Original title: Metrics Ventures Research Report | Akash Network: A Decentralized AI Computing Protocol Combining Narrative Hype with Value Investment

1 Introduction: The AI sector has become the core narrative of this cycle

The hype surrounding the AI sector perhaps needs no further elaboration—its price performance speaks volumes. On February 23, 2024, NVIDIA’s stock price surged past $800, with a market cap exceeding $2 trillion, making it the fastest company in history to double its market cap from $1 trillion to $2 trillion. In the Crypto arena, AI-related tokens have performed exceptionally well in recent months, with leading tokens like RNDR, TAO, and FET all reaping gains of over 3 times. Every significant event in the AI field leads to a surge in related tokens.

AI has become the most significant human technological revolution of this cycle, consequently becoming the prime target for speculative investment. Blockchain and AI are actively exploring integration possibilities, benefiting the Crypto world with advancements in AI technology and driving up prices of related leading tokens. Perhaps a few months ago, we were debating the feasibility of combining AI and blockchain, but now, such discussions seem less important. AI has already become the central narrative of this cycle, overshadowing everything else with market sentiment and enthusiasm.

In our previous analysis of the AI sector (starting with Vitalik’s article, what are the sub-sectors worth paying attention to in Crypto × AI?), we reviewed the four directions in which Vitalik subdivided the AI sector:

• AI as Participants: AI games, AI prediction competitions
• AI as Interfaces: Various AI applications
• AI as Game Rules: Autonomous Agent underlying protocols, zkML/opML
• AI as Goals: Decentralized data protocols, decentralized computing power protocols, and decentralized AI models

“AI as Goals” represents Crypto’s decentralized transformation of AI, which is the most attractive and speculative narrative. While it may not yet pose a strong competitive threat to centralized businesses in terms of practicality, it has birthed many innovative projects with viable business logic. Among them are projects that have formed a moat and will become strong investment choices in this cycle’s trend.

2 Overview of the Track: Decentralized Computing Power is the Core Direction for Positioning in the AI Sector

In the diverse landscape of Crypto × AI, decentralized computing power stands out as a direction that simultaneously fulfills both narrative hype and offers substantial investment value.

Firstly, the AI industry’s escalating demand for computing power is a pressing concern, with shortages and high costs posing significant challenges. On the supply side, the production of AI GPUs is monopolized by NVIDIA, with major players controlling AI computing power distribution. This centralized control extends to pricing through highly centralized and monopolized cloud platforms. Meanwhile, on the demand side, the need for both model training and inference is rapidly growing, fueled by competitions intensifying the demand for computing power. Additionally, smaller-scale model training and fine-tuning require cost-effective computing power support, while the widespread adoption of AI applications further increases the need for computing power for model inference.

Secondly, among the myriad of AI sub-sectors, decentralized computing power represents the most closely integrated and logically coherent convergence of Crypto and AI. The utilization of token incentives to stimulate computing power supply, or the broader concept of DePIN, has already proven its feasibility in previous stages of decentralized storage tracks like Filecoin. Unlike Wrapper-type AI applications or Agent underlying protocols, where tokens play a relatively minor role, decentralized computing power projects exhibit a deep bond between cryptocurrency and the entire business logic, effectively leveraging Crypto’s incentive mechanisms to reshape the AI landscape.

Following the recent NVIDIA conference, decentralized computing power has experienced a surge in momentum, with leading projects witnessing significant increases in value and a host of new computing power-related projects emerging. The current characteristics of the decentralized computing power track include a multitude of projects, similar project business logics, intense competition, and leading projects establishing moats in computing resource supply and demand stability.

In terms of business logic, these projects share fundamental similarities: they utilize cryptocurrency as an incentive mechanism, allowing suppliers with CPU and GPU resources to offer computing power. This enables small and medium-sized enterprises to utilize computing power provided by lessors without requiring permission, resulting in significantly lower computing power costs compared to centralized providers. Furthermore, the increasing demand for decentralized model training, which places higher demands on communication and parallel computing, is transitioning from training to inference. As a result, current projects are predominantly focused on distributed inference, exhibiting a high degree of homogeneity.

