# Settlement Layer Comparison

**Radius is a settlement layer — the infrastructure where stablecoin micropayments reach finality.** This document compares Radius to other settlement options for agent-to-agent micropayments.

---

## What is Settlement?

Settlement is when a payment becomes final and irrevocable. In traditional finance, settlement happens when money moves between bank accounts and can no longer be clawed back. In blockchain systems, settlement happens when a transaction is included in a block and confirmed by validators.

**For agent micropayments, settlement determines:**
- **How fast** an agent can continue its workflow after paying
- **How much** each transaction costs to process
- **How reliably** the system performs under load
- **Whether costs** are predictable or variable

**The key question:** When an AI agent pays $0.001 for an API call, where does that payment settle, and how do the settlement layer's properties affect whether that transaction is viable at scale?

---

## Settlement Requirements by Use Case

Different agent use cases have different settlement requirements:

| Use Case | Throughput Needed | Latency Tolerance | Cost Sensitivity | Example |
|----------|------------------|-------------------|------------------|---------|
| **Consumer agent purchases** | Low (1-100 TPS) | High (seconds to minutes) | Low (fees can be % of transaction) | Agent books a flight for $400 |
| **API micropayments** | Medium (1K-100K TPS) | Medium (1-3 seconds) | High (fee must be <10% of transaction) | Agent pays $0.01 per API call |
| **Streaming payments** | High (100K-1M TPS) | Low (sub-second) | Extreme (fee must be <1% of transaction) | Agent pays $0.0001 per LLM token in real-time |
| **Agent mesh networks** | Very High (1M+ TPS) | Very Low (<500ms) | Extreme (fee must be negligible) | Agents paying each other for routing/compute in distributed networks |

**The scaling problem:** As you move down this table, fewer settlement options remain viable. Consumer purchases can happen on any blockchain or even traditional rails. Internet-scale agent micropayments require purpose-built infrastructure.

---

## Settlement Options: Head-to-Head Comparison

### Performance & Cost

| Settlement Layer | Real-World Throughput | Finality Time | Cost per Transaction | Cost per Million Txns |
|-----------------|---------------------|---------------|---------------------|---------------------|
| **Radius** | 2.8M+ TPS (tested, linearly scalable) | ~500ms | ~$0.00001 | ~$10 |
| **Base** | ~100 TPS (current, roadmap to 3K+) | ~2 seconds | ~$0.001 | ~$1,000 |
| **Solana** | 1,000-4,000 TPS (real-world average) | ~400ms | ~$0.00025 | ~$250 |
| **Arbitrum** | ~40 TPS (current) | ~13 seconds (L1 finality) | ~$0.0001-$0.001 | ~$100-$1,000 |
| **Optimism** | ~50 TPS (current) | ~13 seconds (L1 finality) | ~$0.0001-$0.001 | ~$100-$1,000 |
| **Polygon PoS** | ~500 TPS (real-world) | ~2 seconds | ~$0.0001-$0.001 | ~$100-$1,000 |
| **Tempo** | Not publicly disclosed | Not disclosed | Not disclosed | Not disclosed |
| **Kite AI** | Not publicly disclosed | ~1 second (reported) | Low (subsidized) | Not disclosed |
| **Lightning Network** | Theoretically unlimited (off-chain) | Instant (channel state) | ~$0.000001 | Near-zero |

### Design Properties

| Settlement Layer | Purpose-Built for Agents | Stablecoin-Native | No Native Token | EVM Compatible | Predictable Costs |
|-----------------|-------------------------|-------------------|-----------------|----------------|-------------------|
| **Radius** | ✓ | ✓ | ✓ | ✓ | ✓ |
| **Base** | ✗ (general-purpose L2) | ✓ (USDC native) | ✗ (uses ETH for gas) | ✓ | ✗ (gas varies) |
| **Solana** | ✗ (general-purpose L1) | Partial (USDC available) | ✗ (uses SOL for gas) | ✗ | Mostly ✓ |
| **Arbitrum** | ✗ (general-purpose L2) | Partial (bridged USDC) | ✗ (uses ETH for gas) | ✓ | ✗ (gas varies) |
| **Optimism** | ✗ (general-purpose L2) | Partial (bridged USDC) | ✗ (uses ETH for gas) | ✓ | ✗ (gas varies) |
| **Polygon** | ✗ (general-purpose L1) | Partial (bridged USDC) | ✗ (uses MATIC for gas) | ✓ | ✗ (gas varies) |
| **Tempo** | ✓ (reported) | Unknown | Unknown | Unknown | Unknown |
| **Kite AI** | ✓ (Avalanche subnet) | Partial | ✗ (uses custom token) | ✗ | Unknown |
| **Lightning** | Partial (BTC-native) | ✗ (BTC only) | ✓ | ✗ | ✓ |

