Edge Podcast - Introduction to io.net - Revelo Intel

Edge Podcast – Introduction to io.net

In this episode of Edge Podcast, which took place on December 7, 2023, DeFi Dad, Nomatic, Ahmad, and Tory discuss DePin, io.net, trade-offs of io.net, and more. Read our notes below to learn more.

Background

  • DeFi Dad (Host) – investor at 4RC Capital
  • Nomatic (Host) –  investor at 4RC Capital
  • Ahmad Shadid (Guest) – CEO of io.net
  • Tory Green (Guest) – COO of io.net
  • io.net – a DePin protocol that helps coordination problems between GPUs

Ahmad and Tory’s background

  • DeFi Dad says that Ahmad Shadid is the founder and CEO of io.net. Tory Green is the COO and CFO of io.net.
  • Ahmad says that before io.net, he developed a machine-learning model to detect institutions trying to fill orders in the market and hide them. He scaled the model to cover 2,000 stocks and 150 cryptocurrencies.
  • Tory says that he started with a corporate career, including banking, private equity, and corporate strategy roles. He co-founded a digital publishing firm during the web2 era. He transitioned from being an investor to an operator, helping portfolio companies with sales, marketing, finance, operations, and strategy.
  • He adds that he spent almost two years as CEO & CFO of Hum Capital, a machine learning company. He gained insights into the AI space during this time. He has been personally interested in crypto since 2016.
  • Tory says that he recognized io.net as a deep-end solution solving real problems. He was impressed by the qualities of Ahmad, including intelligence, vision, and drive. He joined io.net full-time three months ago due to his belief in its potential.

What is DePIN?

  • Ahmad says that DePin refers to a collective effort to replace existing infrastructure through hardware investments and incentivization with tokens. Examples include 5G networks, routing 5G through people, and replacing traditional telecom infrastructure. Hardware participation in networks brings value to people.
  • He emphasizes the significance of real infrastructure in DePin projects. He mentions visiting a $BTC mining facility worth $1.3 billion as an example. Real assets that can be touched are valuable components of DePin projects.

Why is DePIN growing on Solana?

  • Ahmad says that Solana‘s friendly and supportive community fosters growth and collaboration. The blockchain offers tools for microtransactions, data storage, and other features. The combination of community support and useful tools makes Salana an ideal platform.
  • Tory adds that DePin represents what blockchains are made for. GPU space is dominated by oligopoly, highlighting the need for disruption. DePinprojects have the potential to bring significant changes to various industries.

Is there a GPU shortage or inefficient utilization?

  • Tory says that GPUs are in high demand due to their use in powering AI and machine learning models. Compute requirements for AI are doubling every 3.4 months, leading to increased demand for GPUs. While there is a shortage of GPUs on the surface, there are hidden sources of supply such as independent data centers and crypto-mining farms.
  • He adds that io.net aims to aggregate these untapped sources of supply to solve the problem.
  • Tory says that by bringing more supply online, io.net aim to create structural arbitrage similar to what Uber did for taxis or Airbnb did for hotels. This approach can simultaneously decrease prices by ten times and increase profits for suppliers.
  • Ahmad adds that Nvidia monopolizes the market by controlling which GPUs can be bought by current providers. They restrict access to newer models unless providers agree not to sell certain GPUs like RTX 3090 and 490.
  • He adds that decentralization allows independent entities to upgrade their infrastructure according to AI compute requirements without seeking permission from Nvidia. Upgrading infrastructure includes changes in storage, RAM, internet connectivity, etc., to meet AI demands effectively.

Solving a coordination problem among GPUs

  • DeFi Dad says that similar to how DeFi disrupted traditional finance by cutting out middlemen and pulling liquidity together, i.net aims to do the same with compute resources. It provides a coordination solution where all participants are incentivized to work together.
  • He adds that io.net identifies inefficiencies in the market, such as supply constraints and centralization of control over GPUs. By addressing these issues, io.net aims to provide a cheaper offering compared to centralized services.

