Interview: The Spherical Journey
Compiled & edited by: Yuliya, PANews
In right now’s more and more capital-intensive competitors inside the AI provide chain, bodily assets equivalent to GPUs and robots have gotten the scarcest and most respected manufacturing components globally. With GAIB asserting its TGE on November nineteenth , this RWA × AI × DeFi mannequin has garnered even larger consideration.
In the new “Founder’s Speak” collection of “The Spherical Journey,” co-produced by PANews and Web3.com Ventures, hosts John Scianna and Cassidy Huang invited Kony Kwong, co-founder and CEO of GAIB, to delve into the story behind the founding of GAIB, how to design a strong danger management mannequin, and the full enterprise path to turning computing energy into a tradable asset.
*Be aware: This video interview was carried out on October thirtieth, and some knowledge and updates might differ from the present state of affairs.
The Founding of GAIB – Bridging the Hole Between AI and Finance
PANews: First, might you please introduce your self and the impetus for founding GAIB? We all know you had been a enterprise capitalist at L2 Iterative Ventures (“L2IV”) and even have a conventional finance background. What expertise led you to find this distinctive downside in the present AI subject?
Kony: Earlier than getting into the VC subject, I labored in conventional finance, together with credit score analysis, fairness analysis, and conventional funding banking. Later, I developed a powerful curiosity in cryptocurrencies, so I joined a big trade to be chargeable for its abroad enlargement, main the acquisitions of different exchanges, pockets corporations, esports gaming platforms, and extra.
Later, my accomplice and I based a cryptocurrency VC fund centered on infrastructure investments, equivalent to ZK, Layer 1, and cross-chain MEV initiatives.
The true impetus for me to discovered GAIB got here in late 2023 and early 2024. At the moment, the crossover between AI and Crypto started to emerge, and everybody was exploring its prospects. As an AI fanatic, I carried out in-depth analysis on this. Since the launch of ChatGPT-3.5, I’ve devoted a big quantity of time and effort to it, even constructing an AI agent myself. Again then, there have been virtually no mature frameworks obtainable, so every part needed to be constructed from scratch, equivalent to how to vectorize knowledge, how to go looking, how to implement RAG (Retrieval Augmentation Technology), and how to construct long-term reminiscence (the context window was very quick at the time).
I found that my two favourite areas had been merging, so I began digging into the Crypto × AI observe. However at the time, most corporations in the market had been doing the identical factor:
- Decentralized computing energy markets (equivalent to Akash and Render);
- Decentralized knowledge labeling (utilizing blockchain as an incentive mechanism to encourage public participation in knowledge assortment);
- Others concentrate on AI brokers .
What shocked me was that virtually no corporations approached the subject from a “monetary” perspective . That is uncommon, as somebody with a monetary background is aware of very nicely that the most vital factor in the early levels of an business is fixing the “capital downside.” The whole AI business, particularly in AI infrastructure, is experiencing exponential development in computing energy demand, which instantly results in an exponential demand for computing assets equivalent to GPUs. This sub-sector is extraordinarily capital-intensive; constructing knowledge facilities, buying GPUs, and deploying infrastructure are all extraordinarily costly.
I assumed to myself, why do not we do one thing about this? To me, the core spirit of blockchain lies in creating new markets and new assets, which is precisely what DeFi and the “DeFi Summer season” have taught us. This led me to the concept of offering monetary providers for AI infrastructure , as this captures a big portion of the worth in the AI provide chain.
I shared this concept with my co-founder, Alex. His background leans extra in the direction of the semiconductor and AI fields; his household runs Realtek , one of the world’s high seven chip producers, and he additionally runs a cloud service firm known as GMI Cloud . He personally skilled all the ache factors I discussed: as a startup cloud service firm, securing funding is extraordinarily tough . There are two causes for this:
- First, GPUs and computing energy had been nonetheless a really new asset class at the time (one or two years in the past), and no one actually understood them;
- Secondly, cloud providers themselves are additionally very new.
So, Alex and I hit it off instantly. We believed we should always begin a monetary firm centered on AI. This has occurred in each different business earlier than, whether or not it is coal, actual property, or anything. And so, we determined to create GAIB final 12 months.
