TAO Hay NEAR: A Glimpse into the AI Race in the Field of Encryption

I’m often asked about the comparison between the Bittensor protocol and NEAR. Are these two protocols competitors? Is TAO or NEAR the best crypto AI project? Which project is currently leading the AI race in the blockchain space? The answer is as complex as its own name: it depends on the specific case. People hate this because humans usually prefer simple answers to complex problems and aren’t very comfortable with nuance. In this article, I’ll try to analyze the subtle differences in an objective comparison between TAO and NEAR, ending with my personal opinion on the matter. I won’t lie to myself that this is an easy job. It won’t be. That’s also why I need to focus on certain specific, measurable aspects. Core Differences Between Bittensor And NEAR First, let’s review the similarities. Bittensor and NEAR were founded and have built AI solutions for many years, with major developments, milestones, notable improvements, and significant AI experience from the founders—both of them have worked at Google. Although both are related to AI, Bittensor and NEAR have different focuses. They’re not really direct competitors like many people still think, and in my view, these two solutions can even complement each other to some degree—I don’t like the “winner takes all” way of thinking, like a zero-sum game. Long-Term Perspective: Bittensor is building decentralized inference and training through subnetworks and incentives based on the issuance of TAO on a dedicated, permissionless L1 server. NEAR is building open-source infrastructure, supporting security, being AI-friendly, and an AI ecosystem that intersects with user-owned web and finance—operating as a versatile, permissionless L1 server. Context Viewpoint: Jacob Steeves, co-founder of Bittensor, has credible and practical AI experience, having worked as a Machine Learning Engineer at Google before 2018, been deeply involved in AI since 2015, shifted to full-time work for Bittensor in 2018, and launched the mainnet in 2021.

• Ala Shaabana, CEO and technical co-founder of TAO, also has a solid academic background, with a PhD in computer science specializing in machine learning. Illia Polosukhin, co-founder of the NEAR Protocol, has demonstrated his exceptional capability by being a co-author of the paper “Attention is all you need” in the journal Transformer, and one of the most cited works in the history of artificial intelligence, helping to build the LLM models as we know them today.
• NEAR began as an AI startup in 2017, then transitioned into a blockchain project in 2018 to meet the need for an abstract, efficient, and decentralized payment infrastructure for AI-related activities worldwide. NEAR’s mainnet launched in 2020.
• Alexander (Alex) Skidanov has a master’s degree in computer science and briefly served as a research engineer advisor at OpenAI before founding NEAR AI with Illia. Utility Viewpoint: Bittensor has an open market where anyone can provide or use machine intelligence through 126 active subnetworks and competing with each other—like a decentralized version of OpenAI’s ChatGPT, Claude by Anthropic, etc.
• Miners run models and provide outputs, validators score them, and the TAO reward distribution network is handled via the Yuma Consensus Mechanism and permissioned nodes. TAO is needed for staking and for paying network fees. NEAR owns smart-contract-based infrastructure where anyone can build AI-friendly applications for finance, certificates, and hundreds of other use cases—similar to Ethereum and Solana, for example.
• Validators maintain safe operation of the infrastructure and are paid in NEAR tokens (Ⓝ) proportionally to the amount of tokens they stake through delegation. NEAR is needed for staking, governance, and paying gas fees for each smart contract call. PMF And the Economics Surrounding TAO And NEAR Product-Market Fit Analysis: NEAR vs. Bittensor NEAR PMF According to data from Token Terminal, NEAR is one of the most widely used blockchains, ranking second by monthly active users (MAU), with over 46 million monthly active addresses thanks to several top applications. All of it can be verified on-chain. NEAR ranks only behind BNB with 50 million MAU, even surpassing Solana with 34.5 million. NEAR Intents, an intent-based chain abstraction protocol, is the fastest-growing cross-chain solution in 2025. It has been integrated by leading companies in the industry at a rapid pace and has received substantial support from the broader industry. As of the time of writing, the protocol has generated nearly $16 billion in on-chain verifiable transaction volume through over 21.55 million swap transactions across more than 35 different blockchains. In just the past 30 days alone, transaction volume reached $2 billion with 570,814 unique accounts. Product-market fit is clear, verifiable, and confirmed by on-chain data and leading companies. NEAR provides elegant solutions to real-world problems: The NEAR protocol has some of the lowest fee tiers and the fastest transaction finality times among blockchains (gas fees around $0.002 and finality in 1.2 seconds). NEAR Intents is connecting fragmented liquidity and becoming one of the easiest, fastest, and most efficient ways to exchange between different blockchains—with very low slippage and fees—which benefits users and the integrated applications.