Although the recent NVIDIA conference has spurred growth in GPU-related projects, we anticipate further aggregation and centralization in the future development of this track. After leading projects establish themselves, smaller projects are likely to decline in the medium to long term. Both computing power supply and users willing to adopt decentralized computing power are scarce resources. In highly homogeneous business logic environments, resources at both ends of supply and demand are likely to prioritize leading projects. Additionally, users require large-scale and stable computing power guarantees, making an overly fragmented track layout less conducive to competing with centralized cloud service providers.

In summary, decentralized computing power emerges as a focal point for positioning in the AI sector. With strategic investment in the medium to long term, leading projects that have already established certain moats will maintain sustained competitiveness. In this context, we believe Akash represents a core target for positioning in this track.

3 Akash Network: Fundamentals and Token Economic Analysis

3.1 Fundamental analysis

Akash Network is a decentralized cloud computing platform designed to integrate underutilized computing resources globally by providing a peer-to-peer marketplace. It aims to establish a publicly transparent market where users can freely publish resource demands, and global resource providers can bid in real-time, thus reducing the cost of cloud services. According to a report by Messari, Akash offers significantly lower costs for the same hardware compared to other cloud providers.

Akash was founded in 2015 and launched its mainnet on the Cosmos ecosystem in 2020. Initially focusing on CPU computing, on August 31, 2023, Akash Network completed its sixth mainnet upgrade, introducing support for a GPU cloud marketplace.

The supply of computing power primarily comes from data centers, miners, and consumer-level computing power. With many mainstream blockchains transitioning to Proof of Stake (PoS), a significant amount of idle mining power has become an unresolved issue. Akash Network is effectively utilizing these idle resources by collaborating with multiple large-scale miners. It has already secured a substantial amount of high-performance computing resources, equivalent to nearly 500 V100 GPUs. Notably, North America’s largest Bitcoin miner, Foundry, has added 48 NVIDIA A100 GPUs to the Akash GPU network. Additionally, there is a considerable amount of underutilized low-end computing power in globally distributed personal computers. Currently, Akash Network boasts over 17,700 CPUs and 258 GPUs, with this number steadily increasing. Furthermore, Akash has launched specific incentive programs, such as a $5 million pilot incentive program, aimed at attracting more computing power providers to join the platform.

On the demand side, Akash Network is actively building its open-source community to attract more developers. This initiative not only strengthens the ecosystem’s moat but also fosters continuous innovation and development momentum for the platform. Additionally, Akash is actively seeking collaborations with other decentralized AI protocols to expand its service range and enhance platform competitiveness. Currently, Akash has established partnerships with two major decentralized layer-one protocols, Gensyn and Bittensor. This collaboration not only brings a significant amount of fixed demand to Akash but also demonstrates the platform’s attractiveness and strength in the decentralized computing power market.

Since the introduction of GPUs in August 2023, Akash’s daily leasing volume has seen a significant increase, accumulating over 162,700 leases to date. Consequently, the daily revenue generated from leasing is continuously growing.

In matching supply and demand, Akash employs a Reverse Auction mechanism, where users create orders and computing power providers bid on these orders. Users then choose suppliers based on bid information and complete the selection process, signing leases thereafter.

In terms of specific business operations, Akash Network’s computing power is primarily used for data preprocessing and model inference, with recent attempts and developments in model training. Starting in August 2023, Overclock Labs partnered with ThumperAI to train basic AI models, ultimately aiming to create an open-source artificial intelligence model called “Akash-Thumper” and share it on Huggingface. If the model training proceeds smoothly, it will define the process of using distributed computing to train models and attract demand for the Akash network, increasing its utilization.

