redc — Executive Summary (Legacy)¶
Renewable Energy Data Cell Prepared: April 2026
Note: Superseded by the multi-phase master BUSINESS-PLAN.md (Phase A colo → Phase B scale → Phase C RE container). This file is reference-only. Execution: execution/phase-a/.
One-Liner¶
redc converts cheap excess renewable energy (solar + biogas) into GPU compute and sells the computing power — earning 3–7× more per kWh than feeding electricity into the grid.
The Problem¶
Two markets are broken in opposite ways:
Renewable energy in Germany is increasingly worthless at the point of generation. Solar curtailment doubled in 2024 (1,389 GWh wasted) and doubled again in 2025. Grid connections are exhausted until ~2030. During peak sun, the market value of solar electricity drops as low as 2 ct/kWh. Simultaneously, ~1,400 biogas plants are losing their 20-year EEG subsidies in 2025–2026, facing 50–60% revenue drops and potential shutdown. Both types of renewable energy operators are stuck with energy they can't sell profitably.
GPU compute is in massive shortage. AI workload demand is growing 2.25× per year. Vast.ai, the leading GPU marketplace, saw 16× customer growth in one year. Enterprises face multi-month queues for GPU capacity. The inference market alone is projected to reach $51.6B by 2032.
The Solution¶
Place GPU servers at renewable energy sites — solar farms and biogas plants. Use the energy that would otherwise be curtailed, sold at rock-bottom prices, or wasted to power AI compute. Sell the compute on established marketplaces and to direct customers at 3–7× the revenue per kWh vs. selling the energy.
The energy doesn't need to travel. The results do.
The solar+biogas hybrid model is key: solar provides the cheapest daytime energy (€0.04–0.06/kWh), while biogas fills nights and gaps (€0.08–0.10/kWh) — delivering 100% renewable, 24/7 compute without any grid dependency.
Why It Works¶
| Factor | Detail |
|---|---|
| Energy arbitrage | Solar feed-in: €0.08/kWh. Grid self-consumption: €0.25/kWh. Compute revenue: €0.40–0.56/kWh. |
| Behind-the-meter | Co-locating at the energy source avoids grid fees, taxes, and surcharges (€0.10–0.12/kWh saved) |
| 24/7 renewable via hybrid | Solar (daytime) + biogas (24/7 baseload) = 100% renewable with no grid dependency. Blended cost: €0.06–0.08/kWh |
| Hardware maturity | Used NVIDIA A100 GPUs available at €12k (down from €18–25k new), still earning €1.00/hr on marketplace |
| Proven marketplace | Vast.ai: 14,000+ paying customers, 20,000+ GPUs listed, explosive growth |
| Dual partner pool | Thousands of curtailed solar farms and 1,400+ post-EEG biogas plants both actively seeking new revenue streams |
Market Opportunity¶
| Segment | Size | redc's angle |
|---|---|---|
| GPU cloud marketplace | Growing 10×+ YoY (Vast.ai alone) | Entry channel: list GPUs, earn immediately |
| EU green compute (CSRD-driven) | Thousands of companies need verified Scope 2/3 reporting | Premium pricing for auditable renewable compute |
| AI inference API market | $51.6B by 2032 | Managed services on own infrastructure |
| Solar curtailment (Germany) | 2,700+ GWh/yr wasted, growing | Unique supply-side opportunity (daytime) |
| Post-EEG biogas plants | 1,400+ losing subsidies in 2025–26 | 24/7 baseload energy supply-side opportunity |
Validated opportunity: Others have independently reached the same thesis — rhöncloud (mid-scale facility, Q3 2026), Projekt Jupiter (hyperscale, Brandenburg), Deutsche Telekom (BTC mining pilot on surplus solar). Nobody is doing the distributed small-scale model redc targets.
Business Model¶
Revenue¶
Primary: GPU compute rental (marketplace + direct customers), tiered by availability: - Firm — battery-backed, 99.9% SLA, premium price - Flexible — solar-hour scheduling, 30–60% discount, for batch workloads - Spot — excess capacity only, deepest discount
Secondary: Grid services (demand response), green compute premium, platform licensing (at scale).
Unit Economics (per A100 GPU, Year 3 steady state)¶
| Solar+grid | Biogas only | Solar+biogas hybrid | |
|---|---|---|---|
| Revenue (70% util, €0.95/hr avg) | €5,825 | €5,825 | €5,825 |
| Energy cost | -€417 | -€245 | -€196 |
| Hardware depreciation (3-year) | -€4,000 | -€4,000 | -€4,000 |
| Operating share | -€1,050 | -€1,050 | -€1,050 |
| Net margin per GPU | €358 | €530 | €579 |
| Net margin % | 6% | 9% | 10% |
At fleet level (48+ GPUs), fixed cost dilution and direct customer rates push net margins to 14–22%. The hybrid model adds ~€221/GPU/year vs. solar+grid — at 128 GPUs (Year 5), that's ~€28k additional annual profit.
