Exploring the Decentralized AI Computing Protocol Akash Network: Where Narrative Hype Meets Value Investment
TechFlow Selected TechFlow Selected
Exploring the Decentralized AI Computing Protocol Akash Network: Where Narrative Hype Meets Value Investment
Among the many niche sectors in CryptoAI, decentralized computing power will be a direction that simultaneously satisfies narrative-driven hype and value investment.
Navigating Web3 tides with focused insightsAuthors: Firehand, Kevin, Charlotte, Metrics Ventures
1 Introduction: AI Has Become the Core Narrative of This Cycle
The热度 of the AI sector may no longer need further elaboration—price performance alone speaks louder than words. On February 23, 2024, NVIDIA's stock price surpassed $800, with its market capitalization exceeding $2 trillion, making it the fastest company in history to grow from a $1 trillion to a $2 trillion valuation. In the crypto space, AI-related tokens have performed exceptionally well over the past few months, with leading tokens such as RNDR, TAO, and FET all achieving gains of over 3x. Every major event in the AI field has triggered rapid growth in related tokens.
AI has become the most significant technological revolution of this cycle and correspondingly emerged as the top speculative sector for capital. Blockchain and artificial intelligence are actively exploring ways to integrate, and the crypto world stands to benefit significantly from key advances in AI technology, driving strong appreciation in leading AI-related tokens. While just months ago we were still debating the feasibility of combining AI with blockchain, such discussions now seem less relevant. AI has already become the core narrative of this cycle, with market sentiment and capital enthusiasm overwhelming everything else.
In our previous analysis of the AI sector (inspired by Vitalik Buterin’s article: What Crypto×AI Sub-sectors Are Worth Watching?), we reviewed the four directions that Vitalik categorized within the AI space:
- AI as Participant: AI gaming, AI prediction markets;
- AI as Interface: Various AI applications;
- AI as Game Rule: Autonomous Agent infrastructure, zkML/opML;
- AI as Target: Decentralized data protocols, decentralized compute protocols, and decentralized AI models.
"AI as Target" represents crypto’s decentralization of AI, which is the most compelling and hype-prone narrative. In terms of real-world implementation, although these projects currently cannot compete effectively with centralized counterparts, many innovative projects with viable business logic have already emerged. Those that have established moats will become strong investment candidates throughout this cycle.
2 Sector Overview: Decentralized Compute Is the Key Direction for AI Investment
Among the various sub-sectors at the intersection of Crypto×AI, decentralized compute is the direction that best satisfies both narrative-driven speculation and fundamental value investing.
First, demand for computing power in the AI industry is growing rapidly, and insufficient supply and high costs have become pressing issues across the AI sector. On the supply side, AI GPU production is dominated exclusively by NVIDIA, with tech giants controlling AI compute resources—some used for training new models, others leased out. Highly centralized and monopolistic cloud platforms now control pricing power over compute. On the demand side, needs for model training and inference are increasing sharply. The race for larger models intensifies competition for compute, while small-scale model training and fine-tuning also require lower-cost compute. The widespread adoption of AI applications further increases demand for inference compute.
Second, among all AI sub-sectors, decentralized compute is where crypto and AI converge most tightly, with the clearest business logic. Using tokens to incentivize compute supply—or more broadly, the DePIN (Decentralized Physical Infrastructure Networks) model—has already been proven feasible in previous cycles through projects like Filecoin in the decentralized storage space. Whether it’s wrapper-style AI apps or agent infrastructure protocols, the role of tokens within those systems is often marginal. In contrast, in decentralized compute projects, cryptocurrency is deeply embedded into the core business logic, truly leveraging crypto incentives to reshape the AI landscape.
Amid the recent surge following NVIDIA’s GTC conference, decentralized compute has reached a new peak, with leading projects seeing strong price gains and a wave of new compute-related projects emerging. The current state of the decentralized compute sector is characterized by a large number of projects with similar business models, intense competition, and established leaders gaining moats in both compute supply and demand stability.
In terms of business logic, these projects are largely similar: they use cryptocurrencies to incentivize individuals and organizations with CPU/GPU resources to supply compute power, enabling SMEs and developers to permissionlessly access affordable compute. Token-based incentives allow prices to remain far below those of centralized providers. Moreover, distributed model training demands higher communication and parallel computing requirements, and since usage is shifting from training to inference, most current projects focus on distributed inference, resulting in high homogeneity.
Although the NVIDIA event boosted several GPU-themed projects, we expect this sector to consolidate and centralize further in the future, with smaller projects declining once leaders emerge. Both compute suppliers and users willing to adopt decentralized compute are scarce, and under highly homogeneous business models, resources on both supply and demand sides will naturally gravitate toward top-tier projects. Additionally, users require large-scale and stable compute capacity—excessively fragmented market structures would make it harder to compete against centralized cloud providers.
