NEAR Protocol AI：Nightshade 分片与 Agent Market 如何构建链上自主交易基础设施

Entering 2026, one of the most notable narrative shifts in the crypto market is the acceleration of the fusion of AI and blockchain—moving from “proof of concept” to “infrastructure deployment.” In this wave of narrative change, NEAR Protocol has locked its core strategic anchor on a more specific direction: making AI agents true users of the blockchain, and endowing them with the ability to hold “data sovereignty” and “transaction autonomy” on-chain.

On February 4, 2026, NEAR AI released version 26.2, providing users with anonymous access channels to ChatGPT 5.2, Claude Sonnet 4.5, and Gemini 3 Pro Preview for the first time. The next day, NEAR officially launched the NEAR AI Agent Market—a decentralized marketplace where AI agents can autonomously bid, execute tasks, and settle directly with NEAR tokens. Agents can scan the market for tasks that match their capabilities, submit bids and delivery times, and be selected by requesters to complete transactions—throughout the process with no human intervention. This sequence of actions quickly placed NEAR at the center of discussions in the “AI-blockchain infrastructure” race.

As of May 8, 2026, according to Gate market data, NEAR is priced at 1.4824 USD, with a 24-hour decline of 1.03%. Its market cap is approximately 1.919 billion USD, and the circulating supply is 1.292 billion tokens. Over the past 7 days, it is up 14.84%; over the past 30 days, it is up 11.30%. The price range over the past 52 weeks is 0.8624 USD to 3.3700 USD, and the historical high price is 20.42 USD. Market sentiment indicators show “neutral.” The gap between this price range and the increasingly heated AI narrative around NEAR forms the core tension in current market discussions.

From AI research to building a chain for AI

NEAR’s AI narrative did not appear out of thin air in 2026. To understand the current strategic layout, you need to trace back to the deep “genetic core” of this public chain.

The identity of NEAR co-founder Illia Polosukhin is itself the starting point of the narrative logic—he is one of the co-authors of the foundational paper “Attention Is All You Need,” which established the Transformer architecture that serves as a foundational technology for large language models such as ChatGPT. In April 2026, Polosukhin made it explicit at the Buidl Asia 2026 conference: “NEAR Protocol was originally an AI project, and only later developed blockchain infrastructure to facilitate data collection and reward participants in AI model training.”

This “genetic” factor determines the narrative logic of NEAR in a fundamentally different way from other public chains. It is not “adding AI capabilities on top of blockchain,” but “using blockchain to serve the AI economy.” In March 2026, Polosukhin further elaborated in an interview: “The users of blockchain will be AI agents. AI will be in the front end, blockchain will be the back end.” In this vision, AI agents interact directly with the protocol—executing payments, managing assets, coordinating services, and even participating in governance voting.

Below is a compilation of key facts along the timeline of the NEAR AI strategy from late 2025 to May 2026:

Coexistence of on-chain activity growth and valuation gaps

On-chain usage data: growth trend established

From verifiable data, NEAR’s on-chain activity has shown a steady growth trend in the first half of 2026. As of April 26, 2026, NEAR’s number of daily active addresses (7-day moving average) has stabilized at around 600,000. As the core channel for AI agent interactions, NEAR Intents’ cumulative transaction volume exceeded 10 billion USD on January 16, 2026. It has processed more than 15.7 million cross-chain transactions, generating over 17 million USD in fees, and has integrated 28 blockchain networks. Within the past 30 days, NEAR Intents recorded transaction volume of 2.15 billion USD, coming from 541,075 unique addresses.

From the perspective of DeFi total value locked (TVL), the data presents two different measures. On one hand, some data shows NEAR’s on-chain TVL reached about 350 million USD in December 2024, a 120% increase from the previous year. On the other hand, other data shows TVL has fallen from a prior peak of about 500 million USD to around 100 million USD. This split in itself indicates that AI agent-driven on-chain interactions are forming a value-creation logic distinct from traditional DeFi—higher transaction frequency, more dispersed amounts per transaction, and more stringent requirements for low latency and low fees—yet its ability to accumulate liquidity has not been fully validated.

