Compiled and edited by TechFlow
Participants: Ahmed Shadid, io.net Founder and CEO; Tory Green, io.net Founder and COO; Aline Almeida, io.net SVP of Engineering; Mahar Jilani, io.net Chief Experience Officer; Gauvrav Sharma, io.net CTO; Basen Oubah, io.net Co-founder and CoS; Hashim Alnabulsi, IO.VENTURES Managing Partner
Guests: Avery Ching, Co-founder and CT of Aptos Labs; Toly Yakovenko, Co-founder of Solana; Lily Liu, Chairman of Solana Foundation; Hidde Hoogland, CEO of DCENT
Podcast source: io.net
Original title: IO Summit 2024 Keynote
Air date: June 4, 2024
Summary of key points
At the 2024 IO Summit, io.net announced three major breakthroughs, launched the world's largest AI computing network IOG (Internet of GPUs), and demonstrated cooperation with Aptos and Solana. These innovations will revolutionize the future of AI, computing, and encryption technology.
Ahmed Shadid, founder and CEO of io.net, shared the latest achievements of io.net at the summit, emphasized the importance of computing power, and put forward a bold vision: to make access to GPU computing power as simple and seamless as accessing the Internet. io.net is committed to building a computing-centric ecosystem that provides a set of tools and services for building, training, and deploying machine learning models on the chain at scale.
As important partners of io.net, Aptos and Solana provide high-performance blockchain infrastructure and decentralized computing capabilities, respectively, to support io.net in expanding its AI computing network globally. Through decentralized cloud computing solutions, io.net fills the current gap between supply and demand of computing resources and provides lower-cost, higher-performance computing services.
Additionally, io.net launched the BC 8 project, the world’s first fully decentralized end-to-end machine learning model that leverages io.net’s IOG infrastructure for training and inference. io.net’s investment arm IO.VENTURES is also driving the development of machine learning startups by providing GPU computing power in exchange for equity and token allocations.
io.net: Three breakthroughs that will revolutionize the AI computing industry
Ahmed Shadid, founder and CEO of io.net, opened the summit by introducing the mission and vision of io.net and expressing his excitement about the team's achievements. He emphasized that io.net brings together the best engineers in the world to push the boundaries of technology and enter areas that no one has ventured into. Together with his team, he is ready to disrupt the AI computing industry and announced three major breakthroughs that will change AI, computing, and cryptocurrency.
In today’s digital world, computing power is the life force that drives technology forward, powering AI that is changing our lives, reshaping industries, and enabling innovators to create a world we have yet to see.
Ahmed asked a simple yet profound question: What if accessing GPU computing power was as easy and seamless as accessing the Internet? This is the background of the birth of Internet GPU (IOG). IUG is the world's largest AI computing network, and Ahmed believes that the world needs fair, democratic, and permissionless access to AI computing.
He further pointed out that in today's digital age, computing power is not only an asset or commodity, it is the equivalent of digital oil. China's spending on semiconductors has exceeded its spending on crude oil, indicating the importance of digital infrastructure. Computing power currently consumes 70% of the funding for startups.
io.net is building a computing-centric ecosystem that provides a set of tools and services for building, training, and deploying machine learning models at scale on-chain, all enabled by the power of Aptos and Solana. Users can deploy models for inference in real time, or train and fine-tune models efficiently. In short, Io represents the currency of computing power, just as the US dollar is the standard for all transactions, Io coin unlocks computing resources globally. io.net hopes to become the universal computing exchange medium for large AI startups and enterprises.
Ahmed also mentioned that about 90% of the world's computing resources are idle, scattered in consumer devices, data centers, and crypto mining farms. io.net's goal is to utilize this idle computing power and release huge economic potential.
Aptos and io.net collaboration: driving the integration of AI and blockchain
Avery Ching, co-founder and CEO of Aptos Labs, presented the vision for the collaboration between Aptos and io.net, highlighting the potential of io.net to use computing power as currency to create a new way to interact with AI products and services. The Aptos Network is an ideal partner for io.net, as Aptos, built on the Move language, provides a more secure, faster, and higher-performance blockchain infrastructure, allowing io.net to focus on innovation rather than dealing with blockchain infrastructure issues.
Aptos has demonstrated its ability to process over 25,000 transactions per second and has processed 2 billion transactions in a single day, which is 13 times more than the volume supported by the Visa network. Aptos also plans to scale to over 1 million transactions per second. With this parallel architecture, Aptos is a partner that io.net can trust to expand its IR offering in the future.
