Bankless
Bankless

NEAR’s New Token Utility and AI Economy | Illia Polosukhin

What if staking a token could keep an AI agent thinking? NEAR cofounder Illia Polosukhin joins David to explain NEAR’s new staking-powered inference model, where yield funds access to confidential and verifiable AI. They unpack how NEAR AI Cloud, IronClaw, Intents, and the NEAR token fit into one in

Topics Discussed

Episode Summary

Executive Summary: The episode explains Near’s upgraded token utility: staking NEAR now pays for confidential AI inference on Near AI Cloud, creating a yield-for-compute model for individuals, agents, and businesses. Ilya frames Near as a vertically integrated stack combining blockchain, confidential/verifiable inference, agent tooling, and a future compute marketplace, all aimed at enabling autonomous AI systems with trust, privacy, and settlement.

Main Topics: NEAR AI Cloud as confidential and verifiable inference (Priority: 5/5): Ilya describes Near AI Cloud as a confidential computing layer built on blockchain primitives and trusted execution environments, where prompts and outputs remain private and inference can be attested end-to-end. Staking NEAR for AI inference (Priority: 5/5): The new token upgrade lets users stake NEAR to receive inference capacity, effectively trading staking yield for AI access rather than paying purely in fiat. Verifiability as a core AI requirement (Priority: 5/5): Beyond privacy, Near emphasizes proof of what model ran, what prompt was used, and what output was returned, with provenance attestation designed for mission-critical and agentic workflows. Vertical integration across AI, blockchain, and intents (Priority: 5/5): Near’s stack connects AI inference, the Intents market, and the blockchain so agents can transact, settle, and operate with verifiable actions and financial rails. Autonomous businesses and agentic infrastructure (Priority: 4/5): The discussion frames Near as infrastructure for always-on autonomous agents/businesses that need continuous inference, finance, execution, and trust. Compute market design and liquidity (Priority: 4/5): Near is building toward a marketplace for GPU/computation that can handle under-specified demand, improve transparency, and create liquidity for compute as an asset. Partner ecosystem and go-to-market (Priority: 3/5): Brave, Venice, Bermuda, and remittance-related partners are highlighted as examples of privacy-sensitive, human-first applications using Near AI Cloud.

Key Arguments: Confidential inference is the primary user benefit: nobody else can see prompts, queries, or responses when using Near AI Cloud. Verifiability matters because AI providers can rewrite prompts, filter outputs, or even inject malicious tool outputs; Near aims to provide attested provenance of the full inference chain. Staking NEAR functions as a utility mechanism: users trade staking yield for AI inference capacity. The supply side is bootstrapped with Near-owned GPUs, with a path for third-party GPU providers to join and earn NEAR emissions. Near Intents converts compute into a tradable/liquid asset and captures transaction fees that support token value. Autonomous businesses need three things: intelligence, finances, and execution; Near aims to provide all three in a single stack. The compute market resembles oil and electricity: many heterogeneous qualities, but unlike oil, compute is consumed immediately and requires a more dynamic market structure. AI may collapse parts of the software stack into personalized, generated interfaces and workflows, but high-quality SaaS products may remain important as AI-enabled systems evolve.

Data Points: NEAR token upgrade: Staking NEAR now pays for inference on Near AI Cloud - Announcement of the new formal integration between the token and AI cloud Privacy model: End-to-end confidential inference - Near AI Cloud is described as keeping prompts and outputs private from others Verifiability metadata: Model hash, prompt hash, output, GPU/CPU provenance - Attestation chain shown in the frontend for inference verification Brave user base: Over 100 million users - Referenced as a major integration partner using Near AI Cloud Revenue capture from emissions: 20% to 50% - Mentioned as the approximate range of NEAR emissions captured and burned by the protocol on the dashboard Universal access model: Proportional to stake - Inference access scales with the amount of NEAR staked Network uptime: Over 5 years - Mentioned in a sponsor read about Near.com infrastructure Cross-chain volume: Over $23 billion - Mentioned in a sponsor read about Near.com Traditional pricing model: Fiat subscription or per-million-token fees - Contrasted with staking-based inference access

Pivotal Quotes: "It’s like universal basic AI, right? Effectively. If you hold near, you have access to some amount of AI inference that is available to you." — Ilya: Explaining the purpose of staking NEAR for inference access "This is AI money, right? The AI money needs to have the sovereign security, which is what blockchain is." — Ilya: Summarizing Near’s value-capture thesis "If you want true autonomous businesses, you need properties that it can run 24-7, it has access to intelligence, and it has access to finances, and it’s able to go and execute actions." — Ilya: Describing the requirements for autonomous agent systems

Implications: Near is positioning NEAR as infrastructure money for AI: a token that buys private inference, secures autonomous agents, and helps route value through compute markets. If successful, Near could become a key layer for privacy-preserving AI commerce and verifiable agent economies.

🔓 Sign Up for Unlimited Episode Search

About Bankless

View all episodes from Bankless