Financial data and valuation references

From a revenue perspective, NEAR Protocol generated approximately 15.6 million USD in protocol revenue in the first four months of 2026 (about 12 million NEAR tokens), while the total protocol revenue for the entire year prior to 2026 was roughly 10 million USD. This means that revenue in just the first four months of 2026 has already exceeded the entire previous year by more than 50%. Analyst Michaël van de Poppe expects that NEAR’s annual revenue by the end of 2026 could rise to between 40 million and 60 million USD. NEAR’s current price-to-sales ratio is about 34x, lower than Solana’s 40x, and far below Ethereum’s 200x.

However, price data reveals a different picture. According to Gate market data, as of May 8, 2026, NEAR is quoted at 1.4824 USD, with a market cap of approximately 1.919 billion USD. Compared with the 52-week high of 3.3700 USD within 2024, this price has retreated by more than 56%. Compared with the historical high of 20.42 USD, the decline is even more pronounced. Even if there is a rebound within a short-term window—up 37.81% over the past 90 days (from a 0.9341 USD low rebounding to a 1.5478 USD high)—the market’s overall pricing is still in a “narrative-led, price-lagging” stage.

Structural advantages: Nightshade sharding and the Intents architecture

The core requirements of AI agent scenarios for a public chain are high concurrency, low latency, elastic scalability, and privacy-preserving transactions. NEAR’s Nightshade sharding technology offers a differentiated solution. Its central design assigns network transactions to multiple parallel shards; by increasing the number of shards, it achieves an approximately near-linear increase in network throughput. The dynamic sharding mechanism allows the network to automatically scale the number of shards up or down based on real-time load, combined with a “stateless validation” design that significantly lowers the operational threshold for nodes.

According to information released by NEARCON 2026, the Nightshade 3.0 upgrade will support functions such as consensus and execution separation, privacy sharding, and privacy-intent transactions. Meanwhile, NEAR has joined the NVIDIA Inception program, focusing on verifiable, privacy-preserving AI inference within trusted execution environments (TEE). This technological direction closely matches AI agents’ needs for privacy computation. As a new transaction format, NEAR Intents allows users to express intents in natural language; a solver network computes the optimal execution path, making cross-chain operations completely hidden from end users.

Of particular interest is that comparing NEAR with the high-performance public chain Solana helps clarify the differences in their technical positioning. On May 5, 2026, the Solana Foundation partnered with Google Cloud to launch Pay.sh, a payment gateway that allows AI agents to call API services on a per-use basis using stablecoins on Solana. If Solana Pay.sh is viewed as the “payment layer for AI agents,” then what NEAR builds is the “intelligence and data sovereignty layer for AI agents”—the former solves “how to pay,” while the latter solves “how to think” and “how to protect data sovereignty.”

Narrative divergence: from “AI infrastructure” to “user-owned AI”

NEAR is the “economic layer for AI agents”

This is a comparatively optimistic narrative logic. Its core argument is: as AI agents become independent participants in economic activity, they need a set of non-sovereign, programmable infrastructure with low trust costs to carry out payments, identity verification, and asset custody. NEAR’s Intents architecture and Agent Market are well-aligned with this need. When an AI agent needs to outsource a computation task, it can publish a request on the Agent Market; other agents then bid, and settlement occurs on-chain—entirely without human intermediaries.

Looking at the broader environment, in the first quarter of 2026 the market attention on the intersection of AI and crypto narratives increased significantly. NEAR Protocol briefly led AI token trading volume in March 2026. Some analysts point out that, under the macro environment of 2026, institutional capital is willing to pay higher premiums for AI infrastructure than for areas such as decentralized gaming. NEAR is viewed by multiple industry research organizations as one of the representative projects in the AI infrastructure layer. Together with projects like Bittensor and Internet Computer, it forms a multi-layer competitive landscape in the decentralized AI track.