Beyond scalability and speed, both Aptos and io.net recognize that the future of AI will inevitably collide with web 3 and decentralization. Maintaining open access and decentralized content ownership is a critical part of the future of AI. Decentralized physical infrastructure networks like io.net are laying the foundation for ensuring AI products and services are accessible around the world. At the same time, both parties are committed to setting new standards for responsible AI.
Aptos and io.net are working together to build data linear use cases such as Proof of Compute and explore how blockchain can provide a transparent and immutable record of assets and works developed by AI. This collaboration provides an opportunity for developers to accelerate innovation at the intersection of AI and Web3, and Avery said they are very much looking forward to building the future with io.net.
Leading a new era of AI computing through decentralized cloud computing
Tory pointed out that current computing demand is 2.5 times the supply, and AI computing demand doubles every 3.4 months, which makes the gap between supply and demand continue to widen.
He believes that traditional cloud computing methods can no longer meet this demand, but this also brings huge opportunities. The total market value of the world's three largest centralized cloud service providers (Amazon, Google and Microsoft) exceeds $7 trillion, and a large part of their value comes from the rise of AI. The demand for computing is so high that China imported more chips than oil last year. Computing is called the "new oil" and its cost is also high. For example, Chad GPT costs more than $700,000 a day to operate. These costs will continue to rise as OpenAI launches its new text-to-video generation platform Sora.
Tory noted that despite the continuous innovation in AI, computing resources are in short supply. io.net fills this gap with an innovative solution called decentralized cloud computing. io.net's infrastructure is valued at over $3.5 billion and has more enterprise-grade GPU clusters than Amazon AWS, Google Cloud, and Microsoft Azure. io.net provides on-demand computing services at a fraction of the cost of traditional cloud service providers, aiming to bring AI to users around the world in a better, faster, and cheaper way.
Ahmed said that io.net’s success is inseparable from the support of the giants in the technology and web 3 industries. io.net is backed by teams such as Hack VC, MultipleCoin Capital 6 MV, OKX Ventures, Aptos Labs, Solana Ventures, M13, and many well-known leaders including Solana founder Anatole, Aptos founders Mo and Avery, Animoca Brands’ Yat Siu and Sebastian (co-founder of The Sandbox). In addition, many senior leaders of the world’s top GPU manufacturers support io.net. With this long list of top VCs and angel investors, io.net has raised more than $40 million in an oversubscribed Series A round, furthering their mission and execution. In its first 10 months, io.net has become a unicorn company, and this is just the beginning.
Solana and io.net: Building a decentralized future for AI computing together
Anatoly, Solana co-founder and CEO of Solana Labs, reflects on the Solana ecosystem’s journey with io.net.
Anatoly mentioned that io.net won the Solana grizzlython hackathon and presented their MVP at the Solana Breakdown conference. io.net chose Solana as a partner because Solana is fast, low-cost and scalable. Solana processes tens of millions of transactions per day with fees less than half a cent and has more than 1.5 million active addresses.
Anatoly emphasized that AI is one of the fast-growing industries and io.net needs a strong blockchain partner to support its expansion. Solana’s scalability and Solana Pay’s seamless payment capabilities will help io.net succeed globally.
Solana Foundation President Lily pointed out that in the rapidly developing Web3 world, finding a platform that not only supports growth but also accelerates growth is critical for pioneers like io.net. Solana is not only a blockchain, but also a launch pad for cutting-edge applications and fast-growing companies, with powerful projects such as Orca, Jupiter, Mediora, and Backpack. Solana provides unparalleled support to ensure that starters have the tools they need to succeed from day one.
Lily also mentioned that the Solana Foundation works closely with teams like io.net to drive innovation through events, hackathons, and direct support. io.net is actively involved in many of Solana’s initiatives and works closely with partners like Colosseum to build the world’s largest decentralized AI computing network, redefine the status quo, and democratize AI.
Achieving sustainable development of AI through decentralized computing networks
Ahmed stressed that io.net is the world's largest AI computing network and treats computing as a currency, so sustainability is crucial here. io.net promotes a healthy ecosystem by rewarding IO workers who contribute computing power, allowing AI developers to get the computing power they need seamlessly and cheaply, while incentivizing them to work hard to build new models and lead the cutting-edge technology. The network now belongs to everyone, relying on a super cluster attack protection system called IOG Cortex, which is deployed on each IO worker node.