Why io.net is cheaper than centralized compute providers

  • Ahmad says that hash rates and earnings vary for different coins. The most profitable coin at the moment is not guaranteed to make $1 a day. $ETH, at its peak, could generate $7 a day for two weeks before dropping in value.
  • He adds that currently, with AI apps, GPUs meeting requirements can earn around 30 cents per hour or $7.2 a day. The goal is to attract clients who book GPUs for longer periods (e.g., six months) to maximize earnings.

io.net vs legacy competitors

  • Tory says that price is a significant advantage compared to competitors, but there are other differentiators. io.net offers a choice – users can recreate an AWS-like data center with comparable performance.
  • He adds that users can bring data centers closer to their location for lower latency, providing a competitive advantage in trading scenarios. User experience is superior compared to deploying on AWS, which often involves complex processes and certifications that take months.
  • Ahmad says that io.net offers GPU speeds faster than Lambda Cloud’s 40,000 or 80,000 GPUs. Startups are switching from Lambda Cloud to io.net due to faster internet speeds and better networking performance. Small details like internet speed matter – providers with fewer servers can offer higher speeds due to less network congestion.
  • He adds that io.net aims to tap into the profitability of AI computing, similar to AWS. By attracting more GPU suppliers, the network can offer consistent profits to suppliers and grow its client base.
  • Ahmad says that clients don’t have to move everything from AWS but can benefit from lower prices, comparable performance, and improved user experience. The network’s growth could eventually lead to building cloud source data centers and additional resources.

Are there tradeoffs to using io.net?

  • Ahmad says that legacy competitors may argue that i.net lacks certain features like auto-deployment, CI/CD, API defense, etc., which they took years to develop. However, io.net is already building these features within the ML industry for training and inference purposes.
  • He adds that while there may be some tradeoffs in terms of the feature range currently available, io.net offers a superior overall experience compared to legacy providers.
  • Ahmad says that with their platform, users can easily access a wide range of GPU supply in 45 countries. No questions asked, simply click deploy and within 10 seconds, the deployment is complete. Traditional methods with AWS or other providers often involve begging for GPU access by submitting pitch decks and LinkedIn profiles.
  • He shares how as a developer, he was hurt by being asked for pitch decks and LinkedIn profiles when all they wanted was access to GPUs. Their platform aims to bring power back to engineers by simplifying the process and eliminating unnecessary requirements. They provide a supportive community on Discord where engineers can seek help from others who are deploying with them.
  • Ahmad adds that by engaging with the community and providing support, they aim to create a better overall user experience for engineers. This approach aligns with the needs and mentality of the current generation of developers who value simplicity and accessibility.
  • He adds that there is an emerging market driven by startups that are willing to invest significant amounts of money into accessing compute resources. This generation is building innovative solutions, making it crucial to cater to their specific needs and ways of working.
  • Tory mentions being part of a disruptive innovation that challenges traditional rules. He believes in offering choice without tradeoffs, providing better, faster, and cheaper solutions than what is currently available. Their platform aims to change the dynamics of the market by bringing supply on-demand.

io.net clustering is a big deal

  • Ahmad says that io.net’s technology allows for clustering 50,000 to 70,000 GPUs into a single cluster. This clustering capability sets io.net apart from competitors who typically offer individual GPU instances. The networking infrastructure built by io.net enables efficient communication between clustered GPUs.
  • He mentions the possibility of building large language models through platform if users have sufficient resources.
  • Ahmad says that the product is designed to cater to those who are building on a specific infrastructure. Targeting clients who have the financial resources but struggle with costs. He emphasizes the importance of aligning with clients in the industry.
  • He shares personal experience of demoing the product to prominent companies like OpenAI, Instacart, and Uber. He expresses confidence in targeting specific clients due to their size and potential for revenue generation.
  • Ahmad anticipates a substantial influx of money into the network as more startups join. He highlights that transactions can be made using traditional payment methods like Visa, without users needing to know it’s crypto-related. He foresees a network where money flow surpasses token emission.
  • Tory considers competitors as either suppliers or aspiring suppliers rather than direct rivals. He asserts that they are uniquely positioned to serve Enterprise AI customers.

Fees for io.net

  • Ahmad says that there are network fees ranging from 5% to 10% for various transactions within io.net.

Check Out These Important Links

Show Information

  • Medium: Youtube
  • Show: The Edge Podcast
  • Show Title: io.net: The World’s Largest GPU DePIN Built on Solana
  • Show Date: December 7, 2023