---

## Detailed Analysis

### Conventional Rails (Stripe, Visa, ACH)

**What they are:** Traditional payment networks built for human-initiated transactions.

**Strengths:**
- Universal merchant acceptance
- Mature fraud prevention and dispute resolution
- Regulatory clarity and compliance built-in
- No crypto/blockchain complexity

**Why they don't work for agent micropayments:**
- **Cost floor:** $0.30 + 2.9% per transaction makes anything under $10 uneconomical
- **Throughput:** Visa's global capacity (~65K TPS) is insufficient for internet-scale agent payments
- **Latency:** Settlement takes 1-3 days; authorization takes seconds but isn't final
- **Programmability:** Can't encode spending rules, conditional logic, or machine-readable policies into payments

**Viable for:** High-value agent purchases ($50+) where convenience and merchant acceptance matter more than cost efficiency.

**Not viable for:** API micropayments, streaming payments, agent-to-agent commerce at scale.

---

### General-Purpose Layer 1s (Ethereum, Solana, Avalanche)

**What they are:** Base-layer blockchains designed for decentralized applications, DeFi, NFTs, gaming, and general smart contract execution.

#### Ethereum L1

**Current state:** ~15-30 TPS, $1-$50 per transaction (varies with gas prices), 13-second finality.

**Why it doesn't work:** Too slow and too expensive for micropayments. A $0.001 API call would cost $1-$50 to settle.

**Viable for:** High-value transactions ($1,000+) where decentralization and security are paramount.

**Not viable for:** Agent micropayments of any kind.

#### Solana

**Strengths:**
- Fast: ~400ms finality
- Cheap: ~$0.00025 per transaction
- High throughput: 1,000-4,000 TPS real-world

**Limitations:**
- **Not agent-specific:** Throughput shared with DeFi, NFTs, gaming — agent payments compete for block space
- **Occasional instability:** Network has experienced multi-hour outages during high load
- **Non-EVM:** Requires Rust/Solana-specific development, can't easily port Ethereum contracts
- **Gas in SOL:** Agents need to hold SOL to pay fees, adding complexity vs. stablecoin-only systems

**Viable for:** Medium-scale agent micropayments (10K-100K TPS) where cost and speed matter more than guaranteed uptime.

**Not viable for:** Mission-critical agent infrastructure requiring 99.99% uptime and predictable multi-million TPS capacity.

---

### General-Purpose Layer 2s (Base, Arbitrum, Optimism, Polygon)

**What they are:** Scaling solutions built on top of Ethereum L1. They batch transactions off-chain and post commitments to Ethereum for security.

#### Base (Coinbase L2)

**Strengths:**
- USDC-native (deepest stablecoin liquidity)
- Primary settlement layer for x402 today (Coinbase AgentKit, Stripe machine payments)
- EVM-compatible (easy migration from Ethereum)
- Strong institutional backing (Coinbase)

**Current limitations:**
- **Throughput:** ~100 TPS currently, roadmap to 3,000+ TPS (not yet delivered)
- **Finality:** ~2 seconds under normal conditions, degrades under load
- **Shared block space:** Agent payments compete with DeFi swaps, NFT mints, onchain gaming
- **Variable costs:** Gas spikes during congestion — your $0.001 API call might cost $0.01 to settle during peak

**Viable for:** Early-stage x402 adoption, agent payments below 100K TPS, use cases that can tolerate occasional latency spikes.

**Not viable for:** Internet-scale agent payments (500K+ TPS), applications requiring guaranteed sub-second finality.

**The migration path:** If you're building on Base today, Radius is EVM-compatible — moving agent payment settlement to Radius is a configuration change, not a rewrite. You keep the same contracts, same wallets, same stablecoin (USDC). You just point settlement at infrastructure purpose-built for this workload.

#### Arbitrum & Optimism

**Strengths:**
- EVM-compatible
- Lower costs than Ethereum L1 (~$0.0001-$0.001 per transaction)
- Mature ecosystems

**Limitations:**
- **Throughput:** 40-50 TPS currently (lower than Base)
- **Finality:** Optimistic rollups require 7-day challenge period for L1 finality (fast finality available via centralized sequencers, but not final)
- **Not stablecoin-native:** USDC must be bridged from Ethereum, adding friction
- **Shared block space:** Same congestion issues as Base

**Viable for:** General smart contract applications, but not optimized for agent micropayments.

**Not viable for:** High-frequency agent payments requiring guaranteed sub-second finality.

---

### Purpose-Built Agent Settlement Layers

#### Radius

**What it is:** A settlement layer designed specifically for machine-to-machine stablecoin micropayments.

**Architecture:** PArSEC (Parallel Architecture with Sequential Execution Consistency) — transactions are partitioned across multiple sequencers, each handling a subset of accounts. Adding sequencers increases throughput linearly.

