The shift to cloud infrastructure has become a massive trend for the internet. But moving generative AI and gaming applications to the cloud can strain computing resources.
That’s why Singapore-based Aethir is focusing on decentralization — of the sort made possible through blockchain and peer-to-peer networking — as a way to speed up games and AI.
One of the ways it does this is to take graphics processing units (GPUs), which are in short supply, from across a whole decentralized network of infrastructure companies and then more fully utilize them to accelerate computing tasks at hand. The company aims to save 60% on costs.
Today, Aethir is also announcing that it is participating in Nvidia’s Inception program, which has helped support thousands of AI startups over the years. Mark Rydon, CEO of Aethir, said the company wants to blaze trails in Decentralized Cloud Infrastructure (DCI) for gaming and AI. By joining the Inception program, Aethir can propel its mission of revolutionizing the gaming and AI industries.
GamesBeat Next On-Demand 2023
Did you miss out on GamesBeat Next?Â Head to our on-demand library to hear from the brightest minds within the gaming industry on latest developments and their take on the future of gaming.
While this is a cool story for gaming companies, it’s also quite interesting as a computing story.
The company’s primary goal is to bring joy to gamers and facilitate swift, secure, and cost-effective interactions with artificial intelligence from anywhere in the world. Their flagship product, Aethir Atmosphere, introduces a groundbreaking concept — a GPU-as-a-Service ecosystem for infrastructure providers while simultaneously democratizing a new Web3 cloud for enterprises.
This innovation aims to offer the experiences of the future to users today, Rydon said. With the support of Nvidia’s accelerator, Aethir will have a platform to connect with venture capitalists and increase its visibility through exclusive events, potentially attracting additional investment and partnerships.
“Joining Nvidia Inception is an important part of our company’s evolution. Nvidia’s unrivaled expertise in GPU technology and deep learning will be an invaluable asset as we continue to innovate and push boundaries in the cloud gaming and AI verticals,” said Kyle Okamoto, CTO of Aethir, in a statement. “This collaboration not only accelerates our technological trajectory but also firmly positions us within an elite network of forward-thinkers and industry leaders. We’re geared up for a future where Aethir and the other members of Inception can set new industry benchmarks together.”
How it works
Addressing a pressing issue in current cloud technology, Aethir offers an infrastructure alternative that solves the cost and latency challenges associated with delivering premium GPU computational loads to latency sensitive industries.
Aethir offers an alternative infrastructure design that becomes faster and cheaper as it expands. Aethir said its infrastructure solution can deliver low-latency, GPU resources via the cloud.
Rydon has been working infrastructure for computing at companies in China and learning about the challenges of scaling cloud infrastructure. At the beginning of 2022, he started a company and raised a round of funding. The company raised a $9 million round at a $150 million valuation. It has scores of engineers in a variety of places including China.
“We build a decentralized cloud infrastructure. To understand what we’re doing, you really need to understand maybe what the current cloud infrastructure looks like,” Rydon said. “The current cloud infrastructure is very CPU focused, or storage focused. That’s great for running enterprise software. It’s great for storing files and data and photos. But what it’s not very good at is delivering GPU resources, in a cost effective way. It really hasn’t had to do that before.”
Even Google hits some hurdles trying to provide outstanding game services via its GPU-based cloud infrastructure. GPUs are in short supply and they’re expensive.
“What we’re essentially doing is solving for the issue using our decentralized infrastructure to essentially rethink cloud infrastructure from the ground up, kind of with the benefit of a rearview mirror,” Rydon said. “We’re rethinking the deployment of that infrastructure, literally, almost from a boots on the ground infrastructure perspective.”
Aethir creates a decentralized network of infrastructure providers and then becomes a broker for a kind of GPU marketplace. This enterprise-grade brokerage can meet the supply-and-demand requirements of some of the biggest game studios in the world, Rydon said. It deals with the real-time problems of gaming and AI inferencing.
Normally, cloud infrastructure is more centralized. If you’re going to supply computing to a lot of cloud gaming subscribers around the world, you create a bunch of data centers and centralize computing resources in those data centers. For enterprises, that meant a lot of CPUs and storage. But with gaming and AI, graphics processing units are the primary computing resource.
“The issue here is that when you care about latency, when you care about performance, the distance that your users are away from that data center has a major impact,” Rydon said. “So edge computing is a way that people solve that. If you put the computational capacity together, then you get it as close to the user as possible. And this is very necessary for things like cloud services in the autonomous driving sector and certain parts of the IoT.”
But he noted that the edge of the network is very expensive and it requires a massive operation, like the older content delivery networks, which offloaded videos to many locations so servers wouldn’t crash during the Victoria’s Secret fashion show.
“What we have with our decentralized infrastructure network is the ability for different enterprises to expose their GPUs to our infrastructure,” Rydon said. “There’s no ecosystem incentive for this hardware to be under one roof. So immediately, you start to mimic an edge computing infrastructure. Because, all of a sudden, different providers are popping up all over the all over the map. And you’re getting this organic coverage, and organic scalability that is incredibly difficult to replicate in traditional cloud and incredibly expensive to replicate in an edge scenario.”
That is a rethinking of the distribution of infrastructure. Rather than concentrate resources at the edge, it mimics the edge with a network of companies that collectively can provide the right resources at the right time as demand surges. As the network gets bigger, the efficiency goes up, Rydon said.
A game like SEA’s Free Fire has 180 million monthly active users. Since it dominates a lot of the infrastructure, it’s hard for rivals to move into the same region. But Rydon said that a decentralized network could compete against the market leader’s infrastructure by aggregating GPUs from multiple providers who come in when the demand and price is right. It’s not unlike the validators who ensure correct computing results in a blockchain network.
As for the focus on gaming, Rydon said. “We have to have a value proposition. We’re a young company, we have to dial in to a specific space. At the moment, we’re dialing into gaming. We love the democratization that this technology allows the gaming sector to experience.”
If a company wants to get more GPUs and it can’t get them, it runs the risk of being out of luck as it needs more computing resources. But if it can tap its neighbors in the decentralized network, then there is more capacity in the market that it could purchase on the fly.
“We are actually built as a solution for that problem,” Rydon said.
And if there is a recession that means a lot of the GPUs in one company are underutilized, that company could farm them out.
Rydon said the company is scaling in Southeast Asia where growth is pretty big right now. The company will be deploying maybe 15 nodes across the region in the new few months, enabling the equivalent of 100,000 monthly active users onto its ecosystem. Then it will become easier for game companies to deploy in places like Indonesia or Malaysia. Rydon said the company has a number of users for its infrastructure in Thailand, Hong Kong, Singapore, Latin America and Brazil. It will also launch in Japan and South Korea soon. In providing computing resources where they are needed, Aethir takes a cut.
Overall, Aethir has more than 10 games on its network with more than 200,000 users.
GamesBeat’s creed when covering the game industry is “where passion meets business.” What does this mean? We want to tell you how the news matters to you — not just as a decision-maker at a game studio, but also as a fan of games. Whether you read our articles, listen to our podcasts, or watch our videos, GamesBeat will help you learn about the industry and enjoy engaging with it. Discover our Briefings.