With the availability of GPUs and competitive pricing, the only remaining barrier preventing developers from using Akash is the “Crypto Barrier.” Akash Network has implemented a series of measures to reduce user difficulty:

• Developing Cloudmos Deploy and Akash Console allows developers to manage instances in the network seamlessly.
• Integration of Cosmos Swap on Metamask enables users to authorize AKT transactions directly on Metamask.
• Support for stablecoin payments, with Noble soon to launch Cosmos-native USDC, significantly lowers the entry barrier for developers.

3.2 Analysis of Token Economics

The AKT token plays several key roles within the Akash ecosystem: serving as a staking medium to enhance network security, governance, settlement unit for leasing, and benchmark for market pricing.

By staking AKT, users can participate in the network’s governance, with voting weight determined by the amount and duration of tokens staked, thus facilitating a decentralized decision-making process.

AKT is primarily used for paying leasing fees, with Akash employing different fee rates (e.g., a 4% fee for AKT payments and 20% for USDC payments) and up to a 13% annual inflation rate to regulate supply and demand. Additionally, a portion of inflation and fee revenues are allocated to a community pool for public funds, incentive measures, and potential token burns, ensuring the ecosystem’s sustainable development and value circulation.

According to Coingecko data, as of March 20, 2024, the circulating supply of AKT is 230,816,799 tokens, with all AKT tokens having been fully unlocked, thus no longer facing significant unlock pressures. The primary source of current circulation increase comes from inflation incentives, with a maximum supply of 388,539,008 tokens. According to Stakerewards data, the current annual inflation rate remains around 15%. Approximately 133.49 million AKT tokens are staked, accounting for approximately 57.8% of the total supply, indicating a relatively high staking ratio.

In terms of liquidity, Akash’s main liquidity is concentrated on KuCoin, Kraken, and Gate, three central exchanges. It’s worth noting that AKT has yet to be listed on major exchanges like Binance, which has led to relatively low visibility in the Chinese market. However, AKT has been listed on leading U.S. exchange Kraken for some time and is already listed on Coinbase. Historical cases show that projects like Bonk and Ondo often experience price rediscovery and significant growth after being included in Coinbase’s roadmap. Based on this trend, AKT’s listing is expected to stimulate market enthusiasm and investor interest, thereby driving its price increase and market value growth.

4 Analysis of Competitive Landscape

According to the previous discussion, the competition in the decentralized computing power field is fierce, with leading projects having a strong moat. We believe that the two core competitive indicators in the decentralized computing power field are: computing power supply and computing power demand.

The importance of computing power supply is self-evident. Having a greater quantity and higher quality of GPUs enables more stable and efficient processing of increasingly complex computing tasks. In a situation where computing resources are extremely scarce, this becomes a stronghold for platforms. Computing power demand is equally important. A recent Coinbase research report pointed out that although the supply of computing power in decentralized platforms has increased significantly, platform revenue has not increased at the same rate, casting doubt on the market demand for decentralized computing. Computing power supply and demand will act as positive feedback loops, driving rapid growth in the entire ecosystem.

In this field, projects that are in a leading position alongside Akash include Render, io.net, and Gensyn. Among them, Akash and Render were established earlier and were not originally designed for AI computing. Akash was initially used for general-purpose computing, while Render was mainly used for video and image rendering. Io.net is specifically designed for AI computing, but with the increase in computing power demand brought about by AI, these projects have all shifted towards AI development. Compared to Akash, io.net, and Render which focus on AI inference, Gensyn specializes in AI model training. Gensyn is attempting to establish a verification layer that ensures the correctness of calculations through probabilistic learning proofs, graph-based precise positioning protocols, and incentives.