ROI of AI infrastructure and the enterprise mannequin of GAIB
PANews: Certainly, the AI subject requires large capital expenditures, and the return on funding (ROI) can take a number of years. What are your ideas on the ROI of AI infrastructure like GPUs? Additionally, how does GAIB collaborate with cloud service corporations and assist them shorten their capital turnover flywheel?
Kony: Truly, most individuals might not know that GPUs, as an infrastructure, are literally fairly worthwhile .
- First, massive AI or gaming corporations require a big quantity of GPUs and usually signal long-term contracts of 2–3 years with cloud service suppliers to make sure secure computing energy. This offers cloud service suppliers with a strong money move.
- Secondly, AI computing energy obtainable on demand is extraordinarily scarce . For instance, a couple of months in the past, ChatGPT launched a brand new characteristic that might rework photographs into particular types, which at one level prompted a computing energy scarcity even for OpenAI itself, demonstrating that the market continues to be in a state of apparent provide scarcity .
- Extra importantly, the typical return on funding cycle for GPUs is definitely 12 to 18 months . Because of this, from a income perspective, the annualized return on funding (ROI) can attain 50% to 100%, which is great in comparison with different asset courses.
Concerning the second query, who can we accomplice with? My co-founder Alex’s firm , GMI Cloud, was naturally one of our first companions and the start line for our challenge. However in the 9 months that adopted, attributable to the exponential development of the market and the surge in demand, we obtained a big quantity of collaboration requests and constructed a powerful pipeline of initiatives. At present, we accomplice with greater than 10 Neo Cloud and Nvidia Cloud companions globally, masking Thailand, Taiwan, Singapore, Hong Kong, and Japan in Asia; the United States and Canada in North America; and Norway, Iceland, and Denmark in Europe.
We are likely to accomplice with corporations which can be usually Nvidia Cloud companions . To an uninformed viewers, this implies these corporations have handed Nvidia’s vetting course of and possess the licenses and capabilities to offer {hardware} and software program providers. Extra importantly, they obtain official suggestions and preferential remedy from Nvidia in accessing the newest GPUs and high-quality clients. Due to this fact, most of our companions are Nvidia Cloud companions.
This brings us to our distinctive benefit in this subject:
- First, our founding crew are operators in this subject themselves, and we have now a deep understanding of its financial mannequin and GPU enterprise.
- Secondly, we have now a community of accomplice purchasers.
- Third, our capital options are extra versatile and higher meet their wants in phrases of each underwriting time and length.
Due to this fact, our pipeline of collaborative initiatives has been constantly rising.
How does GAIB assure returns and management dangers?
PANews: You talked about partnering with cloud service suppliers round the world, however the value buildings for electrical energy, knowledge facilities, and many others., fluctuate significantly throughout completely different areas. How do you guarantee a comparatively constant return on funding (ROI) throughout completely different companions? Additionally, you talked about partnering with Nvidia Cloud companions. Does this imply you have got a set of requirements to ensure the credibility and operational capabilities of your companions?
Kony: Sure, it’s totally tough to make sure that the ROI is totally constant throughout completely different cloud service suppliers as a result of they every have completely different charges and value buildings. For instance, the prices of electrical energy, services, and knowledge facilities in Asia are utterly completely different from these in the United States.
Due to this fact, when transacting with these cloud service suppliers, our focus isn’t on value, however on web money move restoration . We’ll assess:
- If we give somebody $100, when and how will they reciprocate with $100?
- All components, together with electrical energy prices, facility prices, and even depreciation prices that will have an effect on asset worth.
As for the transaction construction, it depends upon the particular settlement we attain with them. Generally we use a fixed-rate mannequin, requiring a hard and fast annualized fee of return. In different instances, we want to take a position instantly in the asset itself and then take a proportion of the complete revenue generated, equivalent to 50% to 70%. This mannequin offers stronger safety.
This comes all the way down to the precise construction of the transaction and our expertise. We are able to request numerous protecting clauses, equivalent to precedence in recovering our whole funding and returns earlier than the different firm distributes any earnings. Briefly, we’ll arrange numerous protecting mechanisms to make sure that our funds are repaid first.
As well as, we have now two laborious standards when doing enterprise with these cloud corporations:
- There have to be actual assets as collateral, and it have to be over-collateralized . Once we present funding, there have to be tangible assets to again it up. Extra importantly, we require over-collateralization. For instance, if the different occasion solely has a GPU price $100, we would solely present $70 to $80 in funding, that means we have now at the least a 1.3 to 1.5 instances over-collateralization ratio. This fashion, in case of any issues, even when the asset is offered at a reduction, our principal nonetheless has a reasonably secure buffer.
