Why the Near $NRR ETF is Different | Hunter Horsley and Sal Ternullo

Bitwise’s NEAR ETF, ticker NRR, attracted about $50 million in inflows during its first three trading days, with more than $10 million entering per day and daily trading volume of $10 million to $20 million, Hunter Horsley said. He called the launch exceptionally strong compared with other single-crypto-asset ETFs launched that year, while stressing that the figures were preliminary and Friday’s results were not yet available. NRR is intended as a bridge for investors to NEAR and an additional channel for the network’s capital markets; it also stakes on investors’ behalf. Horsley linked early demand to improving sentiment, interest in AI, and product awareness, despite a market that was not in a raging bull phase and Bitcoin remaining well below its all-time high.

Bitwise’s involvement with NEAR began before the ETF, through institutional research and valuation work, a European ETP, and participation as a validator. Horsley said Bitwise is NEAR’s fourth-largest validator and contributed to the MPC network used by NEAR Intents; the firm manages about $16 billion and has roughly 150 employees. Ternullo described Intents as a system in which users specify an outcome and a market of solvers determines how to deliver it. It supports digital-asset swaps and, after Ando and 400 tokenized equity products were introduced, swaps between digital assets and tokenized equities. He reported more than 32 billion in Intents volume, without specifying the currency, and said usage currently comes from people using NEAR.com and from B2B partners.

Ternullo said Intents’ transaction take rates are scaling comparably to those of a centralized exchange, with net protocol revenue funding NEAR buybacks. The tokens are held in a permanent capital vehicle rather than burned, so he cautioned against describing the mechanism as straightforwardly “deflationary”; the community could later choose to burn the accumulated tokens or use them differently. Separately, a governance proposal would reduce NEAR inflation from 2.5% to 1.6% along a two-year linear curve. Ternullo also said the ecosystem should examine whether a fixed-supply NEAR asset is feasible.

Ternullo said he considered NEAR two to four times undervalued when it traded around $1, based on human blockchain use and NEAR Intents, without including AI agents or NEAR’s position at the AI-crypto intersection. David noted that NEAR later traded just below $5 after reaching $5.50. Horsley said investors increasingly assess crypto networks through use cases, value creation and capture, and ecosystem health. He cited NEAR’s utility, buyback mechanism, and active development, and compared its roughly $5 billion market capitalization with AI platforms that have traction, estimating that NEAR was valued at about one-tenth of their average. Ternullo argued that NEAR’s investment case is substantially less compelling without its AI exposure.

The AI thesis rests partly on the prospect of a fragmented technology landscape rather than a single dominant model, blockchain, or stablecoin. Horsley sees NEAR as a simplifying layer that could let people and agents use different AI models and contexts while moving across chains and assets. He acknowledged that this interoperability role would matter less if one blockchain, one stablecoin, and ChatGPT alone dominated. Ternullo expects agent-driven use of Intents could grow by a 10x multiplier rather than linearly, and expects NEAR AI data to emerge that could support a case based on revenue and product-market fit similar to the one he sees in Intents. The view that a diverse, interconnected system would be healthier than one dominated by a single provider was presented as an intuition, not an evidence-backed conclusion.

For enterprise AI, Ternullo described a move toward tiered model use: frontier models for tasks without highly confidential or personally identifiable information, and open-source models hosted in confidential-computing infrastructure for sensitive work. He expects sophisticated enterprises to add router or gateway layers to enforce policy and control model selection; using different models for different tasks could improve investment returns and unit economics. NEAR’s private inference offering hosts widely adopted open-source models and lets clients run a verifier to check that prompts and outputs are handled without telemetry or data leakage. Ternullo predicts privacy-guaranteeing architecture for shared services will become a fundamental requirement over the next two or three years. He cited Google search trends showing privacy reaching a peak at a multiple of any other peak as a rough indicator of rising consumer interest, framing privacy as sovereignty over information, analogous to the freedom and control over money associated with Bitcoin.