> For the complete documentation index, see [llms.txt](https://docs.world3.ai/world3/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.world3.ai/world3/routerlink-litepaper/1.-why-a-routing-network-is-needed-now.md).

# 1. Why a Routing Network Is Needed Now

AI application development is increasingly an **operations problem**, not a model selection problem.

## Provider Dependency

Dependence on a single AI provider means provider outages become product outages. Policy changes, pricing updates, or regional restrictions immediately propagate to user-facing failures.

**Quantified Risk:** With 175+ AI provider incidents annually and average enterprise downtime costs of $35,000/hour, a single provider strategy exposes businesses to $500K+ in potential annual losses.

In Web3, this issue becomes existential. On-chain agents and automated protocols cannot pause when centralized APIs fail.

## Opaque Performance and Cost

In centralized AI systems, it is difficult to verify how requests were processed, through which path, and at what actual cost. For on-chain systems, opacity is unacceptable—execution, cost, and contribution must be publicly verifiable.

## Limited Participation

Today's AI infrastructure economy is dominated by a small number of centralized providers. Users consume. Providers control. Infrastructure contributors have little meaningful participation.

RouterLink transforms AI access into a networked, measurable, and verifiable protocol.