CapIX Whitepaper: AI Capacity Exchange

Spare compute and AI capacity is scattered across idle machines, reserved GPU windows, lab hardware, and inference endpoints. CapIX helps providers turn that supply into usable infrastructure. Providers list available compute or model routes, buyers take a CPX-priced quote, and the selected capacity becomes redeemable infrastructure.

Spare supply -> quote -> CPX settlement -> usable capacity

Sellers are compute providers, GPU hosts, CPU lane operators, labs, model endpoints, and teams with reserved capacity they are not fully using. Buyers are researchers, builders, startups, agents, eval pipelines, data teams, and applications that need infrastructure without waiting on conventional cloud procurement.

Supply: spare compute and inference / Demand: builders and workloads / Price: CPX

Smart Route is CapIX's proprietary job-routing system. Instead of forcing an entire script, prompt, or batch job onto one expensive route, it evaluates task shape, price, availability, latency, reliability, and compatibility, then returns the cheapest reliable route mix that can complete the work.

Workload -> Smart Route -> lower-cost execution path

A seller brings spare compute, model-serving capacity, or reserved provider time to the exchange. CapIX verifies that the supply can be delivered, converts it into a CPX-denominated listing, and makes it visible to buyers. When the listing fills, the seller earns CPX for delivered capacity.

Verify supply -> publish listing -> earn CPX on fill

A buyer compares live asks by price, route quality, and delivery type. Direct compute buyers receive secure machine access after payment. Research and batch workloads can use Smart Route to reduce cost before execution. Inference buyers receive streamed model output after the selected route is settled. In every case, the quote becomes the receipt for delivery.

Compare ask -> settle CPX -> receive capacity

CPX is the exchange settlement rail. Buyers pay CPX to redeem capacity. Sellers earn CPX when their capacity fills. The public experience stays simple: take a quote, approve payment, receive capacity, and keep a clear receipt of the transaction.

Pay CPX -> provider earns -> buyer receives capacity

CapIX prices capacity by hardware tier, reliability, latency, available block-hours, route quality, and visible demand. The current beta shows CPX/hr for compute and CPX/request for inference. Smart Route adds a second layer by comparing the cost of running the whole job on the highest required lane against a split across cheaper compatible lanes.

Market quote = tier + utilization + reliability + latency

The current release is a launch preview of the CapIX market loop. It demonstrates provider listings, buyer quotes, CPX settlement, Smart Route pricing, compute delivery, and streamed inference. The next phase expands provider verification, capacity guarantees, market depth, and production-grade access controls.

Launch preview today; deeper provider guarantees next