• It also now has a Secret Exchange mode, which I’ll write about in another article, bringing privacy to the system. NEAR AI Cloud protects users through privacy and verifiable inference, allowing users to ensure they don’t leak their AI interactions to data-collecting centralized companies;
• In addition, it allows users to verify that they’re getting the correct model they paid for, not a counterfeit, downgraded version—bringing cloud computing experiences closer to running models locally. PMF BITTENSOR To me, Bittensor’s PMF metrics are somewhat more ambiguous, but we still have some reported data as a basis. As mentioned earlier, the network has 126 subnetworks and a total of 520,000 accounts. However, Pine Analytics wrote in a report that “the top 10 subnetworks control around 56% of the total emissions [TAO],” in a report dated March 23, 2026. These are data we can verify on-chain. As Pine Analytics pointed out, most of the other demand metrics are mainly made up of unverified off-chain data, self-reported by the independent subnetworks themselves. Chutes is the leading subnetwork and accounts for 14.4% of total TAO emissions. The development team claims to have more than 400,000 users, including over 100,000 users accessing via API. According to the report, these users perform 5 million requests per day, processing 9.1 trillion tokens. Chutes provides serverless inference on open-source models such as DeepSeek, Mistral, and LLaMA, ranking as a top inference provider on OpenRouter. It also has performance evaluation results that do well compared with some competing centralized solutions, but it still ranks behind the top models while charging quite high (not subsidized) rates to end users. Pine Analytics estimates the non-subsidized selling price of Chutes is $1.41 per million tokens, while LLaMA 3.3 70B Turbo charges $0.88 per million tokens and DeepSeek V3 costs between $0.40 and $0.80. Targon and Templar are the next two leading subnetworks, with interesting use cases, but their competitive advantages over similar solutions outside the Bittensor ecosystem—such as centralized providers, aggregators, or running local open-source models—are still uncertain. Economic Analysis: TAO vs. NEAR TAO ECONOMICS Interestingly, Pine Analytics’ rebuttal argument against TAO focuses mainly on the economic aspect. The vague insights about demand mentioned in the PMF section are not as important—and even arguably predictable for a project in its growth phase. Taken individually, there’s nothing particularly alarming. These issues become even more serious when we honestly look at its economic side, and Pine Analytics’ report “The Bear Case for Bittensor (TAO)” did an excellent job analyzing why. I don’t have much to add. I checked the data—these are real data—and I largely agree with the report’s thesis. Therefore, I especially recommend that you read this report. Overall, TAO has an interesting economic model to support its use cases, but the model still hasn’t been proven capable of generating enough revenue to cover network costs, largely because TAO holders bear the burden through share dilution. Moreover, this is still an unproven model for long-term sustainability, because every halving reduces subsidies and challenges the built-in capital dynamics without a clear revenue model. It’s not that Bittensor isn’t a good project. Quite the opposite. But in its current state, IT SEEMS like the price has been significantly overshot at >$300 per token, with a market cap >$3 billion and fully diluted value (FDV) >$6 billion. NEAR ECONOMICS On the other hand, NEAR’s economics are easier to assess and are steadily improving. The protocol has a tail emission rate of 2.5% aa, down from 5% in October 2025, after upgrading the protocol with more than 80% of all NEAR validators approving, including thecoding.pool.near . This supply inflation is controlled by a clearly defined fee burn mechanism: 100% of the native NEAR token transaction fees will be burned. 70% of the total gas fees are used for contract calls (30% belongs to the contract deployer). Similarly, NEAR Intents has activated a revenue-sharing mechanism, where they receive 50% of the total NEAR Intents transaction fees across all integrated user interfaces → use that fee amount to buy back NEAR tokens in various different token forms → and may burn some portion of the repurchased tokens. In short, I believe both Bittensor and NEAR are excellent projects, with bold goals and impressive achievements so far. I respect the vision of building user-owned decentralized infrastructure for artificial intelligence, and I’m very excited about everything I’ve seen from both projects.