Ahmed compares the io.net network to an ant colony, where all worker nodes work in coordination like ants to complete tasks without interruption. The IOG framework enables engineers to use open source computing power, which means that the Python code you run on your laptop can automatically run in the GPU Internet. This network is not only able to provide services to more than 100 million users with extremely low latency, but also coordinates tasks distributed across thousands of GPU clusters in 139 countries, creating a cloud service for ML model inference.
io.net uses ray.io, an open source core technology without which many AI-based products such as Chat GPT, Uber, Instacart, etc. would not exist. io.net takes this technology to a new level, building the world's most advanced, scalable and secure distributed computing framework.
io.net: Improving AI application distribution and training efficiency through IOG
Aline Almeida, senior vice president of engineering at io.net, discussed how the io.net platform distributes AI and Python applications across a broad network of GPUs and enhances these capabilities through IOG (Internet of GPUs).
io.net forked Ray to run better in a decentralized network and provide a more user-friendly experience for application libraries. IOG introduces backend customizations to best integrate with io.net's unique network, including a specialized Outer Scalar and the Io cloud that supports graphs and scheduling algorithms. io.net also improved the dashboard to provide developers with Io-specific operational metrics and geo-node details.
IOG simplifies the distribution of AI applications through direct interfaces with the Io cloud and code templates, making AI expansion not only feasible for engineers at large companies, but also open to AI creators around the world. io.net's clusters can scale to thousands of GPUs, providing powerful computing power, supporting decentralized agents and enabling low-latency model inference, from confidential computing and zero-knowledge operations to AI model training and tuning. This is an open source journey, and global collaboration will shape the future of decentralized AI.
Aline also showed how to use the IUG package in practice to train a computer vision classification model and tune its hyperparameters to improve accuracy. By connecting to an IOG super cluster with 1000 GPUs, io.net can complete 1000 training runs of different configurations in 1 hour instead of 41 days. This shows the power of distributed computing, which is not only faster but also 90% cheaper.
Finally, Aline announced that the IOG SDK Python package will be released to the public next quarter and invited interested users to join the beta waiting list.
io.net: AI cloud computing platform that redefines user experience
Mahar Jilani, Chief Experience Officer of io.net, shared how io.net leads the future of Web3 through innovative design and builds the world's largest AI computing network. He emphasized that technology alone cannot shape the future, but people are the ones who really shape the future, and design is the core of connecting people and technology. io.net is committed to providing a high-quality user experience that enables engineers to easily deploy the world's most advanced clusters.
io.net’s Io Cloud platform has 100 times more GPUs than major crypto market players and is almost on par with Google Cloud, Amazon AWS, and Microsoft Azure. Engineers can deploy a cluster of 20,000 GPUs in 90 seconds, providing an end-to-end solution from model training to inference.
Mahar demonstrated how to deploy a cluster using Io Cloud, where engineers can choose powerful options such as Ray, Mega Ray, and Kubernetes, and select the GPU model and connection speed as needed. io.net also supports clusters for Apple chips, enabling millions of Apple device owners to contribute computing power and earn revenue. The whole process is simple and fast, with no KYC required, users only need to make their selections, and io.net will handle the rest.
io.net’s cloud services are up to 80% cheaper than major players in the market, providing the same performance, quality and reliability at a significantly lower cost.
io.net: Turning GPUs into digital assets for global AI computing
Gauvrav Sharma, CTO of io.net, introduces io.net’s revolutionary platform that aims to democratize GPU computing power and make it a global standard. io.net’s IOG (Internet of GPUs) platform allows GPU owners to connect their devices to the Internet, turning GPUs into valuable assets that generate a steady stream of revenue.
Whether it is a large data center, mining farm or professional miner, anyone with an underutilized GPU can join io.net and become the digital oil that drives the AI revolution. io.net breaks traditional boundaries and empowers users around the world through the revolutionary principles of Web3. Gauvrav compares io.net's GPU network to an ant colony, where each GPU is maximized like an ant, and you can join this community of workers no matter where you are.
Gauvrav shows how to connect and monitor devices. Users can connect a new device through an intuitive process, select the operating system (Mac, Linux, or Windows) and device type (GPU or CPU), then run a script to download and install Docker and all drivers, and finally run the Docker command to connect the device. On the Devices page, users can monitor device activity and reputation scores (driven by online time), and monitor and withdraw accumulated earnings on the Earnings and Rewards page. Users can connect a Solana wallet to io.net to claim rewards and view the computing tasks a device has participated in and the earnings earned.
BC 8: Leading a new era of 3D content creation through decentralized AI
Basen Oubah introduced BC 8, the world’s first fully decentralized end-to-end machine learning model that leverages io.net’s IOG (Internet GPU) infrastructure for training and inference. BC 8 is a paradigm of decentralized AI, using Aptos as a payment rail to handle inference compute fees, royalty fees for model owners, and proof of computation records, generating 500,000 images and reaching 2 million transactions per day.
BC 8 supports third-party model integration and is hosted on io.net’s decentralized cloud. BC 8’s journey began with a model driven by stable diffusion technology, trained on a 10 TB dataset, setting a new benchmark for text-to-image realism, and has been adopted by tens of thousands of community members for generating character images for social media identities.
