jakiro
Author: Aylo; Compilation: Tao Zhu, Bitchain Vision Realm
Last week, AETHIR announced that their ARR (annual recurring income) was $ 36 million, which will make them among the top 20 encryption agreements in this year’s revenue.
Is I attracted your attention now?Very good, this article is worth reading.
With the rapid development of artificial intelligence towards AGI, the demand for computing resources has increased, resulting in the gap between people who can use powerful GPU chips and people who cannot use it.AETHIR is an innovative decentralized physical infrastructure network (DEPIN), which aims to realize democratic visits to cloud computing resources.
AETHIR’s establishment vision is to make it easier to access as required and more affordable. It builds a distributed network that can aggregate enterprise GPU chips from various sources.The network aims to support the increasing demand for artificial intelligence, cloud games and other computing dense applications.
In this interview, Mark Rydon, co -founder of AETHIR, sat down to discuss the unique method of Aethir.
Given the increasingly fierce competition in the recent decentralized GPU, how is AETHIR different from other participants in this field?
This is a very good question.I will answer in two parts.First of all, I will explain the problems we want to solve because this is the key to understanding.I believe that you and your audience know that there are scarcity problems in the world.Technology giants are competing for key GPU resources.This is a large -scale competition that creates more and more intelligent artificial intelligence until we reach AGI or ASI, and then everything in the world will change.
The interesting thing about this competition is that it is driven by a simple principle.If you add more GPUs and data in the ecosystem, artificial intelligence will become smarter.This is like a line from the lower left corner to the upper right corner, with the increase of artificial intelligence and intelligence.The type of GPU required is crucial.You can’t do this on consumer GPUs or low -power graphics cards.All large companies in artificial intelligence competitions use enterprise GPUs in terms of training or application.Over the past year and a half to two years, the specific model they have been looking for is NVIDIA’s H100 GPU.
The key point is that enterprises are companies that mainly have huge demand for AI computing.They are establishing business on this calculation infrastructure.Therefore, we must ask what calculations they want, what type of GPU, and what quality, performance and normal operation time they need to meet their internal indicators.This is just like a server with a high normal running time guarantee to avoid service interruption.The same is true of GPU and any computing provider; they must meet strict services, quality, normal operation time and performance requirements.
Unfortunately, most of the computing networks in the web3 space aggregate consumer GPUs.This is the easiest way to build a network -to provide tokens to people’s communities to contribute to the GPU of the GPU of the idle game.This will quickly attract a large number of GPUs and build a powerful community that is excited about token rewards.This is why many computing networks that exist today have gathered consumer GPUs.
The challenge is that if you want to have a real business, you need to sell aggregation calculations.However, the consumer GPU network quickly reached a low limit because 99.9% of the company did not want to buy calculations from consumer decentralized networks.They cannot guarantee that the GPU will not be closed at night, or the bandwidth will not be limited by family activities such as streaming media Netflix.This leads to a considerable disconnection between corporate demand and the content provided by the GPU network.
Source: layer.gg
From the first day, we have decided not to gather any consumer GPU.Each GPU connected to AETHIR is an enterprise -level, integrated through corporate network infrastructure, and is located in a data center suitable for enterprise workloads.The largest AI companies, telecommunications companies, and technology companies can use our network to do anything they need without sacrificing performance or quality.In fact, they have gained higher performance and better overall experience.
For example, IO.NET needs to overcome a large number of FUDs of its huge consumer GPU network.When they want to prove that their network can handle the actual business, they rent a corporate GPU from AETHIR.Therefore, all companies on IO.NET are provided by AETHIR.This is public knowledge in the ecosystem.
AETHIR has been committed to serving corporate customers since the first day.
There is another thing: when I explained what we do, this has confused people in the AI field.According to the definition, distributed GPU means that we do not have any GPUs.Our data center has about 43,000 GPUs, but we do not have any GPU.Among them, we have more than 3,000 H100, which is the largest H100 collection in Web3 so far, which is almost 10 times that of our closest competitors.This is why so many large AI companies use our infrastructure because we can truly provide them with services.