- The corporate will need to have current signed purchasers and a great cost historical past . This ensures that the GPUs we make investments in are usually not simply sitting idle burning cash, however are getting used successfully and producing a steady money move to repay the funds we offer.
If an organization does not meet both of these two standards, we won’t proceed with transaction negotiations.
From “spices” to tokenization, GAIB’s core philosophy
PANews: This danger management mannequin sounds very sturdy. Additionally, we’re in the identify GAIB, which appears to be associated to the well-known science fiction novel *Dune*. May you clarify the origin of this identify and your views on the function of GPUs in the AI worth chain?
Kony: Sure, we’re all followers of Dune, and the identify GAIB was certainly impressed by Dune . In truth , additionally it is an acronym for GPU , AI , and Blockchain . You might say we’re a ” World AI Infrastructure Blockchain” platform.
This analogy could be very apt. In the Dune universe, “spice” is the most treasured and vital commodity. Equally, in our AI period, computing energy is every part. Whether or not you are utilizing ChatGPT, Claude, or Perplexity, the core constructing block you are speaking about is computing energy. Due to this fact, computing energy performs a really comparable function to “spice.”
As for the place of computing energy in the AI provide chain, I like to make use of a “smile curve” to explain it. Because of this the worth is principally concentrated at each ends of the curve.
- On one finish is the software layer, as a result of they management pricing energy and customers. Nonetheless, the curve in this market is not steep sufficient but, as most apps solely began to actually revenue this 12 months, having beforehand been burning by money and not but attaining large-scale business adoption. However we’re seeing extra and extra apps emerge.
- At the different finish of the curve, on the left, is AI infrastructure, together with GPUs, knowledge facilities, and even robotics manufacturing corporations.
Regardless of how the software layer evolves, they’ll all depend on the core AI infrastructure. As I discussed earlier than, whether or not you are utilizing ChatGPT’s mannequin or Claude’s mannequin, they in the end depend upon the underlying GPU chip for energy.
I like to make use of the analogy of a Visa card: irrespective of which financial institution points your Visa card, Visa earns somewhat cash each time you make a transaction. The identical is true for GPUs; each time you name any mannequin or use any AI software, the GPU is operating, offering computing energy, and producing income . For this reason we concentrate on core AI infrastructure, which has huge enlargement potential as functions proceed to develop.
How does GAIB turn GPUs/computing energy into on-chain assets?
PANews: Financializing AI infrastructure seems like a terrific start line. So, what’s the subsequent step? How do you intend to construct an entire monetary stack on high of this?
Kony: That is a wonderful query. Internally at GAIB, we consult with ourselves as an “financial system” as a result of we act as a bridge connecting off-chain RWA with the on-chain DeFi financial system. Our course of for dealing with these assets primarily includes three steps:
- Asset digitization : We should first convert these bodily assets into digital kind. In any other case, their knowledge, asset worth, and different data can’t be mirrored and used on the blockchain.
- Asset financialization : After assets are placed on the blockchain, the subsequent step is to remodel them into helpful monetary devices or merchandise. For instance, can they generate returns? Can they be used as collateral for lending? We develop completely different merchandise based mostly on this.
- Injecting liquidity : As soon as these assets exist on-chain, in the event that they haven’t any utility or buying and selling channels, they’re primarily only a bunch of ineffective knowledge. Due to this fact, we have now been increasing the makes use of of these assets, together with integration with lending protocols, DEXs, derivatives protocols, and many others., to actually combine them into the on-chain financial system and kind a closed loop.
These are the three issues we’re doing. By way of this core infrastructure, we are able to deal with any kind of AI infrastructure asset. We began with computing energy and have confirmed this path is possible, having already efficiently tokenized assets price roughly $30 million onto the blockchain.
Now, we’re making ready to broaden into what we imagine is the subsequent massive development—robotics . In the event you assume of AI as the “mind,” then robots are the “physique” that interacts with the bodily world. Very similar to GPUs, robots have bodily {hardware} and are poised for a large transformation in their monetization fashions. The long run of robotics shall be utterly completely different from the massive robotic arms of conventional manufacturing; it is going to grow to be rather more consumer-oriented.