Differentiated path: data sovereignty and privacy protection

NEAR’s second narrative dimension is a “user-sovereign AI platform”—allowing users to own and control their AI agents and the data generated by those agents. The near.com super app launched in February 2026 is the terminal expression of this strategy: users can manage wallets within a single entry point, execute confidential transactions, and receive AI recognition and decision support based on their own data.

This narrative is persuasive in light of the growing concern over data privacy. In April 2026, at Buidl Asia 2026, Polosukhin pointed out that current blockchain systems are designed with “complete transparency” as a principle, but in scenarios where AI agents represent individuals to manage finances, such transparency will “create security and privacy vulnerabilities.” NEAR’s proposed confidential-computing-based blockchain infrastructure—allowing transaction settlement to hide asset amounts, transaction details, and wallet identities—is precisely a structural response to this risk. The Confidential Intents feature introduced by NEAR builds an execution environment with restricted visibility through private sharding, making cross-chain transactions invisible to the outside during settlement.

Application validation and data transparency remain core challenges

The market currently has the following major points of divergence regarding NEAR’s AI narrative:

First, the ecosystem’s application data lacks transparency. Key indicators such as the activity level of AI agents on the Agent Market, transaction frequency, and task completion rates have not yet formed a publicly available, standardized disclosure system. The market needs clearer signals to judge the platform’s actual usage, rather than relying solely on narrative-driven expectations.

Second, the divergence in TVL data reflects the fragility of the ecosystem. Although NEAR has a first-mover advantage at the AI agent narrative layer, unstable liquidity accumulation capability is a signal that needs to be observed. Some analyses suggest that NEAR’s current valuation essentially “directly bets on future utility rather than on currently realized activity.”

Third, there is a timing mismatch between valuation and price. Some market participants believe that NEAR’s current 1.9 billion USD-level market cap already discounts a large portion of the expected value growth from the future AI agent economy, but the transmission mechanism between on-chain user growth and token price remains unclear. Analyst van de Poppe lists NEAR as a value asset in the crypto AI track based on its relatively low price-to-sales ratio, but his own revenue model also assumes a 300% to 500% year-over-year compound annual growth rate—whether this growth rate can be sustained carries substantial uncertainty.

Industry impact analysis: from a “public chain narrative race” to modular division of labor in “AI economic infrastructure”

Impact on the competitive landscape of public chains

NEAR’s AI agent strategy is changing the competitive dimensions within the public chain arena. Previously, public chain competition mainly revolved around TPS, gas fees, and the scale of ecosystem applications. In 2026, the key variable becomes: which chain can become the AI agents’ “native habitat.”

From the perspective of modular division of labor, an initial competitive pattern is taking shape. Solana establishes a foothold in the AI agent payment execution layer by leveraging high throughput and deep cooperation with Google Cloud (Pay.sh). The Ethereum Foundation, through its dAI team, promotes agent coordination standards such as ERC-8004 and x402, but AI remains only one of many priority areas. NEAR, at the protocol layer, designs an Intents architecture and Agent Market tailored for AI agents, while continuously investing in privacy computing. In this modular division of labor, public chains are not direct substitutes; instead, they form complementary components of the AI economic infrastructure.

Impact on the narrative structure of the crypto industry

NEAR’s practice is validating an industry-level narrative transformation: the core users of smart contract platforms are shifting from humans to AI agents. This is not merely a technical adjustment; it is a reorganization of the value chain. If the assumption that “AI agents are the main users of blockchain” holds, then the criteria for assessing a public chain’s competitiveness will expand from “friendly to human developers” to “natively supporting AI agent interactions.”

At the same time, security challenges introduced by AI agents are driving technological iteration. The consensus/execution separation architecture introduced by Nightshade 3.0, the privacy execution layer of NEAR’s Confidential Intents launched in February 2026, and verifiable inference within trusted execution environments (TEE) after NEAR AI joins the NVIDIA Inception program—these technological directions indicate that the AI agent economy requires not only public chain infrastructure, but also the coordinated evolution of multi-dimensional institutional infrastructure such as identity, compliance, and security.

