Every 1,000 devices + 1 G292-280 server = one cluster. $875K net revenue per cluster annually. EBITDA+ in 12 months. The math works at every scale.
One RevoFi cluster combines edge computing power with central AI inference to deliver maximum revenue per deployed asset.
1,000 Devices
NVIDIA Jetson Link/Link Pro
Role: Edge AI inference, dVPN, CDN caching, venue WiFi
1x G292-280 Server
8x NVIDIA A100 80GB HBM2e
Role: Heavy AI inference, LLM serving, Triton Inference Server
Mode A: Triton Client
Jetson preprocesses, A100 infers (large models 70B+)
Mode B: Local Inference
Jetson runs lightweight models directly (dVPN, CDN, light vision)
Fleet managed via SaltStack (infrastructure) + revos-agent (workloads)
Full CAPEX breakdown showing hardware revenue offsets server costs on Day 1.
| Device Fleet | Per Unit | x 1,000 |
|---|---|---|
| Hardware BOM | $400 | $400,000 |
| Assembly + QC | $25 | $25,000 |
| Import/Shipping | $35 | $35,000 |
| Total Landed | $460 | $460,000 |
| Retail Revenue | $1,000 | $1,000,000 |
| Gross Profit | $540 | $540,000 |
| Server Infrastructure | |
|---|---|
| G292-280 (8x A100 80GB) | $180K-$220K |
| Rack + Switches + Install | Included |
| Electricity/month | $800-$1,000 |
| Electricity/year | $12,000 max |
Total CAPEX
~$680K
Hardware Revenue
$1,000,000
Day 1 Net Position
+$320K
Four revenue streams combine to generate $1.75M gross per cluster annually.
Hardware Margins
RDC/WaaS Billing
B2B Venue/Retail
Cloud AI Inference
Total Gross
$1,750,000/year
Net Revenue
$875,000/year
Timeline
EBITDA+ in 12 Months
Two-phase distribution model ensures investors recover capital quickly while incentivizing long-term channel partner commitment.
Before Capital Repayment
After Capital Repayment
Channel Partner revenue share increases from 4% to 20% after investor payback. This 5x long-term incentive secures high-quality venue partners for the full device lifecycle.
Payback Timeline: 12-15 months at 80% preferred return → perpetual 20% share for remaining 8-9 year device lifecycle.
Complete pricing schedule for all compute resources. First platform to unbundle VRAM from GPU compute.
Conversion Rate: 1 RDC = $0.000001 USD
| Resource | Asset | USD Price | RDC Price | Billing |
|---|---|---|---|---|
| Edge Fleet Tier (Jetson Devices) | ||||
| CPU | Jetson ARM vCPU | $0.015/hr | 15,000 RDC/hr | Per-Second |
| GPU | Jetson 128-core Maxwell | $0.05/hr | 50,000 RDC/hr | Per-Second |
| Memory | Jetson LPDDR4 | $0.004/GB-hr | 4,000 RDC/GB-hr | Per-GB-Second |
| Storage | Micro SD (Persistent) | $0.021/GB-mo | 21,000 RDC/GB-mo | Per-GB-Second |
| Connectivity | Ethernet (Venue ISP) | $0.10/GB | 100,000 RDC/GB | Per-KB |
| Connectivity | Cellular (Twilio SuperSIM) | $0.12/MB | 120,000 RDC/MB | Per-MB |
| Server Tier (G292-280) | ||||
| CPU | Intel Xeon vCPU | $0.025/hr | 25,000 RDC/hr | Per-Second |
| GPU (Compute) | NVIDIA A100 (FLOPS) | $0.30/hr | 300,000 RDC/hr | Per-Second |
| VRAM (GPU) | A100 80GB HBM2e | $0.005/GB-hr | 5,000 RDC/GB-hr | Per-GB-Second |
| Memory | Server DDR4 | $0.004/GB-hr | 4,000 RDC/GB-hr | Per-GB-Second |
| Storage | NVMe (Ephemeral) | $0.08/GB-mo | 80,000 RDC/GB-mo | Per-GB-Second |
| Storage | SSD (Persistent) | $0.021/GB-mo | 21,000 RDC/GB-mo | Per-GB-Second |
| Connectivity | Server Fiber | $0.10/GB | 100,000 RDC/GB | Per-KB |
First platform to separately bill GPU compute from GPU memory. Developers with optimized 10GB models pay 75% less than bundled 80GB competitors.
| Provider | A100 80GB/hr | VRAM Unbundled? |
|---|---|---|
| AWS | $3.02 | No |
| Vast.ai | $0.69 | No |
| RevoFi | $0.70 | Yes — $0.005/GB-hr |
Self-reinforcing economics where hardware margins fund unlimited private services, driving more hardware purchases.
25% Cap
$800-$1,000/mo electricity
to serve ALL device owners
Day 1 Profit
$580-620 margin per device
covers 7-8 years of usage
70%+ Reserved
Majority capacity ring-fenced
for paid B2B/WaaS workloads
| Service | Monthly Quota | Avg Usage | % Hit Cap | Usage Bar |
|---|---|---|---|---|
| AI Tokens | 1,000,000 | 300K-600K | <8% | |
| dVPN | 500 GB | 80-150 GB | <5% | |
| Messaging | 50K messages | ~8 GB | <3% | |
| Cellular | 1 GB | 400-700 MB | <12% |
"Free" private services
drives hardware purchases
Hardware margins
fund server infrastructure
25% cap
preserves 70%+ for paid workloads
Paid workloads
generate RDC/USDC revenue
RVS rewards
offset user costs
More purchases
repeat the cycle
The Flywheel
99%+ of users never hit their caps. It feels unlimited. They buy more devices. We print money.
5-year cluster deployment timeline showing path from 2 clusters to 13.5 clusters.
| Year | New Clusters | Cumulative | Devices | Net Revenue | Key Driver |
|---|---|---|---|---|---|
| 2026 | 2 | 2 | 2,000 | $1.75M | Hardware flywheel |
| 2027 | 1 | 3 | 3,000 | $3.10M | Venue conversions |
| 2028 | 1.5 | 4.5 | 4,500 | $4.65M | Cloud WaaS uptake |
| 2029 | 4.5 | 9 | 9,000 | $6.05M | Capacity scaling |
| 2030 | 4.5 | 13.5 | 13,500 | $7.50M | Multi-device ownership |
5-Year Cumulative
10-Year Cumulative
Distribution: 70% consumer / 30% business • Consumer $680/yr • Business $880/yr • Server $250K-350K/yr
Three deployment models demonstrating cluster economics across different use cases.
300 Leetop A203 nodes from existing inventory. Zero additional CAPEX for compute hardware.
Purpose: Production testbed, platform validation
Hotels deploy clusters in lobbies + rooms. Seamless WiFi + AI, no portals or fees.
Scale: Single hotel to 500+ locations
Computer vision + AI inference at point of sale. Edge processing for real-time analytics.
Margin: Highest per-location revenue
Ready to deploy a cluster? Request a custom deployment quote or explore our equity and revenue opportunities.