Key AI Investment Opportunities

This report analyzes the most prominent AI themed assets in the digital markets and the use cases they aim to unlock. The following are the core sub-categories of leading AI-related cryptocurrencies:

1. Decentralized AI protocols
2. DEPIN projects for computation
3. Supply chain solutions
4. AI meme coins

1. Decentralized AI protocols

• Projects included in this category are Layer 1s built to directly incentivize or facilitate the development of AI. Below are the most prominent projects aiming to do this:

1.1 NEAR (NEAR)

• NEAR’s initial vision was to enable user-owned AI by creating an open and permissionless environment for AI development. After encountering limitations with existing smart contract blockchain solutions, NEAR decided to pivot and build its own blockchain to solve these limitations. Recently, NEAR has decided to pivot back to AI to fulfill their original vision.n.
• Near’s end-goal is to achieve open source AGI. They plan on achieving this through the following steps: 
• Build the ‘AI Developer’ (teaching machines to code)
• Build the AI Researcher (use the AI Developer to teach machines to do research)
• Leverage the AI Researcher to progress towards open-source AGI
• Takeaway: Near protocol itself has established itself as a prominent Layer 1, however, we are still waiting to see the progress of their AI endeavors as the pivot back to AI has only happened recently.

1.2 Artificial Super Intelligence (ASI)

• Artificial Super Intelligence is the token representing Fetch AI, Singularity Net, and Ocean Protocol which recently merged. While the projects have an overlapping interest in facilitating AI development, their specific aims and technology differ.

1.3 Fetch AI (FET)

• A machine learning network enabling the development of a digital economy powered by AI. It aims to create the infrastructure and software necessary for developers to build connected devices and services through autonomous economic agents (AEAs) that can learn, adapt, and transact on the FET blockchain on behalf of their owners.
• Bosch and Deutsche Telecom are members of the Fetch.ai foundation and Fetch is also part of MOBI, a foundation including companies such as BMW, Bosch, FORD, and others aiming to accelerate the adoption of blockchain in their industries.  
• Despite their strong partnerships, developer activity based on 73 Fetch.ai Github repos' weekly commits is quite low as seen below:

Figure 1: Github commits FET (source: Stack.money)

• Although Fetch.ai has impressive partnerships, there is little evidence indicating any significant adoption of their technology. Considering Fetch has been around since 2019, it is difficult to see them gaining technological adoption in the future.

1.4 SingularityNet (AGIX)

• Singularity Net has two key components: AI Marketplace and Publisher, and the Domain Specific Language (DSL). The AI marketplace connects users and publishers, allowing users to purchase specific AI-related services monetized by publishers via AGIX coins. The DSL aims to create a network of AI agents that can communicate, learn from each other and complete complex on-chain tasks.
• The AGIX marketplace currently has 87 products published. Of the products published most are currently offline at the moment and have a low number of reviews (only 3 have over 100), indicating little traction of the products over the last years. 
• Since SingularityNET has been building since 2018 and has struggled to gain traction, we see it unlikely their products and technology will be adopted over the next years.

1.5 Ocean Protocol (OCEAN)

• Initially focused on building a data marketplace for secure and decentralized data exchange, Ocean expanded to provide the infrastructure necessary for AI applications. More specifically their SDK enables developers to build the following:
• Token-gated Dapps
• AI dapps: includes dapps for AI training, models, predictions and feature vectors
• Data marketplaces
• Decentralized storage of user profile data
• The Ocean marketplace only has 9 addresses that have published data on the platform. The most purchases a single set of monetized data has gotten is 99, which is a real-time ETH/USDT order book to take bid orders. 
• Apart from the marketplace, there are 4 apps highlighted to have used Ocean’s tech stack. Three of them are built by Ocean. The most impressive application is the Acentrik Market, an award-winning data exchange platform built by Mercedes-Benz for enterprises.
• Having Mercedes Benz use their technology stack to build a data exchange platform is impressive. However, apart from this, Ocean Protocol has gained limited adoption of its technology and products.

1.6 Bittensor (TAO)

• Bittensor is a mining network that aims to provide the infrastructure necessary for a decentralized AI environment to compete with big tech corporations. It follows a similar model to Polkadot’s subnets as those holding the largest staked TAO can build and monetize AI products on a ‘Subnet’. Those with a Subnet can leverage Bittensor’s network of miners who are rewarded for providing their computational resources to train AI models and deliver inference. The Bittensor network also includes validators tasked with assessing and ranking the work contributed by miners when training a model, creating a ranking system among the network’s participants. 
• Bittensor is still in its early stages, but it has a strong community. Over 80% of the supply is staked with validators, and 39 subnets have been registered. Some subnets aim to create use-cases such as text prompting, data scraping, predictions, LLMs, image generators, and 3D animation.
• It is difficult to assess its long-term competitiveness, due to the network still being in its early stages. While its model is unique and theoretically promising, it remains to be seen whether its subnets can compete with centralized tech companies. However, as Bittensor aspires to take on the big tech corporations, is listed on T1 exchanges, and was fairly launched (ie, no pre-allocated or vested tokens), it is an appealing choice from a narrative-based point of view. However, as Bittensor currently in its first halving cycle there is significant sell pressure from miners that receive roughly 7200 TAO tokens per day, resulting in roughly a 3% increase of supply in the network per month.