**Performance:**
- **Throughput:** 2.8M+ TPS tested, linearly scalable
- **Finality:** ~500ms guaranteed, doesn't degrade under load
- **Cost:** ~$0.00001 per transaction (~$10 per million transactions)
- **Predictability:** No native token, fees paid in stablecoin, no gas price spikes

**Design principles:**
- **Stablecoin-native:** Settles only in USDC, PYUSD, or other fiat-pegged tokens
- **No native token:** Transaction fees paid in the same stablecoin being transferred
- **EVM-compatible:** Contracts written for Ethereum/Base run without modification
- **No MEV:** No mempool, no front-running, deterministic transaction ordering

**Viable for:** Internet-scale agent micropayments (500K+ TPS), applications requiring guaranteed performance SLAs, use cases where cost predictability is critical.

**Not viable for:** General-purpose smart contracts requiring complex cross-contract interactions (DeFi), applications requiring maximum decentralization over performance.

**Status:** Testnet. x402 facilitator in active development.

#### Tempo

**What it is:** A purpose-built "agent commerce chain" (limited public information available).

**Known properties:**
- Positioning as agent-native settlement infrastructure
- Performance specs not publicly disclosed

**Assessment:** Direct competitor to Radius in the purpose-built settlement category. Differentiation will come down to disclosed performance, EVM compatibility, and ecosystem adoption.

#### Kite AI

**What it is:** An Avalanche subnet designed for agentic commerce.

**Known properties:**
- SPACE framework: Stablecoin-native, Programmable constraints, Agent-first authentication
- "Payment Lanes" to guarantee block space for specific applications
- ~1 second finality (reported)
- Low fees (subsidized via "piggy bank" reward system)

**Limitations:**
- **Not EVM-compatible:** Avalanche subnet requires custom tooling
- **Subsidized economics:** Long-term fee structure unclear
- **Performance not disclosed:** Throughput capacity not publicly benchmarked

**Assessment:** Focused on agent identity and access control as much as settlement. Positioning as a full-stack agent platform rather than pure settlement infrastructure.

---

### Specialized: Lightning Network (Bitcoin Layer 2)

**What it is:** A network of payment channels built on Bitcoin. Payments happen off-chain between channel participants, with only channel open/close transactions settled on Bitcoin L1.

**Strengths:**
- **Near-zero cost:** Payments within channels cost fractions of a cent
- **Instant finality:** Channel state updates are immediate
- **No blockchain congestion:** Off-chain payments don't compete for block space

**Limitations:**
- **BTC-only:** Can't settle stablecoin payments without wrapped tokens (adds complexity)
- **Channel management:** Requires liquidity locked in channels, routing can fail if paths unavailable
- **Network effects:** Requires widespread Lightning adoption for agent-to-agent payments to work

**Viable for:** Bitcoin-native applications, very high frequency ultra-low-value payments (<$0.01) where Lightning infrastructure is already in place.

**Not viable for:** Stablecoin-based agent micropayments (the dominant model for x402 and AP2 adoption), applications requiring deterministic payment success without routing failures.

---

## Settlement Requirements by Scale

Different scales of agent payment adoption require different settlement infrastructure:

### 0.01% of global API traffic → ~50,000 TPS

**What this looks like:**
- Early x402 adoption by a few dozen API providers
- A few thousand agents making micropayments
- Aggregate: 50-100K transactions per second at peak

**Viable settlement layers:**
- ✓ Base (within current capacity with headroom)
- ✓ Solana (well within capacity)
- ✓ Radius (massive over-provisioning, but works)

**Conclusion:** General-purpose L2s are sufficient. No need for purpose-built infrastructure yet.

---

### 0.1% of global API traffic → ~500,000 TPS

**What this looks like:**
- x402 becomes standard for hundreds of APIs
- Tens of thousands of agents transacting continuously
- Major API providers (Stripe, Twilio, AWS APIs) adopt micropayment models
- Aggregate: 500K-1M transactions per second at peak

**Viable settlement layers:**
- ✗ Base (theoretical roadmap capacity is 3K TPS — not enough)
- ✓ Solana (at upper edge of capacity, no headroom for spikes)
- ✓ Radius (within tested capacity with room to scale)

**Conclusion:** This is where general-purpose chains start to struggle. Radius becomes necessary for guaranteed performance.

