Industries/ AI, Data & Compute/ Decentralized Compute

Decentralized Compute on PYRAX

Decentralized compute pools idle GPUs and servers from many operators into an open, permissionless supercomputer, breaking the hyperscaler bottleneck for AI and rendering workloads. PYRAX is the settlement and trust layer PYRAX Compute runs on: buyers escrow PYRX, workers are matched by VRAM and trust tier, and every job settles against a ComputeReceipt whose output is validated by the 4-rung ladder — so payment follows proven work, not promises.

Market size
$9.4B (2024)
Projection
$100B+ by 2032 · ~28% CAGR

The market

$9.4B
Decentralized physical infra (DePIN) compute
$50B+
Global GPU shortage backlog
~50%
Avg. datacenter GPU utilization
up to 80% cheaper
Cost gap vs. hyperscale cloud

Source: Grand View Research / Precedence, 2024. Figures are indicative and provided for context.

What's broken today

AI teams face long GPU waitlists and premium on-demand pricing from a handful of clouds.
Vast amounts of consumer and enterprise GPU capacity sit idle with no trustless way to monetize it.
Buyers cannot verify a remote worker actually ran the requested job on the claimed hardware.
Marketplaces without escrow expose both sides to non-payment and non-delivery.

How PYRAX transforms it

Concrete network elements mapped to this business.

PYRAX Compute compute marketplace

Buyers post jobs and the network matches workers, meters usage in compute units at a fixed 8 PYRX/CU, and settles peer-to-peer with no cloud middleman taking margin.

On-chain escrow + 4B-PYRX compute pool

A dedicated protocol pool bootstraps demand while per-job escrow locks buyer funds up front and releases them to workers only against a verified ComputeReceipt.

VRAM / trust-tier scheduler

Jobs are placed on workers that can actually hold the model in memory and meet the required trust tier, so large models never land on hardware that will fail them.

ShardedExecutor GPU pooling

A model too big for one GPU is split across a cohort of consumer cards with a member-independent combine, letting pooled RTX-class hardware serve workloads that once needed a datacenter card.

4-rung verification ladder

Attestation, redundant re-execution, and ZK proofs give buyers a dial between cheap-and-fast and cryptographically-proven, without changing how they pay.

Buildathon: dApp ideas

Ship-ready concepts for decentralized compute on PYRAX.

GPUMarket

01

Open order book where buyers escrow PYRX for GPU-hours and PYRAX Compute matches them to available workers by VRAM and price.

ComputeEscrow

ShardPool

02

Cohort scheduler that runs a single oversized model across many consumer GPUs via the ShardedExecutor.

ComputeScheduler

ProofOfWorkload

03

Job runner that returns a ComputeReceipt with a chosen verification rung so buyers can trust results from unknown workers.

ComputeReceiptZK

IdleYield

04

One-click worker client that lets home and studio GPUs earn PYRX for verified CUs when otherwise idle.

ComputeEdge

RenderSwarm

05

Distributed 3D/render farm that shards frames across the pool and settles per completed frame.

ComputeEscrow

TierBid

06

Reverse-auction contract where workers bid on jobs and stake into their trust tier for priority placement.

EVMScheduler

BurstCloud

07

Autoscaling backend that overflows spiky AI demand from a private cluster into the PYRAX Compute market with capped spend.

ComputeWASM

SlashGuard

08

Dispute contract that re-executes a challenged job on independent workers and slashes the escrow of a proven-faulty provider.

EscrowZK