Financial Projections (5-Year, Two Variants)¶
Variant A: Aggressive Growth (external funding, full-time from Year 3)¶
| Year 1 | Year 2 | Year 3 | Year 4 | Year 5 | |
|---|---|---|---|---|---|
| GPUs | 16 | 48 | 48 | 80 | 128 |
| Sites | 1 | 3 | 3 | 5 | 6 |
| Revenue | €51k | €130k | €280k | €505k | €884k |
| EBIT | -€22k | -€34k | +€62k | +€124k | +€302k |
| Investment (cum.) | €245k | €631k | €631k | €1.0M | €1.5M |
Variant B: Steady Growth (side-project, self-funded, 3–5 hrs/week)¶
| Year 1 | Year 2 | Year 3 | Year 4 | Year 5 | |
|---|---|---|---|---|---|
| GPUs | 4 | 8 | 16 | 24 | 32 |
| Sites | 0 | 1 | 1 | 1–2 | 2 |
| Revenue | €21k | €43k | €91k | €144k | €199k |
| Cash flow | +€14k | +€20k | +€55k | +€96k | +€113k |
| Investment (cum.) | €52k | €104k | €208k | €312k | €416k |
Recommended: Start with Variant B. Months 1–18 are identical. Decision point at Month 18 with real data. See 008 for detail.
Phased Approach¶
| Phase | When | Investment | Goal | Go/No-Go |
|---|---|---|---|---|
| 0 — Validate | Month 1–6 | €52k (self-funded) | 4 GPUs, prove ≥65% utilization on marketplace | If util <50% after 6 months → stop |
| 1 — First site | Month 7–18 | €193k | 16 GPUs at solar farm, first PPA | If unit economics negative → stop |
| 2 — Expand | Month 19–36 | €386k | 48 GPUs, 3 sites, first direct customers | If no direct customers → marketplace only |
| 3 — Scale | Month 37–60 | €400–900k | 128 GPUs, 6 sites, platform layer | If can't scale → profitable single-site business |
Each phase has a clear kill criterion. The business never commits capital it can't recover by selling hardware.
Team¶
| Name | Role | Contribution |
|---|---|---|
| David | Founder, technical and operations | Software/infra, marketplace ops, Phase A–B execution; Phase C may add external energy/partnership support |
Key gap to fill (Phase 2+): Operations / sales hire for customer acquisition and site management.
Competitive Advantage¶
- Validated thesis — rhöncloud, Projekt Jupiter, and Deutsche Telekom independently prove the renewable-energy+compute concept works. redc executes it at lower cost and risk.
- Multi-source energy partnerships — solar farms (curtailment) + biogas plants (post-EEG) provide a dual partner pool. Physical co-location creates long-term, relationship-based moat. Site access and credible EE conversations matter at scale.
- 24/7 renewable via hybrid — solar+biogas combination delivers verifiable 100% renewable compute around the clock. Nobody else offers this at small/mid scale.
- Behind-the-meter economics — 70–90% energy cost reduction vs. grid (depending on energy mix)
- Distributed model — small, standardized deployments at multiple renewable energy sites. Nobody else is doing this.
- Dual tailwind — solar curtailment doubling yearly (2,700 GWh, €3.1B cost) and 1,400+ biogas plants losing subsidies. Both trends create willing partners with no fix in sight.
- EU data residency — GDPR-compliant, verifiable green compute for CSRD reporting
- Bounded downside — hardware has liquid resale value; no irreversible commitments
Funding Ask¶
| Phase | Amount | Use of funds | Type |
|---|---|---|---|
| 0 | €52k | 4 GPUs + server for validation | Self-funded |
| 1 | €193k | 16 GPUs, container, first solar farm deployment | Seed (bank loan / angel) |
| 2 | €386k | 32 additional GPUs, 2 more sites | Growth (bank / equity) |
Total Phase 0–2: €631k over 30 months. Security for bank financing: GPU hardware with tangible resale value (€9–12k per unit).
Key Risks and Mitigations¶
| Risk | Mitigation |
|---|---|
| GPU marketplace rates decline | Shift to direct customers and managed services |
| Low utilization | Phase 0 validates demand before major investment; multi-platform listing |
| Hardware failure | Warranty, maintenance reserve, spare parts |
| Renewable energy partnership unavailable | Model works on grid too; dual partner pool (solar + biogas); multiple conversations in parallel |
| Connectivity at remote sites | Site selection criterion; Starlink as backup; batch workloads are bandwidth-tolerant |
Next Steps¶
- Validate immediately (€52k risk): Purchase 4 used A100 GPUs, list on Vast.ai, measure utilization for 6 months
- Begin energy partner conversations (when Phase C nears): Identify 3–5 candidate solar farms and/or post-EEG biogas plants in southern Germany (outreach, MaStR, industry contacts)
- Incorporate GmbH: Prepare for Phase 1 financing
- Build investor deck: Based on this analysis series
Supporting Analysis¶
| Document | Key finding |
|---|---|
| 001 — Profitability by Scale | Energy arbitrage is real (3–7×), but consumer GPUs don't work. Datacenter GPUs at mid-scale do. |
| 002 — Hardware Unit Economics | Used A100 80GB is the optimal GPU. Break-even at 67% utilization, ≥8 GPUs. |
| 003 — Solar Farm Partnership | Behind-the-meter PPA at €0.05/kWh. Solves curtailment crisis for solar farms. No blockers. |
| 009 — Biogas Partnership | Post-EEG biogas plants offer 24/7 baseload at €0.08–0.10/kWh. Hybrid (solar+biogas) = 100% renewable, lowest blended cost. |
| 004 — Competitive Landscape | Thesis validated by others (rhöncloud, DT pilot, Projekt Jupiter). Distributed small-scale model is uncontested. |
| 005 — Financial Projections | Profitable Year 3. €884k revenue, €302k EBIT, 34% margin by Year 5. |
| 006 — Risk Analysis | No showstoppers. All risks have viable mitigations. Downside is bounded. |
| 008 — Growth Variants | Aggressive (128 GPUs Y5, needs full-time) vs. Steady (32 GPUs Y5, side-project). Recommend: start Steady, switch to Aggressive if data warrants. |