BC 8’s latest innovation is generating 3D models from text that can be used in AAA game engines such as Unreal Engine 5, Unity Engine, CRI Engine, and Frostbite, opening up a world of possibilities for creators. Users simply visit the BC 8 AI website, upload an image or enter a prompt, and a 3D model is generated in seconds, with the option to improve quality.
In the future metaverse, whether it’s gamers looking for new weapon skins or Blender designers conceiving unique shapes, BC 8 will meet these needs and exceed expectations. BC 8 is a community-driven platform that provides a level playing field for content creation and truly shapes the digital future. Basen also mentioned that AV from Aptos will further introduce their collaboration.
Aptos: The blockchain backbone that powers decentralized AI and game development
Avery Ching describes how io.net deployed its generative AI product BC 8 on the Aptos blockchain. Every inference transaction is recorded on the Aptos blockchain, processing seamless payments and providing proof of computational record. Since its launch, BC 8 has processed nearly 500,000 transactions per day, demonstrating the advantages of Aptos in speed and scale. BC 8 is just the beginning, and future new machine learning models will generate high-quality 3D models and AAA game assets, redefining the game development industry.
The Aptos blockchain provides instant, secure, and scalable computational proofs for these operations in the background. Avery emphasized that this is just one of many use cases supported by Aptos, and there will be hundreds of similar use cases in the future. The Aptos ecosystem will provide a set of tools and services for building, training, and deploying machine learning models on the chain, supporting payments for hundreds of thousands of on-chain models, making Aptos the backbone of the Internet's GPUs. Each product will be supported by its native token on Aptos, driving the growth and adoption of all on-chain AI products.
io.net’s physical infrastructure: Etch Data Center
Hidde Hoogland introduced io.net’s Etch data center, part of the world’s largest GPU distributed computing network. She walked us through the server room, which can accommodate up to six cold-aisle containment units with a total of 120 racks. The Etch data center primarily utilizes GPU computing power leased by the Web3 and io.net platforms.
The data center has 2 MW of redundant power connections, with each cold aisle containment unit connected via power strips under the raised floor. Perforated floor tiles between racks allow fresh air to reach the GPUs. These clusters are filled with GPUs leased through the io.net platform, both to serve customers and generate revenue.
Hidde also demonstrated a battery backup room that can keep the data center running for at least 2 hours in the event of a local grid outage. The battery system is charged by the data center’s solar roof, providing enough energy to make the facility net neutral in the summer.
io.net’s investment arm: IO.VENTURES promotes the development of AI startups
Hashim Alnabulsi, managing partner at IO.VENTURES, introduced IO.VENTURES as the investment arm of io.net, dedicated to the growth of machine learning startups. AI startups typically spend more than 80% of raised funds on computing power, and even smaller projects can consume more than $500,000 per month, which seriously hinders the development of AI. io.net proposed a better solution.
IO.VENTURES’ proposal is to provide GPU computing power to startups in exchange for equity and token distribution, an approach that is more economical than traditional venture capital and returns a lot of value to GPU suppliers. With io.net decentralizing AI computing, IO.VENTURES aims to promote the development of the Io economy by investing in fast-growing AI and machine learning startups.
Hashim also introduced an even bolder plan: the creation of the IO.VENTURES Mega GPU Fund. Unlike traditional funds that rely on a small number of limited partners, IO.VENTURES has access to the entire io.net network, unlocking $3.5 billion worth of computing power. This grassroots movement connects the world's idle GPU supply with innovators, unlocking unlimited potential while democratizing AI.
IO.VENTURES aims to expand globally through decentralized investing and focus on projects that challenge the status quo, especially those that align with its vision and want to transform the industry. Hashim calls on the world's bravest AI talents to join this transformational journey. IO.VENTURES plans to connect with partners around the world, including accelerators, research institutions, large enterprises, and progressive public sector projects, to create an unprecedented supportive ecosystem.
The future is both challenging and exciting for IO.VENTURES, and they look forward to meaningful collaborations with groundbreaking projects to embark on this transformative journey together.
io.net: The dream of opening a new era of AI has come true
Ahmed Shadid expressed his excitement about the dream come true with io.net. A few months ago, io.net was just a dream, but today, it has become a reality and has grown faster than anyone expected.
Today’s event announced several major product releases: IO Coin, IO Cloud, IO Worker, and most importantly, the introduction of the Internet GPU (IOG), which will empower all of us to build and create a better world. Ahmed also showed a preview of the upcoming IOG Python framework, which will be released later this year.
He called on everyone to join this revolution, visit io.net immediately, and change the world together.
