io.net deep dive: GPU aggregation, Aethir H100s and network metrics
Quote680 words74 likes
Details io.net's aggregation of consumer, datacenter and Aethir H100 capacity, with 28,000 cluster-ready and over 294,000 verified GPUs. Cites 378,000+ inferences, 1.9M onchain transactions and about $5.25M in block-reward revenue to providers.
Key claims
800 of 1,000 committed Aethir H100s are live
Over 2M IO distributed as block rewards, about $5.25M
Cluster-ready GPUs pass a Proof of Work challenge every 6h
Yesterday, I wrote about the trillion $$$ DePIN opportunity.
Driven by exponentially growing demand for GPU resources and cloud compute from an exploding AI industry.
One of the leading providers of decentralized computing power is @ionet.
By sourcing compute (GPUs) in a decentralized manner from consumer devices to data centers or GPU-as-a-service providers, @ionet aims to make GPU resources available to AI/ML startups at a competitive price.
Among the largest enterprise-grade GPU-as-a-service providers in Web3 and one of the partners of @ionet is @AethirCloud.
Just in May, @ionet announced a significant expansion of its GPU infrastructure in partnership with @AethirCloud.
In the first phase of the rollout in April, @ionet has already integrated over 300 NVIDIA H100 GPUs from @AethirCloud onto its fast-growing network.
With the expansion of the partnership, @AethirCloud and @ionet have vowed to deploy a total of 1'000 H100 GPUs, out of which 800 are currently live.
As an aggregator of decentralized computing power that also integrates other large GPU networks such as @rendernetwork or @Filecoin, the number of GPUs on @ionet is significantly higher than on these other platforms.
What are H100 GPUs you are wondering?
H100 GPUs are the latest high-performance graphics processing units (a.k.a. GPUs) from @nvidia, specifically designed to handle heavy AI and machine learning workloads.
Why does this matter?
H100 GPUs can deliver up to 756 TFLOPS (FP32 Tensor Core) of compute, providing significant resources that can be leveraged for AI model training and inference tasks.
They are suitable for running advanced AI models, including large language models (LLMs), complex deep learning frameworks or high-end scientific simulations.
If @ionet can make these GPU resources provided by @AethirCloud available to AI developers that are potentially priced out from centralized service providers such as @awscloud or @googlecloud, this can be a game-changer and finally a sustainable Web3 business model.
It's a huge opportunity to gain market share in an industry that has generated over $76 Billion in revenue in Q1 2024 alone.
Where does @ionet stand today?
Before, I said that @ionet has roughly 28'000 GPUs in the network.
More specifically, what I meant by that is that there are 28'000 cluster-ready GPUs.
What does "cluster-ready" mean?
It means that the GPUs pass the Proof of Work challenge every 6h and are readily deployable.
Additionally, there is >294'000 verified GPUs that pass Proof of Work, but are not quite ready to be deployed yet. This gives an indication of the potential that slumbers on @ionet.
To date, @ionet has processed >378'000 inferences and processed almost 1.9M on-chain transactions.
Most of these inferences come from one application: BC8.
A generative AI platform built by the @ionet to demostrate the power of its DePIN infra.
Each generated request has a Proof of Compute that proves on-chain which GPU in the network performed the job.
BC8 is one of currently six IO elements.
What are IO elements?
Basically all products that are built on top of @ionet's decentralized network of GPUs and core ecosystem components (e.g. the $IO token).
Soon, @ionet will release a dev-ready toolkit called IOG Package, which will enable anyone to build IO Elements, opening up its DePIN layer to permissionless innovation.
Why are there so many GPUs in the network?
Well, it seems to be a good business.
To date, >2M $IO in rewards have been distributed as block rewards, which amounts to roughly $5.25M in revenue for GPU providers and workers.
Note: This doesn't even include the cumulative earnings from @ionet's users, which total >$1M in stables.
While certain issues remain and the reliability of the service needs to improve, this is definitely a high potential use case and definitely one of the narratives to watch this cycle.
Once the technology fully matures and companies outside of crypto become more open to integrate Web3 services, the door to a soon-to-be trillion $$$ market is wide open.
🧠 NFA but DYOR anon!
QUOTING @zerokn0wledge_
The AI industry is valued in the $100s of Billions. The Gaming industry is valued in the $100s of Billions. 🤖 What do they have in common? Both require cloud infrastructure & computing power. An emerging trillion $$$ market. Already in Q1 2024, enterprise spending on cl
The AI industry is valued in the $100s of Billions. The Gaming industry is valued in the $100s of Billions. 🤖 What do they have in common? Both require cloud infrastructure & computing power. An emerging trillion $$$ market. Already in Q1 2024, enterprise spending on cl
4 images on X