Some AI companies are confusing to be the importance of our so -called “common machine”.If you are conducting large -scale training, such as OpenAI or similar projects, and need 500 or more H100, then these GPUs must be at the same data center.You cannot have a H100 in Japan, one in the United States, and 200 in India.AI cannot conduct effective training on distributed hardware.This is a major technical challenge that other DEPIN has been working hard to solve. It is still an unsolved problem.I think this is a huge opportunity, but it is very complicated.
Since AETHIR has focused on enterprises from the beginning, we understand that providing services for corporate customers not only means that there is a bunch of disconnected enterprise GPUs.We also need to consider the co -ownership of the machine in our network.Therefore, AETHIR has a large number of large -scale high -performance GPU clusters.This means that we are not just a distributed network with global enterprise GPUs; our network has a large -scale co -owner cluster, enabling us to handle those large artificial intelligence work for companies that need to coexist with machines.
What inspiration created AETHIR?
So in fact, it is the cloud game network that makes us excited about the distributed GPU calculation.The team met when I lived in Beijing for about seven years.I moved there to start my first company and finally began to work on expanding the cloud game network.
To say a long time, we have an idea that we can solve the performance and scalability challenges of the cloud game network by distributing hardware in a decentralized manner.From a high level, the premise is that delay is the killer of these networks.The farther the calculation of the user is, the worse the user experience is due to the increase in the delay.This idea is that if you eliminate the motivation for centralized calculation, you can have a more distributed network.As the network becomes larger and more scattered, the possibility of users closer to calculation will increase, thereby reducing delays and improving performance.
The centralized solution focuses on focusing all resources in one position to achieve a scale economy, but this does not increase value from the perspective of user.It actually limits network performance.If you want to build a network optimized for user experience, you will distribute calculations to everywhere so that users can always approach it.We believe that if we can solve distribution and unit economic challenges, we can solve the problem of preventing services such as Google Stadia and other services.
This is our starting point. We soon realize that we have the correlation with the field of artificial intelligence and began to build products there.
Another good news worthy of attention is that the global game population is about 3.3 billion.Most of them (about 2.8 billion) use low -end equipment to play games, which means that they cannot play mainstream AAA games. Similarly, AAA developers cannot reach these players.
The most feasible solution to unlock all these capitals is to use cloud games to eliminate user hardware requirements.Just use our existing technology and make it more cost -effective to expand.Now, you have a technology that can unlock these 2 billion players, no matter where they are.This is exactly the role of building a network in a decentralized manner.
The current AETHIR game index
We are bringing a gaming experience separation from the hardware of billions of game players worldwide, and this method is fundamentally impossible to achieve.This is why I am so optimistic about the game field; this is our original vision.
Do you think the demand for decentralization calculations will increase significantly?Or do you think we have reached this point, but customers need more time to use this technology?
In my opinion, to be honest, if we talk about decentralized cloud solutions, or just talk about Aethir, then they are mainly about education.You don’t need to know that AETHIR is a decentralized cloud provider, you can cooperate with us.We cooperate well with Web2 -90% of our customers are web2 companies, and they are very satisfied with the services we provide.
Looking at the wider artificial intelligence ecosystem, where is the turning point?There are some crazy statistics.
A few weeks ago, I read a research paper that said above that according to the growth of calculation demand, by 2030, the power on the earth will not be enough to meet the computing needs of artificial intelligence.This is too crazy.
These macro numbers show that a large amount of capital is deployed into the ecosystem.These numbers are almost incomprehensible.But if you zoom in a little bit, you will find two types of computing needs: training and reasoning.Training is a process that makes artificial intelligence more intelligent, such as upgrading from ChatGPT-4 to ChatGPT-5.Reasoning is the process of artificial intelligence to complete its work, such as answering questions.
Most people like you and me use ChatGPT or other large models, right?For example, through the Microsoft ecosystem’s Chatgpt or through Google’s Gemini.Most of our interactions are carried out with universal language models from very few companies.But if we look forward to the next year, considering the industry’s index -level growth, I guess you will interact with artificial intelligence more than now.