From the perspective of computing power supply, Akash currently has 17,700 CPUs and 258 GPUs. In comparison, Render has not disclosed hardware data publicly, while io.net has more GPU computing power. As of March 20, 2024, io.net has 51,738 GPUs and 10,206 CPUs, and has also partnered with Render and Filecoin, gaining access to 4,458 GPUs and 197 CPUs from Render, and 1,024 GPUs from Filecoin. The quantity and quality of computing resources far exceed Akash’s, but it is worth noting that io.net is attracting computing resources through highly attractive airdrop incentives, and the number of GPUs on the platform is rapidly changing. We still need to observe how much computing power will remain on the io.net platform after the airdrop ends. In contrast, Akash’s computing power comes from more stable and cooperative relationships, and the computing resources have been steadily growing.

Regarding the usage of computing power, Render has not disclosed relevant data yet, but its current business focus remains on image rendering. Additionally, it promotes usage in the field of artificial intelligence by establishing a computing client, providing an API that allows other projects to access Render’s GPU network to support AI inference, training, fine-tuning, and other use cases. Projects currently integrated with Render include io.net, Beam, FedML, Nosana, Prime Intellect, and Exabits, which are currently under consideration.

The overall network utilization rate of io.net is around 30%-40%, with the computing power accessed from Render and Filecoin being almost unused.

The GPU utilization rate of the Akash network remains around 40%-60%, which is a relatively leading level among decentralized computing power platforms. However, in recent days, there has been a temporary decline in utilization due to a significant increase in GPU computing power supply. The CPU network utilization rate also remains at a high level of 50-60%.

(Data source: Calculated based on Akashstats)

From a valuation perspective, Akash currently has an FDV and MC both around 1.2 billion USD. RNDR’s FDV is approximately five times that of Akash, while its MC is about three times higher. io.net and Gensyn have not yet issued tokens. io.net recently completed a Series A financing round with an estimated valuation of 500 million USD. However, with the market sentiment towards io.net on the rise, the price of secondary tokens after opening is expected to far exceed this value.

5 Conclusion

Based on the analysis, Akash is set to seize the central theme of the current cycle and offers a promising long-term investment opportunity.

In terms of the big picture and basic aspects, the AI sector has emerged as a key focus in this period, with decentralized computing being a pivotal area within it. This focus is driven by factors such as Nvidia’s influence on computing power and excitement surrounding the integration of crypto with AI through the DePIN concept. Akash stands out as a leading player in decentralized computing. In the upcoming shift within the AI sector, decentralized computing will surely be a significant area of interest, and Akash is well positioned to attract substantial investment due to its competitiveness and strong market position.

Financially, Akash has completed all investor and team token unlocks, reducing concerns about significant selling pressure during this cycle. With over half of the tokens staked, the daily unlocked AKT supply, considering a 15% inflation rate, amounts to approximately 94,609 AKT. At a price of $5 per token, this translates to around $500,000 in daily inflation-driven selling pressure, which is relatively moderate. Additionally, with the Coinbase listing on March 20th, Akash has gained exposure to the US market, enhancing liquidity. The fact that AKT has not yet been listed on Binance suggests there is still significant anticipation for further exchange listings.

Looking ahead, it will be essential to monitor Akash’s efforts in attracting computing resources and expanding customer relations. Particularly, there is a need to remain cautious about risks arising from Akash’s potential lack of competitive advantage compared to similar projects. Projects such as io.net and Render pose strong competition to Akash, especially considering io.net’s ability to attract significant computing resources due to its airdrop expectations. Given Akash’s origins in CPU computing, its ability to sustainably expand its computing resources on GPU and maintain stable use cases and clients will be crucial in competing with other projects.

Disclaimer：

1. This article is reprinted from [Metrics Ventures], Forward the Original Title‘Metrics Ventures Research Report | Akash Network: A decentralized AI computing power protocol that coexists narrative hype and value investment’, All copyrights belong to the original author [Firehand, Kevin, Charlotte]. If there are objections to this reprint, please contact the Gate Learn team, and they will handle it promptly.
2. Liability Disclaimer: The views and opinions expressed in this article are solely those of the author and do not constitute any investment advice.
3. Translations of the article into other languages are done by the Gate Learn team. Unless mentioned, copying, distributing, or plagiarizing the translated articles is prohibited.