# The Death of Passive Yield: Why DeFi Needs Real Work | Jun Liu - GAIB - Channel: [Ethereum Denver](https://streameth.org/ethereum-denver) - Date: 2026-03-09 - Duration: 12:44 - Watch: https://streameth.org/watch/yt-bRBxqlx92B4 - YouTube: https://www.youtube.com/watch?v=bRBxqlx92B4 ## Description πŸš€ Get Ready for ETHDenver 2026! πŸš€ We're already hard at work preparing for next year's biggest Web3 event! Keep your eyes peeled for more info on ETHDenver 2026β€”it’s going to be epic! 🌟 ## Transcript Okay. Okay. So, now we have June. I introduced you earlier. So, now actually it's your time from Gaib. Um, he's going to be talking about a new product of finance where onchain returns are built on tangible activity, not speculation. Take the stage. >> Hello everyone. Um this is Jun. I'm the co-founder and CTO of Gaib. And today I'm gonna share um share about the data passive yield and why D5 needs why D5 needs real work. Yeah. So where where the passive yields come from when we're talking about uh in in D5. So it usually represents the APY with a minimal effort and it usually comes from like two parts. First uh it comes from either uh the variable market dynamics or the token incentive. So first is like uh the lender interest income right uh like borrower paying the interest to the lenders and it's usually um fueled by the leverage or some market dynamics. you want to farm something and the most representative protocols are like a and morpho and second yields come from like uh AMN liquidity provider fee right like doing the LP um it's usually comes from like the trading fees pays by the users if they swap uh and then share with the LPS and yeah and unis swap curves are the most representative protocols and the third EOS come from the staking usually the onchain consensus rewards from different L1s. Um yeah so it actually just have new issuance of of the of their tokens or yeah as the gas tokens to the validators as the third yields. So for example like lido right you just like easily stick in to lido and it give you like native yields of ease. So these three uh fees comes from like uh like basically based on like different market dynamics because it comes from either new issuance or it comes from like users uh and the liquidity and the markets and then they actually trades on amnies from money markets and the fourth one is usually uh called like yo aggregator or yo aggregators or like strategies or like votes. like these kind of fees and these fees come from like yield farming or liquidity mining, right? Uh like these votes like people deposit the money into the votes, right? And they will be like curators or strategies, right? They create some active strategies and they optimize um optimize the yields on chain and move the money here and there and the most that like probably the most representative one is but like these kind of yields are unpredictable right because the market is like volatile and it's often subsidized by like token incentives which means it's a sufficient for long-term capital um attraction and there's know like real external cash flows. So if we want to sum it up the D5 native use uh it usually have like these four like several limitations. First is like the emission driven burnout. Let's say like uh if they if there's no like token incentive or the price of token drops a lot and it will have like dilution risk, right? No one's want to like mine like like a tokens without any uh without any value or price. So the most of the capital will just assict. And the second limitation is reflex uh reflexive uh cyclicality. It means that it has like high beta because it very hard to predict whether the market is going up or going down or uh whether another protocol is going to be popular or not. So it based on like uh it yeah highly based relies on the market condition and there's no correlation with the external economic growth. And the third limitation is the organic yield compression. It means when a landing when a pair on DAX goes mature uh it will has like less volatility. That means there will be less people to swap on it or there will be too many people's LP into that pool. That means the trading fees cannot offset the impermanent loss. Which means if you are uh an LP of that pair, you will probably have some loss if the pro uh if the if the pairs imbalanced. And the four limitation is the blackout risk. And it usually happens to um these like votes or curator curated votes because if you put money into like these curated votes, the curators will move money away, right? And then put it into different protocols and these protocols are usually built on top of different uh D5 Legos and you actually don't know like how many Legos are in that boat, right? So there will be a lot of like composability risks. Um yeah so there are like a lot of like limitations uh for defi uh native yields and it in the early stage of defi right the yields uh are probably be like double digits right like 30 or even to 50% APY so it worth these risks but right now it's just like single digits right