More than 100,000 GPUs from data centers and private clusters will be plugged into io.net’s new Decentralized Physical Infrastructure Network (DePIN) beta.
As Cointelegraph previously reported, the startup has developed a decentralized network that sources GPU computing power from geographically dispersed data centers, cryptocurrency miners, and decentralized storage providers to provide machine learning and artificial intelligence computing. power.
The company announced the launch of its beta platform during the Solana Breakpoint conference in Amsterdam, coinciding with a new partnership with Render Network.
Tory Green, Chief Operating Officer of io.net, gave an exclusive interview to Cointelegraph after delivering a keynote speech with Angela Yi, Head of Business Development. The pair outlined the key differences between io.net’s DePIN and the broader cloud and GPU computing market.
Related: Google Cloud expands Web3 launch plan with 11 blockchain companies
Green views cloud providers such as AWS and Azure as entities that own the supply of GPUs and rent them out. Meanwhile, peer-to-peer GPU aggregators were created to address the GPU shortage, but “soon ran into the same problem,” as the executive explained.
proudly presented @ionet_official exist @solana #breakpoint2023 yesterday!
Whether you’re a GPU provider or a machine learning engineer – listen to a live demo of the platform and join in Now.
Watch full video pic.twitter.com/E1XsgJLJNu
— io.net (@ionet_official) November 4, 2023
The broader Web2 industry continues to look to exploit underutilized resources to take advantage of GPU computing. Still, Green believes that none of these existing infrastructure providers are clustering GPUs in the way that io.net founder Ahmad Shadid has pioneered.
“The problem is they’re not really clustered together. They’re mostly single instances, and while they do have a clustering option on their website, the salesperson will most likely call all the different data centers to see what’s available,” Green added.
Meanwhile, Web3 companies like Render, Filecoin, and Storj provide decentralized services rather than focusing on machine learning. This is part of the potential benefit of io.net to the Web3 space, as an entry tool for these services into the space.
Green pointed to AI-centric solutions such as Akash Networks (average clusters have 8 to 32 GPUs) and GenSyn as the closest functional service providers. The latter platform is building its own machine learning computing protocol to provide peer-to-peer “super clusters” of computing resources.
Through an overview of the industry, Green believes that io.net’s solution is very novel and can cluster different geographical locations in a matter of minutes.This statement was tested by Yi, who created a cluster of GPUs from different networks and locations during live demonstration On stage at Breakpoint.
As for using the Solana blockchain to facilitate payments to GPU computing providers, Green and Yi noted that the sheer scale of transactions and inference that io.net will facilitate will not be handled by any other network.
“If you are a generative art platform and you have a user base, you will be prompted every time you make these inferences, and there are microtransactions behind it,” Yi explained.
“So now you can imagine the size and scale of the deals that are going on there. That’s why we feel Solana is going to be the best partner for us.”
Partnered with Render, the DePIN network of decentralized GPU providers, to provide io.net with compute resources deployed on their platform. The Render Network is primarily designed to obtain GPU rendering operations at a lower cost and faster speed than centralized cloud solutions.
Yi describes the collaboration as a win-win situation, with the company looking to leverage io.net’s clustering capabilities to take advantage of GPU computing that it has access to but cannot be used for rendering applications.
io.net will launch a $700,000 incentive program for GPU resource providers, and render nodes can expand their existing GPU capabilities from graphics rendering to AI and machine learning applications. This plan is for users with consumer-grade GPUs, Nvidia RTX 4090 and below hardware.
As for the broader market, Yi highlighted that many data centers around the world have large amounts of underutilized GPU capacity. Many of these locations have “tens of thousands of high-end GPUs” sitting idle:
“They’re only utilizing 12% to 18% of their GPU capacity, and they have no way to really utilize their spare capacity. It’s a very inefficient market.”
Io.net’s infrastructure will primarily cater to the needs of machine learning engineers and enterprises, who can take advantage of a highly modular user interface that allows users to choose the number of GPUs, locations, security parameters and other metrics they need.
Magazine: Beyond Crypto: Zero-knowledge proofs show potential from voting to finance
Svlook
