What to measure. Centralized metrics vs protocol-era metrics.
| Category | What It Measures | Centralized | Protocol-Era |
|---|
| Data | Volume and quality | Proprietary datasets | On-chain verified collections |
| Financial | Return on investment | Revenue, margins | Token yields, burn rate |
| Network | Infrastructure scale | Data center capacity | Device count, coverage |
| Community | Participation health | N/A | Operator distribution, governance |
Data Metrics
Volume
| Metric | Centralized | DePIN Protocol | Why Better |
|---|
| Dataset size | Proprietary, opaque | On-chain attestations | Verifiable |
| Collection rate | Internal only | Real-time protocol data | Transparent |
| Geographic coverage | Corporate footprint | Global device map | Community-driven |
Quality
| Metric | Centralized | DePIN Protocol | Why Better |
|---|
| Accuracy | Internal QA | Cryptographic verification | Trustless |
| Labeling quality | Contracted reviewers | Staked attestation | Incentive-aligned |
| Freshness | Batch processing | Streaming telemetry | Real-time |
Financial Metrics
Centralized Data
| Metric | Benchmark | What It Shows |
|---|
| Revenue | Scale AI ~$870M (2024), targeting $1.5B ARR | Market validation for data services |
| Gross margin | 60-80% | Data leverage — collect once, sell many times |
| Growth rate | 50-100% YoY | AI demand driving data demand |
| Valuation | Scale AI $29B (post-Meta investment) | Market pricing of data infrastructure |
DePIN Data Protocol
| Metric | Benchmark | What It Shows |
|---|
| On-chain revenue | Growing | Real demand for protocol data |
| Token burn rate | Burn > issuance = healthy | Sustainable token economics |
| Revenue per device | Varies by vertical | Operator economics |
| Protocol revenue share | % distributed to operators | Alignment strength |
| Centralized | Protocol-Era | Shift |
|---|
| Revenue | On-chain revenue | Transparent, verifiable |
| Gross margin | Protocol take rate | Distributed to operators |
| Customer count | Data consumer count | Permissionless access |
| Growth rate | Device deployment rate | Community-driven growth |
Network Metrics
| Metric | What It Measures | Target |
|---|
| Device count | Infrastructure scale | Growing month over month |
| Geographic coverage | Spatial completeness | Expanding to new regions |
| Data throughput | Network capacity | Increasing with demand |
| Uptime | Reliability | >99% device availability |
| Device diversity | Resilience | Multiple device types per region |
Market Sizing
| Segment | 2025 Estimate | 2030 Projection | Growth Driver |
|---|
| AI training data | $3.5B | $13B+ | Frontier model demand (23% CAGR) |
| Data labeling | $5-7B | $20B+ | Scale AI model expanding |
| GPU compute (NVIDIA DC alone) | $115B | $300B+ | Training + inference demand |
| Decentralized storage | $500M | $5B+ | Data sovereignty requirements |
| DePIN data networks (total on-chain) | $72M FY2025, $150M/mo Jan 2026 | $15B+ | 270% YoY market cap growth |
Opportunity Assessment
Scoring Dimensions
| Dimension | Weight | AI Data Score | Evidence |
|---|
| Market Attractiveness | 20% | 8.5 | $115B+ NVIDIA DC alone, $13B+ training data by 2034 |
| Technology Disruption | 20% | 8.0 | DePIN networks 300%+ YoY growth |
| VVFL Alignment | 25% | 7.5 | Loop works, quality verification is the gap |
| Competitive Position | 20% | 7.0 | Infrastructure phase, first-mover available |
| Timing Risk | 15% | 7.0 | Build phase 2025-2027, institutional adoption 2027+ |
Aggregate: 7.6/10 — Strong Conviction
Opportunity Matrix
| Opportunity | Score | Timing | Key Risk |
|---|
| DePIN sensor networks | 8.0 | Now | Device unit economics |
| Distributed GPU compute | 8.5 | Now | Hyperscaler competition |
| Data labeling protocols | 7.0 | 1-2 years | Scale AI dominance |
| Decentralized storage | 6.5 | Now | Filecoin adoption curve |
| AI data marketplaces | 7.5 | 1-2 years | Liquidity and pricing |
Watch Signals
| Signal | Bullish | Bearish |
|---|
| DePIN device growth | >50% QoQ | Plateaus |
| On-chain data revenue | Exceeds token issuance | Issuance dominates |
| Enterprise adoption | Fortune 500 using DePIN data | Remain in pilots |
| Regulatory | Data sovereignty laws strengthen | Status quo |
| AI demand | Frontier models need more data | Synthetic data suffices |
| Principle | What to Measure |
|---|
| Data compounds | Dataset growth rate, model accuracy improvement |
| Collection is physical | Device count, geographic density |
| Quality beats quantity | Verification rate, premium over commodity data |
| Ownership creates alignment | Operator retention, revenue per device |
| Compute follows data | Edge processing %, inference latency |
Context