We lately introduced a partnership with Primech, a Nasdaq-listed firm that primarily manufactures cleansing robots. We’re exploring tokenizing these robots as a result of they’re utilizing a brand new enterprise mannequin we name “Robotic-as-a-Service” (RaaS). This mannequin permits us to have each {hardware} assets and a secure month-to-month income stream, making it an ideal match for making a product that gives customers with constant AI-related income.
AI Greenback, GAIB Token, and Ecosystem Outlook
PANews: That sounds very thrilling. You talked about integration with lending protocols, and DEX integration is comparatively permissionless and straightforward to implement. However how are you progressing with the lending market? Are you able to reveal some particular cooperation agreements?
Kony: In the lending market, we’re about to combine with Morpho and many different comparable protocols. Moreover, there are numerous sorts of lending protocols obtainable to us on completely different blockchains. For instance, Plume Chain has a lending protocol particularly designed for RWA. Due to this fact, we’re working to combine with as many blockchains as potential to make our assets as broadly relevant as potential.
PANews: Final month, some instances emerged in the NFT area involving artistic methods to unlock liquidity from “illiquid assets.” I am questioning if anybody might create an “AI technique” that leverages these tokenized AI assets to revenue from transaction charges and then reinvests the earnings into extra AI infrastructure?
Kony: That is an fascinating concept. It is one of the causes we launched an AI-powered stablecoin, or artificial greenback—which we name the “AI Greenback.” The objective of launching the AI Greenback is to make it a common “security web” masking all assets on our platform. The worth of the AI Greenback shall be backed by all the differing types of tokenized AI infrastructure assets we have launched, together with computing energy and bots.
This provides customers a unified unit that they will simply use to earn rewards, and it may be built-in into any DeFi protocol they need. Due to this fact, AI Greenback is a single gateway we offer to customers to the whole world of AI infrastructure.
PANews: How can customers earn rewards by AI Greenback? Do they should stake it in your platform?
Kony: Sure, just like different fashions. With AI Greenback, you possibly can stake it to acquire a staking certificates model. This staking certificates will constantly generate returns from the underlying computing energy and bot assets.
PANews: So, what’s the imaginative and prescient and function of GAIB’s native token in the whole ecosystem?
Kony: The GAIB token is a vital ingredient in our whole ecosystem. It isn’t only a common governance token; it has real-world utility.
As I discussed earlier, GAIB is an infrastructure platform. One of the core parts of this infrastructure is our node community, which we name the “validation community” or “node coordination community.” This community requires all tokenized GPUs to constantly run a node and report knowledge to our community to make sure that these assets are actual, functioning correctly, and producing returns.
To make sure the safety of this community, we require customers to stake our GAIB tokens. We make use of sure restaking protocol mechanisms. Because of this GAIB tokens present financial safety for the community we offer.
Secondly, the GAIB token is, of course, at the coronary heart of all incentives inside our ecosystem. Whether or not it is further earnings, further incentives, further rewards, or DeFi integration actions, all the behaviors we encourage shall be pushed by the GAIB token.
Due to this fact, the GAIB token is at the core of GAIB’s operation not solely at the technical infrastructure degree, but additionally at the governance and incentive ranges.
PANews: Lastly, we see many decentralized computing energy suppliers in the market, equivalent to Io.web and Akash, however they appear to be extra centered on Web3 cloud infrastructure. Do you assume GAIB, an organization centered on serving the Web2 market, will intersect with these Web3 initiatives in the future?
Kony: I believe their causes for current are completely different. Decentralized computing marketplaces like Akash or Io.web had been initially meant to behave as aggregators, bringing collectively numerous idle sources, whether or not consumer-grade or institutional-grade GPUs, and offering customers with a unified API to entry this computing energy.
This mannequin might certainly go well with sure customers as a result of it may be cheaper for small-scale deployments or small-scale use instances. Nonetheless, if it’s good to make large-scale deployments, equivalent to these requiring tens of 1000’s of GPUs to coach a big mannequin or to offer production-level providers, you’ll doubtless nonetheless want to speak to the massive conventional cloud corporations or rising cloud corporations we’re working with.
Due to this fact, I imagine the market is massive sufficient to accommodate their respective area of interest merchandise and providers.