---

### 1% of global API traffic → ~5,000,000 TPS

**What this looks like:**
- Agent micropayments become the dominant model for internet commerce
- Hundreds of thousands of agents transacting millions of times per day
- Most HTTP requests become monetizable via 402 protocol
- CDNs, DNS, package registries all collecting micropayments
- Aggregate: 5M+ transactions per second sustained

**Viable settlement layers:**
- ✗ Base (not designed for this scale)
- ✗ Solana (theoretical max capacity exceeded)
- ✓ Radius (linear scaling handles this by adding sequencers)

**Conclusion:** Only purpose-built settlement infrastructure works at this scale. General-purpose blockchains weren't designed for this workload.

---

## Decision Framework

**Start here: What's your expected peak throughput?**

| Your Expected Peak TPS | Recommended Settlement Layer |
|----------------------|----------------------------|
| < 1,000 TPS | Base, Solana, or any general-purpose L2 |
| 1,000 - 10,000 TPS | Base (if staying within Coinbase ecosystem), Solana (if non-EVM is acceptable), Radius (if EVM + guaranteed performance needed) |
| 10,000 - 100,000 TPS | Solana (accept occasional instability), Radius (guaranteed uptime) |
| 100,000 - 500,000 TPS | Radius (only tested option at this scale) |
| 500,000+ TPS | Radius (linear scaling via sequencer sharding) |

**Then ask: What else matters?**

| If you need... | Choose... |
|---------------|----------|
| **EVM compatibility** (easy migration from Ethereum/Base) | Radius, Base, Arbitrum, Optimism |
| **Stablecoin-native** (no gas token to manage) | Radius |
| **Predictable costs** (no gas price spikes) | Radius, Solana, Lightning |
| **Sub-second finality guaranteed** | Radius, Solana (usually), Lightning |
| **Maximum decentralization** | Ethereum L1, Bitcoin + Lightning |
| **Largest existing ecosystem** | Ethereum L2s (Base, Arbitrum, Optimism) |
| **Lowest cost per transaction** | Lightning (BTC-only), Radius (stablecoin) |

---

## Why Radius Exists

General-purpose blockchains optimize for flexibility, decentralization, and permissionless innovation. They support DeFi, NFTs, gaming, DAOs, and thousands of other use cases.

**Agent micropayments have different requirements:**
- Predictable sub-second finality (not "usually fast")
- Linear scalability to millions of TPS (not "we'll scale eventually")
- Near-zero predictable costs (not "gas prices vary")
- Stablecoin-native (not "bridge USDC and pay gas in ETH")
- No MEV or front-running (not "protect yourself with sophisticated strategies")

**Radius is a bet that agent micropayments become large enough to justify purpose-built infrastructure.** If agent payments remain a small niche, general-purpose L2s can handle it. If they become a dominant model for internet commerce, only purpose-built settlement infrastructure will work.

---

## Migration Path

**If you're building on Base today:**

Radius is EVM-compatible. Your contracts, wallets, and integrations work without modification. Moving settlement from Base to Radius is a configuration change:

```solidity
// Base configuration
USDC_CONTRACT = 0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913; // USDC on Base
SETTLEMENT_RPC = "https://mainnet.base.org";

// Radius configuration  
USDC_CONTRACT = 0x...; // Same USDC on Radius (bridge or native)
SETTLEMENT_RPC = "https://rpc.radiustech.xyz";
```

**You keep:**
- Same smart contracts (EVM-compatible)
- Same wallets (same address format)
- Same stablecoin (USDC)
- Same x402 protocol integration

**You gain:**
- Guaranteed sub-second finality
- Linear scalability to millions of TPS
- Predictable costs (no gas spikes)
- Dedicated throughput (no competition with DeFi/NFTs)

---

## Summary

**For agent micropayments at scale, settlement infrastructure must be:**
1. **Fast:** Sub-second finality to keep agent workflows moving
2. **Scalable:** Millions of TPS without degradation
3. **Predictable:** Costs and latency don't spike under load
4. **Simple:** Stablecoin-native, no gas tokens to manage

**Current state of the market:**
- **Conventional rails** (Stripe, Visa): Too expensive and slow for micropayments
- **General-purpose L1s** (Ethereum, Solana): Fast enough for early adoption, but not designed for internet-scale agent payments
- **General-purpose L2s** (Base, Arbitrum, Optimism): Better than L1s, sufficient for current x402 volumes, will hit limits as adoption scales
- **Purpose-built layers** (Radius, Tempo, Kite AI): Designed specifically for agent micropayments at internet scale

**Where Radius fits:** The only settlement layer with publicly disclosed performance benchmarks at 2.8M+ TPS, purpose-built for stablecoin micropayments, EVM-compatible for easy migration, and stablecoin-native with no gas token complexity.

**When you need Radius:** When your expected peak throughput exceeds 100K TPS, when you require guaranteed sub-second finality, or when cost predictability is critical for your business model.

**When you don't:** If your peak throughput stays below 10K TPS and you can tolerate occasional latency spikes, Base or Solana will work fine.