Multi-scenario evolution: three possible paths

Based on current data, technology roadmaps, and market conditions, NEAR’s AI agent narrative may evolve along the following three paths during the second half of 2026 to 2027:

Path one: Accelerated deployment of the AI agent economy (optimistic scenario)

Key assumptions: user and transaction growth for NEAR Intents and Agent Market continues; at least 2 to 3 large-scale AI agent applications achieve product-market fit on the NEAR chain; Nightshade 3.0 completes its core feature deployments on time (consensus-execution separation, privacy sharding).

In this scenario, NEAR’s number of active on-chain addresses is expected to grow further, and protocol revenue could achieve a multiple-level increase compared with the baseline of 2026 as a whole. Market perception of NEAR would shift from “a public chain with an AI background” to “the core infrastructure for the AI agent economy,” with the valuation logic correspondingly reconstructed.

The main risks of this path lie in execution: whether the pace of product iteration can keep up with the rapidly changing needs of AI agent developers, and whether enough “killer” agent applications emerge in the ecosystem. In addition, the fragility suggested by TVL data indicates that even if on-chain interactions are active, liquidity accumulation capability remains a test.

Path two: Narrative and usage volume misalignment, sustained valuation pressure (neutral scenario)

Key assumptions: the AI agent narrative maintains market attention, but the growth rate of on-chain data slows toward a plateau; Agent Market runs steadily without explosive growth; NEAR’s differentiated positioning remains clear in the market, but the price fails to reflect fundamental improvements.

In this scenario, NEAR would continue to be in a phase of “strong narrative and weak price.” This does not negate its long-term value, but it means the market will need clearer catalysts to break through valuation ceilings. The current price-to-sales ratio of about 34x (lower than Solana’s 40x) may provide some downside buffer, but it is not enough on its own to drive an upside move.

Path three: AI narrative migration and a reshaped competitive landscape (pessimistic scenario)

Key assumptions: stronger competitors emerge in the AI agent infrastructure track (for example, Ethereum Layer 2 ecosystems emerging with Rollups specifically optimized for AI agents, or other Layer 1s launching more attractive native AI agent features); the market’s attention shifts to the “middleware layer” or “agent interaction protocol layer” of AI agent applications rather than the Layer 1 public chain itself; NEAR’s ecosystem encounters bottlenecks in attracting top AI projects.

In this scenario, NEAR may face a “first-mover dilemma”—having invested earliest in the AI agent narrative but failing to capture a sufficient portion of value growth in the sector. Even so, under a pessimistic scenario NEAR still has fundamental support: Nightshade sharding technology itself remains competitive even in non-AI scenarios, and the chain abstraction capability of the Intents architecture does not depend on the AI agent narrative standing alone.

It should be emphasized that the three paths above are speculative scenario analyses based on information available as of May 2026, and do not constitute any form of prediction or investment advice. The market contains unforeseeable variables, including macro environment changes, risks in technological implementation, and sudden shifts in the competitive landscape—actual outcomes may deviate from any of the scenarios above.

Conclusion

NEAR Protocol’s 2026 bet on AI agents is, at its core, answering a major industry question: as the subject of economic activity gradually shifts from humans to AI agents, what role should blockchain play? NEAR’s answer is that it should not merely be a tool that passively records transactions, but should become the underlying economic operating system for coordinating, bidding, executing, and settling between agents. Starting from the genetic roots of Transformer—co-authored by Illia Polosukhin—through the deployment of the NEAR AI Agent Market, and then to the Nightshade 3.0 privacy sharding roadmap, this chain aims to build a structural moat at the intersection of AI and blockchain.

This answer needs time to be validated. On-chain data is growing—active addresses reaching 600,000, Intents transaction volume surpassing 10 billion USD, and revenue in the first four months exceeding the entirety of 2025—but the divergence in TVL data and the lack of transparency in AI agent application data remind observers to remain cautious. Whether NEAR can transform from a “narrative pioneer” into a “value capturer” will depend on its continued execution across product development, ecosystem building, and developer experience. What we are watching is not only the evolution of a single public chain, but an important experiment in the integration of the crypto industry and the AI industry.