—Introduce more high-end computing power to play the role of mining
—Sell orders that have already generated cash flow at a discount, recover the principal in advance, and play the role of de-leveraging forward
—Use its own network effect to connect customers with idle computing power
Let’s analyze Aethir’s recent efforts in this regard based on these three points:
1. Introducing high-end computing power
Aethir’s two recent developments are New horizon and in-depth cooperation with GAIB & GMI cloud.
New horizon aims to introduce more high-end GPU holders to Aethir and hold more callable machines in various countries and regions. Judging from the application process on the official website, the machines currently introduced are mainly gaming graphics cards above RTX 3080, as well as data center-level graphics cards of H100, H200 and even GB200 NVL2. From the document, Aethir labs has given the corresponding computing power reward coefficient, which is probably determined according to the current memory and floating-point performance tests on the market. The cooperation with GAIB & GMI Cloud introduced the H200 machine. GMI Cloud is Nvidia's official cloud service partner, has received investment from Supermicro, and is the first service provider in the Asia-Pacific region to obtain H200. Therefore, Aethir has also become one of the first project parties to introduce H200 into the decentralized computing network.
According to the current computing power market, H200 has a memory capacity of up to 141G and a memory bandwidth of 4.8T/s, which is nearly twice that of H100. It helps to train models with larger parameters more quickly, so AI large model companies are willing to pay a premium. From this perspective, it is more conducive to hedging the risk of falling prices of other low-end computing power.
2. Early withdrawal of principal
Aethir's decentralized cloud includes more than 3,000 NVIDIA H100s and more than 43,000 consumer graphics cards, some of which are self-owned machines. Faced with the impact of new models and the shrinking demand of AI companies, it is necessary to issue corresponding income products to recover the principal of purchasing machines in advance.
GAIB can tokenize GPUs through the operation of computing power finance, allowing them to circulate further in the DeFi system. This is essentially a win-win operation, which brings external underlying income assets to the DeFi system, and also allows GPU machine holders to recover their principal in advance and purchase new machines to expand scale advantages.
3. Network Effects
From the perspective of economic model, Aethir's high-end machine access is B2B, which requires filling out a form and signing a contract with Aethir Lab, and then Aethir Lab sets the price according to the region and model. Here Aethir can give full play to the network effect and use its own platform to find AI customers, while at the same time obtaining positive cash flow by charging service fees and price differences to GPU providers.
The good thing here is that the machine provider first needs to purchase ATH for staking. The service income generated by the machine needs to pay a 20% service fee to the Aethir network, and it will be issued in the form of ATH. The more high-end machines there are, the greater the buying order will be. Take H200 as an example. According to the formula given by Aethir lab, eight cards of H200 need to pledge 1469235.0552 ATHs, which is close to 73000u according to the market price, which basically reaches 25% of the machine's value.
Summarize
In short, the computing power industry will have a high degree of Matthew effect, and the same is true for crypto computing power projects. If there is no GPU physical business, it is fine, at least there will be no losses, but the valuation of FDV supported by the machine will return to the Meme level ; if there is a physical business, then it can only go up, use high-end machines, strive to create revenue, while locking in the selling pressure of the local currency, and recover the cost in advance through computing power derivatives, so as to fight against the computing power deflation caused by the increase in Nvidia's market value.
