This week marks a major milestone for Nodexo.
In its weekly update, the decentralized GPU compute marketplace on Bittensor announced that it is rolling out Proof of GPU v3, a foundational upgrade that reshapes how compute is validated, how providers get paid, and how the subnet evolves toward Nodexo 2.0.
Rather than incremental changes, this release represents a structural shift toward verifiable, demand-driven decentralized compute.
A Quick Recap: What Nodexo Does
Nodexo is a decentralized GPU compute marketplace built on the Bittensor network’s subnet 27.
It allows GPU owners to contribute hardware and earn rewards based on verifiable performance, while users can rent on demand compute without relying on centralized cloud providers.
Validators benchmark GPU performance using Proof of GPU, ensuring that real hardware is doing real work.
The result is a network that turns GPU compute into an on-chain commodity, connecting global supply and demand for AI workloads.
Proof of GPU v3 Launches This Week
Proof of GPU v3 goes live with subnet version 2.4.0, introducing a new rolling, recipe-based proof system designed to continuously verify GPU compute. The key improvement is speed.
Compute capability is now validated in near real-time, typically within one to two minutes removing much of the latency that previously limited scale and participation.
All miners and validators are required to update, with a short window of expected network disruption during the rollout.
What Proof of GPU v3 Changes
Proof of GPU v3 is not an AI service or model hosting layer. It is a neutral proof system for pure GPU compute, focused exclusively on bare metal GPUs and virtual machines: No hosted models, no prompt APIs, and no opaque abstractions.
Under the new system:
a. GPU providers run short-deterministic tasks in the background,
b. Cryptographic proofs of compute are generated continuously, and
c. Validators verify and score performance in near real-time.
This ensures that compute claims are verifiable, timely, and publicly auditable.
Off-Chain Compute, On-Chain Trust
One of the most important changes with Proof of GPU v3 is architectural.
GPU providers remain fully off-chain. They install the Nodexo agent, register hardware, accept jobs, and get paid directly by users. Proof of GPU anchors run transparently in the background.
Validators no longer act as payout mechanisms. Instead, they publish trust signals based on verified performance.
From the provider perspective, Nodexo now behaves like a modern GPU marketplace rather than a mining protocol.
Emissions are Burned, Demand Takes Over
With Proof of GPU v3, protocol emissions are burned.
GPU providers earn from real-user demand, not subsidies. The subnet, Subnet 27, transitions from being a paymaster to functioning as a trust oracle, identifying reliable providers and filtering out unreliable ones.
This shift reduces incentive gaming and aligns rewards with actual usage.
Ethereum-Backed Verification is Coming Next
Proof of GPU v3 is designed to anchor compute proofs to Ethereum data availability.
Proofs are aggregated off-chain, batched, and published to Ethereum, making the compute record public and difficult to tamper with. This enables independent verification by validators, enterprises, and external observers.
Ethereum anchoring is a key component of the Nodexo 2.0 roadmap.
The Bigger Picture: Nodexo 2.0
Proof of GPU v3 is the foundation for Nodexo 2.0, which is still in development.
The long term vision is a pure GPU compute cloud offering bare metal and virtual machine access, backed by public proof and direct payments. Future upgrades include confidential compute using TEE (Trusted Execution Environments) and deeper enterprise readiness.
Why This Weekly Update Matters
This week’s release on Subnet 27 is not just a technical upgrade.
It represents Nodexo’s shift away from subsidy-driven growth toward a model built on verifiable compute, real demand, and transparent trust.
Proof of GPU v3 lays the groundwork for decentralized GPU infrastructure that can scale without sacrificing integrity.
Be the first to comment