We will soon interact with artificial intelligence in a more meaningful and agent way.Artificial intelligence will do more for us, such as booking tickets, providing assistance services and processing customer service calls.It will be much more than now.
If you pay attention to calculation upstream, unless a company only uses the ChatGPT API to create applications, they are likely to build their own AI products, which means that they have their own computing infrastructure needs.As a result, as the reasoning space increases, it will become more scattered.At present, most of the infrastructure for training comes from a few large companies.Although some companies are developing new competitors in these large language models, the explosive growth of AI applications in reasoning in reasoning will lead to more scattered computing markets.
This means that we will see a large demand for competitive prices and channels for start -ups and small corporate -friendly contracts.This may be the most imminent turning point I saw.
Another product in the AETHIR ecosystem is A-Phone.Can you introduce us in detail who is this product and the target audience?
A-PHONE builds and expands directly on our infrastructure.It uses our cloud game technology to transmit real -time rendering to the device in a low latency.This is very cool because it is completely related to access.For example, you can have a $ 150 smartphone, download the AETHIR application, and then open it to access the device equivalent to $ 1,500.All hardware restrictions on local equipment have disappeared because you have cloud functions that support the application.
You can open almost infinite applications on the AETHIR mobile phone, and will not exhaust your battery.All computing, processing and storage are completed in the cloud, basically providing you with a super mobile phone, you can call any application you need at any time.
Whether it is a game or a video conference education platform, it is very cool.It eliminates the obstacles of hardware on people’s access content, tools or practical procedures, especially for most mobile users who occupy Internet users.
What do you think is the key to the success of AETHIR so far: your technical solution, or your business development efforts?
In my opinion, as a company, we focus on two aspects.The first is the corporate element I mentioned before.This means making some very difficult decisions in the early days.As I said, it is much easier to aggregate consumer GPUs.It is much more difficult to gather these corporate GPUs that we already know that it is difficult to find and access.We took a more difficult road in the early days, which made us risk at the beginning of operation.But because of this, we have done hard work, and now the situation is better.Not many companies are determined to do such adventure in the early days, which is of great significance to us.
Secondly, we have always focused on the actual business -actual utilization rate, actual contract, and actual income.From the beginning, this concern is very important to us.This is why we choose the corporate path.We hope to make full use of Web3 technology to provide a solution to change the industry, not only the web3 solution, but also the first -class industry solutions in AI and the game.
Our business development team has played a vital role in persuading partners to join our ecosystem, especially in the early days.In terms of technology, we make the process of connecting computing resources seamless.At present, more supply than we can accept hopes to enter our ecosystem.In the future, our goal is to become an ecosystem that really does not require licenses and completely decentralized, and we will achieve this goal.But at the beginning, we must be pragmatic.Open the calculated gate and let a large number of GPUs exhaust your tokens. This is not a good business measure.
We regard ourselves as a supply -dominated organization.We always try to make more supply than demand.We don’t want to refuse demand, but we don’t want a huge gap between supply and demand.We want to develop supply and demand relationship wisely and steadily.We are not just adding unlimited GPUs to boast our numbers; this is not the right way.
We plan to issue some major announcements in the next few weeks to show our commitment to transparency.This is really interesting to people, and it shows that AETHIR is a company that people want to participate.
Can you introduce us to AETHIR tokens?How does it integrate into the ecosystem and how to generate value?
This is actually a larger version of the theme, and you will see it soon.I can’t talk about this topic anymore now, but what I can say is that because of the need to deal with tokens, many projects have been difficult to cooperate with large web2 entities before.
This is a continuous challenge in this field, and we have a very exciting and novel solution.I think people are very optimistic when they see it.Therefore, it will allow us to bring a lot of transaction volume to tokens.
Our biggest customer is web2 customers, and I think this will not change.We need to ensure that we are engaged in this business and allow the value to accumulate into the AETHIR tokens and its support ecosystems.This is our promise. I think you will see something very interesting about how to achieve this next week.