1.7 Qubic (QUBIC)

• Qubic is a blockchain utilizing what they call a ‘useful proof of work’. Similar to TAO, miners in the Qubic network contribute computing power towards training and developing AI. However, in the case of Qubic the computational power goes directly towards creating ‘Aigarth’, an AI running on top of Qubic, which Qubic claims is a new paradigm for Artificial intelligence. Qubic’s founder is well known in the space as he was the co-founder of IOTA.
• Since launching in late 2022, Qubic has managed to gain a significant market cap with over 450K active wallets according to its block explorer. However, there is little information regarding the state of its highly experimental Aigarth project.
• Qubic is still in its early stages and highly experimental as there is little information regarding the progress and state of ‘Aigarth’.

2. DEPIN (Decentralized GPU computation and file storage)

• With the demand for GPU resources skyrocketing, many smaller businesses cannot afford the resources necessary to maintain AI products. Decentralized GPU networks present a cost-effective pay-as-you-go alternative to major centralized cloud providers like Amazon AWS and Microsoft Azure. 

2.1 Render (RNDR)

• Redner is primarily intended to provide rendering for artists. The network is closed-sourced and permissioned. The Render team is well connected in the industry and were speakers at Nvidia’s GTC conference earlier this year.
• According to Render’s dashboard, over 34M frames have been rendered using the protocol since its inception, showing a clear use case and demand for its platform.
• Render has gained traction and usage since coming out. However, as their network of GPUs is primarily for rending art, newer projects focused on providing GPU computational resources are better positioned to capture the market for AI-related computation.

2.2 Akash (AKT)

• Akash offers permissionless cloud computation resources through a decentralized supercloud computing network. It is open-source and decentralized, focusing on AI/ML training and inference. 
• The increased demand for computing power has been reflected in Akash as well. Below are some statistics provided by Akash.

• Figure 2: Total USD spent renting compute on Akash (source: Akash)

• Figure 3: CPUs leased (source: Akash)

Figure 4: Akash active leases (source: Akash)

2.3 Aethir (ATH)

• Similarly to Akash, Aeithir is a cloud computing network that provides on-demand GPU computational resources. However, Aethir differs from Akash as the network is permissioned.
• No data was found, as access to the platform is only available to GPU operators and institutions onboarded.

2.4 Io.net (IO)

• Io.net offers on-demand cloud computing services through a decentralized network. The token is listed on major exchanges, but the founder's controversial history raises caution. There are numerous videos of him convincing people in the Arab crypto community to send him money in return for significant profits in the future. Since these allegations came to light, the founder Ahmad Shadid has stepped down.
• According to IO.net’s explorer, they have significant traction. Their data suggests over 16K cluster-ready GPUs and over 1.6K verified CPUs that are readily deployed at the moment, which is significantly higher than competitors. Their data reports over 940k hours of compute has been served on their platform since launching.
• While the reported data of their platform’s usage is impressive, the validity of this data should be further scrunitized and caution is advised due to Ahmad Shadid’s history.

2.5 Netmind (NMT)

• Netmind is a decentralized blockchain for AI training, fine-tuning, and inference. It aims to create a global network of computing power by targeting idle computers to provide resources for researchers and businesses. Netmind is also building products such as an LLM chatbot, digital human models, and an AI-powered healthcare system.
• Netmind is regarded highly by some in the industry. An OpenAI researcher praised Netmind’s Github and the Nvidia team visited their office in London.
• Despite being relatively new, Netmind has been gaining impressive traction. Some of the institutions they have partnered with to conduct joint research or provide GPU resources are the following: Kings College London, University of Edinburgh, University of Cambridge, TGO, The Ohio State University
• Currently, Netmind has 1786 GPUs available on the platform with 95% of them being utilized according to their dashboard. There are over 5000 GPUs queued to be added to the platform. Since February, the number of GPUs on the platform has grown from 240 to 1786 today as shown below:

• Figure 5: GPUs available on Netmind (source: Netmind)
• Netmind’s impressive list of partners and the growing number of GPUs indicate the project is on a good trajectory. Furthermore, Netmind’s team stands out from competitors as they appear to be highly regarded by some in the AI industry.

3. Supply chain Solutions

3.1 OriginTrail

• Originally set out to be a blockchain for supply chain, Origintrail have built a decentralized knowledge graph (DKG) to create verifiable AI solutions for supply chains, construction, healthcare, and research. Their solution aims to enhance accuracy and reliability by fetching information from their DKG to which anyone can publish knowledge. This differentiates it from the many existing AI solutions that are trained on the same datasets.  
• According to OriginTrail, they boast a highly impressive list of partners and clients. A few notable implementations and partnerships are the following: Bsi, Parity, Walmart, Scan, WFH, Google, Oracle, etc.
• Furthermore, their dashboard showing data of the usage of DKG shows impressive usage. In total, over $5M worth of fees have been spent publishing on the DKG, with current daily publishing fees ranging in the lower end of 5 figures. Below is a chart showing the amount of TRAC spent Publishing on the DKG per month:
• Origin Trail has an impressive network of partners and implementations using their technology. As their adoption is further backed by a strong trajectory of TRAC spent on their network, OriginTrail shows promising long-term potential.
• Figure 6: TRAC spent publishing (source: Othub)

4. AI meme coins

• Meme coins and AI coins have been the strongest-performing sectors in the market during Q1 2024. Betting on the sub sector of AI Meme Coins makes sense should this trend continue. 

4.1 Turbo

• Turbo is a meme token created by ChatGPT. Since launching in early 2023, Turbo has successfully established itself as the leading AI meme coin. It has been listed on Bybit and DWF labs has invested in the token.