In summary, decentralized compute is a key strategic direction for positioning within the AI sector. For long-term investment, leading projects with existing moats will maintain sustainable competitiveness. Based on this logic, we believe Akash is the core asset to target in this space.
3 Akash Network: Fundamental and Tokenomic Analysis
3.1 Fundamental Analysis
Akash Network is a decentralized cloud computing platform designed to aggregate underutilized global computing resources via a peer-to-peer marketplace, creating a transparent and open market where users can post resource requests and global providers can bid in real time, thereby reducing cloud service costs. According to Messari reports, Akash offers significantly lower hardware costs compared to other cloud providers.
Founded in 2015, Akash launched its mainnet within the Cosmos ecosystem in 2020. Initially focused on CPU computing, Akash completed Mainnet 6 upgrade on August 31, 2023, officially enabling support for GPU cloud markets.
On the supply side, compute resources primarily come from data centers, miners, and consumer-grade hardware. After major blockchains transitioned to PoS (Proof-of-Stake), vast amounts of idle mining hardware became available. Akash has partnered with multiple large-scale miners to efficiently utilize these idle resources, securing nearly 500 V100-equivalent high-performance GPUs. Notably, North America’s largest Bitcoin miner, Foundry, contributed 48 NVIDIA A100 GPUs to the Akash GPU network. Meanwhile, individual PCs scattered globally hold significant untapped low-end compute capacity. Currently, the Akash network hosts over 17,700 CPUs and 258 GPUs, with numbers steadily rising. Additionally, Akash has launched targeted incentive programs, such as a $5 million pilot incentive program, to attract more compute providers.
On the demand side, Akash fosters an open-source community to attract more developers, strengthening its ecosystem moat and fueling continuous innovation. It is also actively collaborating with other decentralized AI protocols to expand service offerings and enhance competitiveness. Akash has formed strategic partnerships with two major Layer 1 decentralized AI protocols—Gensyn and Bittensor—securing substantial fixed demand and demonstrating its appeal and strength in the decentralized compute market. Since introducing GPU support in August 2023, Akash’s daily lease volume has surged significantly, accumulating over 162,700 leases to date, with daily revenue steadily increasing.
For matching supply and demand, Akash uses a reverse auction mechanism: users create orders, compute providers bid competitively, and users select suitable bids based on price and other criteria before signing leases.
In practical applications, Akash’s compute power is mainly used for data preprocessing and model inference, though recent efforts are underway to explore model training. Since August 2023, Overclock Labs has collaborated with ThumperAI to train foundational AI models, aiming to develop an open-source AI model named "Akash-Thumper," which will be shared on Hugging Face. Successfully completing such training would define workflows for distributed model training and drive greater demand and utilization of the Akash network.
With GPU availability and competitive pricing in place, the final barrier preventing developer adoption is the so-called "Crypto Barrier." Akash has implemented several measures to reduce user friction:
- Developed Cloudmos Deploy and Akash Console, allowing developers to manage instances on the network seamlessly;
- Integrated Cosmos Swap into MetaMask, enabling users to authorize AKT transactions directly within MetaMask;
- Supports stablecoin payments; Noble’s upcoming native USDC on Cosmos will further lower the entry barrier for developers.
3.2 Tokenomics Analysis
The AKT token plays multiple critical roles within the Akash ecosystem: as a staking medium to secure the network, governance participation, unit of account for leasing settlements, and benchmark for market pricing.
By staking AKT, users participate in network governance, with voting weight proportional to the amount and duration of their stake, promoting decentralized decision-making.
AKT is primarily used to pay leasing fees. Akash regulates supply and demand through mechanisms such as fee differentials (e.g., 4% fee when paid in AKT vs. 20% when paid in USDC) and a high annual inflation rate of up to 13%. A portion of inflation rewards and transaction fees is allocated to a community pool, funding public goods, incentive programs, and potential token burns, ensuring long-term sustainability and healthy value circulation.
According to Coingecko data as of March 20, 2024, AKT has a circulating supply of 230,816,799. All AKT tokens have been fully unlocked, eliminating future large-scale unlock pressures. Current supply growth comes mainly from inflation rewards, with a maximum supply capped at 388,539,008. As per Stakerewards, the current annual inflation rate remains around 15%, with approximately 133.49 million AKT staked (~57.8% of total supply), indicating relatively high staking participation.
