NEAR Is Betting Everything On AI Agents | Roundup

NEAR Protocol is one of the few tokens up year-to-date and has repositioned toward AI and cross-chain interoperability over the past 12 to 24 months, a shift from its original go-to-market strategy. Recent price movement is attributed partly to Arthur Hayes publicly predicting a significant pump and to Jensen Huang acknowledging Ilya, who is cited in the original transformer paper, causing NEAR to pump roughly 50 percent in a single day. The speakers are careful to note that current momentum reflects a mix of good technology and favorable narrative timing rather than adoption-driven growth.

NEAR's technical architecture combines three layers: chain signatures using multi-party computation for cross-chain key management, an intent-based solver network where value is accruing to the NEAR token rather than off-chain solvers, and trusted execution environments for private verifiable AI inference. NEAR chose TEEs over zero-knowledge proofs because ZK proofs for AI inference workloads are considered too expensive to bring on chain, and NEAR is described as the current leader in TEE-based private secure inference. NEAR is currently buying back approximately half of its token emissions, and if AI agent activity on-chain grows, buybacks could eventually cause demand to outweigh supply.

Despite positive price action, fees on the intents side are actually down in the most recent period, raising questions about whether revenue is following the narrative. Confidential TVL on NEAR is approximately 30 million dollars, described as relatively low. Approximately 0.0001 percent of all stablecoin volume is currently agent-driven, meaning the AI agent narrative is significantly ahead of actual on-chain usage. NEAR claims tens of millions of users but may not own the direct end-user relationship, since distribution runs through partners like Venice AI and Zcash wallets, creating unresolved questions about revenue sharing.

The private secure inference use case is argued to suit enterprises more than crypto-native users, because enterprises fear that models like OpenAI capture their proprietary data and could commoditize their applications. However, the speakers flagged that confidential computing faces extremely competitive pressure from hyperscalers like Microsoft, questioning why enterprises would choose NEAR over those alternatives. NEAR is also described as functioning more like a middleware or remote control layer over other chains rather than a self-contained base layer economy like Ethereum or Solana, and because inference happens off-chain it could theoretically be verified on other chains as well, raising questions about NEAR's differentiation.

NEAR has historically struggled with focus, having pursued around 20 different directions at one point. The speakers questioned whether its current AI branding is too generic, noting that NEAR AI does not specify what it does or who it targets, and drew a parallel to NEAR's previous operating system tagline, which was similarly broad. Very few chains successfully straddle crypto and AI audiences, with Bittensor cited as one example and NEAR potentially becoming a second. The speakers see NEAR's real competitive threat as coming from X402 and Base rather than Bittensor-adjacent projects, and if AI agent adoption on blockchain rails begins on Base or through X402 instead of NEAR, they believe NEAR's competitive position becomes more defensive rather than leading.

The speaker predicted that sovereign agents will begin using blockchain within roughly six months, and if NEAR enables those agents to run businesses on-chain and create their own economies, token value accrual becomes significantly more interesting. NEAR has already built execution-based infrastructure but still needs to build permissioning systems for agents acting on users' behalf. The speakers described building toward AI agents on-chain as slightly too early, which they characterized as exactly the right time to be building.

The speakers view NEAR's underlying fundamentals positively over the long term but treat the next six months as telling for where its AI positioning actually lands. If NEAR can spark genuine adoption of blockchain rails for AI agents, they see that as a strong bullish signal. If it cannot, the AI blockchain label risks becoming a narrative wrapper on a layer one that has not yet converted the pitch into measurable activity, noting that approximately zero dollars are currently flowing through NEAR from cross-chain agent routing.