NEAR Is Betting Everything On AI Agents | Roundup

NEAR protocol is one of the few tokens up year-to-date in the current market, and most of its significant technology developments have occurred in the past 12 to 24 months. The project has shifted from a standalone layer-one chain toward a cross-chain network, a thesis few predicted years ago. Chain signatures are described as multi-party computation cross-chain control that abstracts key management so users can transact across multiple chains simultaneously from NEAR itself without managing backend complexity. NEAR functions more like middleware or a remote control over other chains rather than a self-contained base layer economy like Ethereum or Solana.

NEAR chose trusted execution environments over zero-knowledge proofs for private secure AI inference and is described as the current leader in TEE-based private inference. Zero-knowledge proofs for AI inference are considered very expensive by comparison, and some blockchain technologists estimate TEEs as only marginally less secure, roughly 0.x percent. NEAR is partnering with Venice AI, a private inference network, and released its own open source AI sandbox called Iron Claw around the same time OpenAI's open source model went live, though Iron Claw had less developer tooling than the alternative it was based on. NEAR's inference currently happens off-chain, meaning validators are not part of the inference process, which raises questions about long-term differentiation since inference could theoretically be verified on other chains as well.

NEAR's tagline is user-owned AI, and its thesis combines cross-chain asset swaps with private secure AI inference, with blockchain intents and AI agent intents described as converging. However, fees on the intents side are actually down in the most recent period despite positive price action, and nothing fundamentally new has happened with NEAR in the last month, though the architecture is described as directionally correct. NEAR claims tens of millions of users but may not own the direct end-user relationship, distributing instead through third-party providers like Venice and the Zcash wallet, raising questions about revenue shares and who controls the end user. Confidential TVL on NEAR is approximately 30 million dollars.

The private secure inference use case is seen as better suited to small businesses and enterprises than to crypto-native users, as enterprises fear that models like OpenAI capture business data and could commoditize their applications. The user-owned AI thesis has been questioned given that crypto users caring about owning their own data has been proved wrong multiple times in crypto history, and hyperscalers like Microsoft could potentially serve the same confidential computing function. NEAR has historically struggled with focus, having pursued around 20 different directions at one point, and its brand name is considered too generic and not specific enough about what it does or who it appeals to.

The broader crypto-AI agent space is described as very early and heavily hyped, driven by crypto investors wanting exposure to the AI trend. Approximately 0.0001 percent of all stablecoin volume is currently agent-driven, indicating a significant gap between hype and actual usage. The agent-to-trading-execution use case has been overhyped and adoption has not materialized as expected over the past six months, with traders mostly still executing trades themselves because AI is probabilistic rather than deterministic. NEAR is being framed not primarily as a competitor to Bittensor but as a cross-chain layer whose real competition is X402 and Base, and essentially zero dollars are currently flowing through any cross-chain agent routing.

NEAR is currently buying back approximately half of its emissions, and its price-to-sales ratios are described as lower than those of Solana and Ethereum. Jensen Huang acknowledged NEAR's chief AI researcher Ilya publicly, and NEAR's token pumped approximately 50 percent that day. Ilya is cited in the original transformers paper from Google. Arthur Hayes and others have publicly called for a significant NEAR price increase, though recent price movement is attributed partly to market mechanics rather than fundamentals alone.

The bullish case depends on whether AI agents operating across multiple chains begin routing through NEAR before rivals establish that position. NEAR still needs to build permissioning systems for agents acting on users behalf, and building for AI agents on chain is described as likely slightly too early, which is characterized as exactly when you want to be building. A prediction was made that sovereign agents will start using blockchain within approximately six months. If that adoption initiates on Base or through X402 instead, NEAR's claim to the AI blockchain category becomes more defensive, and the next six months are described as the critical window for determining whether NEAR's AI narrative translates into real traction.