In terms of liquidity, Akash’s primary trading activity is concentrated on centralized exchanges KuCoin, Kraken, and Gate.io. Notably, AKT has not yet been listed on Binance or other top-tier exchanges, limiting its visibility in Chinese-speaking regions. However, AKT has been listed on Kraken for some time and recently went live on Coinbase. Historical precedents show that projects like Bonk and Ondo experienced significant price revaluation and growth after being added to Coinbase’s roadmap. Following this trend, AKT’s listing on Coinbase could ignite market enthusiasm and investor interest, driving price and market cap appreciation.
4 Competitive Landscape Analysis
As previously noted, the decentralized compute sector is highly competitive, with leading projects enjoying strong moats. We identify two core competitive metrics in this space: compute supply and compute demand.
The importance of compute supply is self-evident—having more and higher-quality GPUs enables platforms to handle complex computational tasks more reliably and efficiently. In an environment of extreme compute scarcity, this becomes a key competitive advantage. Demand is equally crucial. As highlighted in Coinbase’s latest research report, despite substantial increases in decentralized platform compute supply, platform revenues have not grown proportionally, raising questions about actual market demand for decentralized compute. Supply and demand must form a positive feedback loop to accelerate ecosystem growth.
Key competitors alongside Akash include Render, io.net, and Gensyn. Akash and Render are earlier projects not originally built for AI computing—Akash started with general-purpose computing, while Render focused on graphics and video rendering. io.net was specifically designed for AI workloads. However, after AI dramatically increased compute demand, all three have pivoted toward AI. While Akash, io.net, and Render primarily target AI inference, Gensyn focuses on AI model training, developing a verification layer using probabilistic proof of learning, graph-based precise localization protocols, and incentive mechanisms to ensure computational integrity.
From a compute supply perspective, Akash currently operates 17,700 CPUs and 258 GPUs. Render does not disclose hardware data publicly. io.net boasts significantly more GPU capacity: as of March 20, 2024, it has 51,738 GPUs and 10,206 CPUs. Through partnerships with Render and Filecoin, io.net also accesses an additional 4,458 GPUs and 197 CPUs from Render, plus 1,024 GPUs from Filecoin—far surpassing Akash in both quantity and quality. However, it should be noted that io.net is attracting compute resources through highly aggressive airdrop incentives. Its GPU count is rapidly changing, and it remains uncertain how much compute will remain post-airdrop. In contrast, Akash’s compute resources stem from more stable and established partnerships, experiencing steady organic growth.
Regarding compute utilization, Render has not disclosed detailed metrics. Its business remains focused on rendering, though it has developed a compute client offering an API for third-party access to its GPU network, supporting AI inference, training, and fine-tuning. Projects currently integrated with Render include io.net, Beam, FedML, Nosana, with Prime Intellect and Exabits pending governance votes.
io.net’s overall network utilization sits around 30%-40%, with almost no usage observed from compute resources acquired via Render and Filecoin integrations.
Akash maintains GPU utilization between 40%-60%, a leading figure among decentralized compute platforms. Recently, utilization dipped temporarily due to a surge in GPU supply. CPU utilization remains consistently high at 50%-60%.
(Data source: calculated from Akashstats)
In terms of valuation, Akash currently has an FDV and MC of $1.2B. RNDR’s FDV is about 5x higher and MC around 3x higher. io.net and Gensyn have not yet issued tokens—io.net’s latest private round valued the project at $500M, but given rising market sentiment, its secondary market debut is expected to exceed this valuation significantly.
5 Conclusion
Based on the above analysis, Akash represents a prime investment opportunity aligned with the core narrative of this cycle and possesses strong long-term value.
From a narrative and fundamental standpoint, the AI sector has become the dominant theme of this cycle, with decentralized compute as its central pillar—driven both by external attention and speculation fueled by NVIDIA, and by the deep integration of crypto and AI within the DePIN paradigm. As a leading decentralized compute project, Akash is poised to be a focal point during the next wave of AI sector rotation, attracting capital due to its competitive advantages and entrenched moats.
From a capital perspective, Akash has completed full unlocking for investors and team members, avoiding massive sell-side pressure in this cycle. Over half of the tokens are staked. At current circulation and ~15% inflation, approximately 94,609 AKT are newly minted daily. Assuming a price of $5, this translates to only ~$500K daily inflationary pressure—manageable and unlikely to cause significant downward pressure. Furthermore, AKT’s listing on Coinbase on March 20 opens access to the U.S. market, improves liquidity, and enhances credibility. With no Binance listing yet, upside potential from future exchange listings remains high.
Going forward, continued monitoring of Akash’s ability to attract compute resources and expand customer relationships is essential. Risks to watch include weakening competitive positioning against peers like io.net and Render. Particularly, io.net has attracted massive compute supply through anticipated airdrops. As a project originally rooted in CPU computing, Akash’s ability to sustain GPU resource expansion and secure consistent use cases and clients will be key to maintaining its edge.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
@MetricsVentures
