Keyboard shortcuts

Press or to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Position in the AI-Payments Landscape

Several approaches exist for letting AI programs pay each other. Exfer is not the only one, but it is the only project today combining base-layer settlement, an open protocol, and a fair launch.

Four main routes

RouteApproachExamplesCritical weakness
BTC + LightningLayer-2 channels on top of BitcoinLightning NetworkOpen/close costs several dollars; weak scripting limits conditional payments
Ethereum + L2Express lanes alongside the ETH mainnet (Arbitrum / Base / Optimism)Coinbase Base + AgentKitL2 still commits to L1; L1 gas spikes propagate; fees are not statically predictable
Centralized AI-payment protocolsTranslation layer letting agents call existing human payment railsStripe Agentic Commerce, Coinbase x402, SkyfireUnderlying rails are centralized — single point of failure, geographic restrictions, KYC still applies
Purpose-built chain for machinesDesign every protocol primitive around M2M from scratchExferCold start; ecosystem is zero; no big-company backing

Each route trips on at least one thing. Which one wins? Probably more than one — they likely coexist long-term across different niches:

  • Centralized protocols (Stripe / x402) will probably capture the bulk of the mainstream market first; they're faster to integrate
  • Off-chain + big-chain combos (Lightning / L2) fit specific verticals — in-game economies, social tipping
  • Purpose-built chains (Exfer) fit scenarios that demand absolute decentralization plus predictable cost — cross-border AI agents, autonomous IoT devices

Base layer vs. application layer

The "AI x Crypto" landscape over the past two years has been crowded, but most projects are competing at the application layer — AI model marketplaces (Bittensor), AI service APIs (Fetch.ai), GPU compute markets (Render). Every one of them implicitly assumes one thing: the base-layer chain is ready.

It isn't:

  • Bitcoin's 10-minute confirmation is too slow for programs
  • Ethereum's gas wars make micropayments infeasible
  • Solana's occasional outages make autonomous systems wary of relying on it

Exfer's position is the base settlement layer — it does not try to do AI; it provides money AI can use. It is unglamorous work, but base-layer openings of this kind are rare — historically every 5 to 10 years:

EraBase-layer openingWhat flourished on top
1995TCP/IPStill the internet's substrate 30 years later
2009BitcoinAll of crypto
2014EthereumDeFi / NFTs
2020Stablecoins (USDT/USDC)Web3 payment layer
2025+M2M payment protocol?AI agent economy?

Most base-layer projects do not succeed; the ones that do tend to define the era.

Where Exfer is unique

Mapping the major options against the relevant attributes:

ProjectDecentralizedStatically predictable feesNon-Turing-complete scriptsNo premine / no VCAI-optimized
Bitcoin
Ethereum / L2Partial
SolanaPartial
LightningPartial
Stripe x402 / SkyfireN/AN/A
Exfer

Only Exfer satisfies all five today. That does not guarantee market success — technical correctness rarely does — but the position is open and unoccupied.

Costs / boundaries

  • Cold start is real. No VC money to push exchange listings, no team marketing budget, no big-name endorsements. Mainstream awareness is near zero. All "potential value" assumes "someone, eventually, wants to use it"
  • Not interoperable with existing chains. Exfer is not a BTC fork or an ETH L2 — wallets, exchanges, explorers all need new integration work
  • Bets on M2M payments becoming a real market. If the AI-agent economy never materializes at scale, or runs entirely on centralized rails, Exfer is on the wrong side of the bet

Further reading