so it probably isn't worth all the risks so how can we do that so we need to uh introduce the actual cash flow the actual yields from the real world that's why we always heard that like 2025 and 2026 will be uh the era of RWA right we have to tokenize real world assets and then put it on chain right and most of them are tokenized like us treasuries or commodities or even private credits and we can also uh because like the fat policy And it actually uh has real impacts to the real asset yields um on on both on everything like US treasuries, private credits, commodities and even equities. But that's leads to a new questions. Oh Yeah. So as as 2026 is probably the era of AI, right? So how can we get the AI yield in the real world and put it on chain? How can we earn a real world uh AI yield? So before we talk about the AI yield uh that's break AI into like the five layer cake. So this is the uh uh the layers like presented by uh by Jensen. So he he said that like AI can be broken uh can be broken down into like five layers and first layer is the power and energy is the physical foundation AI like the cooling the water the electricity the land resource the yeah these are the natural resource and artificial resource that the actual data center needs to build on top of and the second layer is the hardware layer which is the GPU chips and accelerators and the memories like these semicondar semiconductor infrastructure powering uh the all the AI data centers neo clouds and the these chips are the hardware layer and the third layer on top of the hardware is the compute infrastructure which is the neo clouds provider AI data center itself large scale like GPU clusters they actually combines these like uh hardwares into a running facilities that can actually provide compute resource to all the uh all of the AI startups, AI companies and uh independent developers and the fourth layer which is the foundation models is a large scale of like pre-trained models like different LLM um computer vision system and other like tuned models like for like different uh from from different companies and kind of like abstract the new like the role compute into the real intelligence And the fifth layer which is the uh layer that most people are familiar with is the AI applications like chat GPT cloud cloud code frontics right and open uh open uh yeah the API tokens connected by open clouds and these are uh are the applications which actually utilize the real intelligence to create a real world value and what GI is doing is focusing on the third layer uh the compute infrastructure. So what we do is we tokenize uh the enterprise level of GPUs and the rental contracts of these like neo clouds and we turning it into a yield bearing staple coins that people can just like hold it and get the yields from these data centers. So uh why is like different from like investing into like data centers because like right now for most of the data centers they are like private companies you cannot directly hold it right uh and for the stock that you can buy those are like the second uh like the second layer they are like cheap and hardwares like for example like Micron like Nvidia like AMDs these are like stocks of like producing or manufacturing uh the actual chips so it's not direct exposure to the data center and the cash flow generated by these like rental contract. So that's why we like direct uh directly tokenize the rental contract and the actual uh GPU racks uh and bring the GPU yields on chain and in this way we create uh a new type of like D5 primitive that can be composable onto other protocols because it actually generate like really high yield because uh the boom of AI. So uh so the main products we have uh there are like two main products one is aid called AI dollar and the other one is the stake AI dollar. So for AI dollar is a synthetic dollar. I will call it stable coin. I will more call it like a synthetic asset and it's backed by like US treasury and cash equivalent. You can easily m it with like USDC and USDT and other approved like stable coin and you can also like redim it at any time with at most like T+1 uh due to the redemption of US treasury. So aid is kind of like the base currency of Gaib ecosystem. Uh you can trade it as an entry point of the tokenized ADA portfolios and we are going to open another like robotics portfolio. So you will also be like purchased by aid and the stick aid is the stick version of aid. It's a 4626 volts that you basically like stick your aid to get exposed to the actual data center deal and it tracks the net asset value instead of maintaining the $1 pack. So uh you will yeah you will get the yields directly from the vault and the value reflects the performance of the underlying AI data centers uh in their rental contract to different AI companies like um open AAI and then prop. Yes. Uh if you want to get more uh details about GI you can go to the link tree and like yeah uh other things cannot be displayed. So just go to the link tree. Yeah. And thanks everyone for listening.