✓ Gate: Market conditions permitting · profitable standalone at any phase
Investment
Phase-gated · equity ~30%, debt/GPU financing ~70% · the minimum viable raise reaches anchor contracts before any CAPEX
1st close
$4–6M
Site survey, wind measurement, permits, grid study, TEE stack, 2 design partners + anchor LOIs, token legal. Walk-away option: if anchors don't sign, we stop.
2nd close
$140–170M
Equity tranche of the $0.45–0.57B Phase 1 (10 MW IT · 800 nodes · 28 MW energy park). Triggered only after anchors ≥30%; the rest via project debt + GPU financing.
Parallel
50–250 nodes
TEE-capable, CC-enabled; ~$420K each all-in = $20–105M capex value. Operators join via token rewards or revenue share. Reduces equity need 1:1.
Parallel
$47–63M
Direct stake in the 28 MW wind/solar + 50 MWh BESS plant via a 20-yr internal PPA with the DC. Hard asset, scales to 140 MW. Target levered IRR 12–15%.
Strongest investor position: Phase 0 seed ($5M) now + Phase 1 equity ($140–170M) conditional on anchor contracts + energy co-invest ($47–63M) + optional hardware nodes.
Exit options
Valuable to several acquirer categories; the form depends on market conditions at the time.
Year 3–4 · 20–35 MW
$1–2B at listing
≥$330M ARR run-rate · 20 MW live · 2 yrs GAAP · underwriter engaged
Comparables: Core Scientific (CORZ $2.5B), Applied Digital (APLD $600M), Iris Energy (IREN $1.1B). Green AI infra earns an ESG premium of 30–50% vs non-green peers.
Year 2–4 · any phase
$0.5–2B
≥$150M ARR + 10 MW live + own energy asset
NVIDIA (AI infra), Oracle (sovereign cloud LATAM), AWS/Google/Microsoft (LATAM DC gap), Aramco/ADNOC (sovereign AI). Energy + land + jurisdiction is what hyperscalers can't buy fast.
Year 3+ · Phase 2+
$0.4–1B
Stable EBITDA · long-term energy asset · contracted AI revenues
Brookfield, Macquarie, KKR Infrastructure value predictable cash flow + hard assets: renewable plant + contracted AI revenue + sovereign DC.
Separate, parallel · Phase 2+
Upside, not modelled
Token TGE · CEX listing
Equity investors hold Bloovian Inc. shares; token holders are separate. Legally and economically separate: no cross-contamination of returns or regulation.
ESG
Focused on what is real, measurable and investor-relevant · no speculative future programmes
CO₂ Scope 2: net zero · PUE < 1.15 · water use: zero
80–200+ jobs · 30,000 ha dual-use · local supply chain
GAAP from Phase 1 · TCFD-aligned · independent board by Phase 2


Team & governance
CFO hired on day 1 for IPO readiness · board and advisors recruited by Phase 1 · core team scales to ~50 FTE at 10 MW
| Role | FTE | Type | Cost | Responsibility |
|---|---|---|---|---|
| ML / inference systems eng. | 1.0 | New hire | $300–420K | GLM-5.2 on 8×H200, tensor parallel, OpenAI endpoint |
| TEE / confidential computing eng. | 1.0 | New hire | $320–450K | H200 CC mode, TDX/SEV-SNP, attestation, sealed-box eval |
| Eng lead / founder | 1.0 | Founder | Equity | Architecture, design partners, investors, strategy |
| CFO / finance lead | 0.5 | New hire | $250–400K | GAAP from day 1, auditor, IPO readiness, IR |
| Energy project manager | 0.5 | Contractor | $80–120K | 28 MW park + 10 MW DC: EIA, EPC, CAMMESA, 132 kV |
| Enterprise BD / sales | 0.5 | New hire | $120–180K | Regulated pipeline, design-partner LOIs, first 2 signed |
| Legal / token counsel | 0.5 | Contractor | $80–150K | Corporate structure, utility token, LATAM regulatory |
AI infra
Former hyperscaler infrastructure executive: technical credibility and operations at scale.
Capital
Ex-investment banker or CFO with NASDAQ listing experience, for Phase 2 IPO preparation.
Energy
Argentine or regional renewables expert: Patagonia permits, CAMMESA, grid tie, project finance.
AI / ML
Research-level expertise in TEE/attestation or AI model serving at scale.
Governance: GAAP financials, audit readiness and investor reporting from Phase 1; board and advisors in place ahead of any institutional raise.
Risk matrix
Phase 0 gate: energy/DC CAPEX starts only when anchor contracts cover ≥30% of Phase 1. Walk-away cost if Phase 0 fails: ~$4–6M, a cheap option to test the thesis.
Phase 1 = $0.45–0.57B, ~70% debt/GPU-financed against contracted revenue. Modular 5 MW blocks, fixed-price EPC, +15–20% Patagonia logistics premium already modelled.
CoreWeave ($1.1B+) and Nebius (NASDAQ) are real. Our moat is six compounding layers that take 24 months and $130–160M (excl. GPUs) to replicate.
The IPO is Phase 3+ and optional. EBITDA positive from first nodes; growth paced by contracts and financing. The IPO accelerates 35→50 MW, it is not required.
The attestation and infrastructure stack is model-agnostic. If DeepSeek disappears we switch models in weeks. Value sits in enclave + jurisdiction + energy cost.
Token sits in a separate entity, structured as utility with specialist counsel from Phase 0. Optional financing: the plan is profitable without token proceeds.
Corporate structure
Two entities, one integrated asset stack. Bloovian Inc. owns all hard assets; the foundation manages the token. They are legally separate.
Operating company · owns all hard assets and IP · revenue-generating · the equity investors own
Separate legal entity · token holders are not equity holders
Investor clarity: buying equity in Bloovian Inc. = owning the land, energy plant, DC, AI revenue and IP. The token is a separate financing instrument. A NASDAQ IPO is equity only.
Impressions
Visualisations of the site, the campus and the data center.

Bloovian Inc.
The only verifiable, jurisdiction-sovereign, 100%-renewable AI inference platform in Latin America: built on own land, own power and a six-layer structural moat that takes 24+ months and $130–160M (excl. GPUs) to replicate.