NEAR lets users pay for AI agents by staking tokens

NEAR Protocol has launched a staking-based payment system for its AI platform, letting users access confidential inference and autonomous agents by locking NEAR tokens instead of handing over a credit card.

The feature, announced on X, converts staked NEAR into monthly compute credits usable across all 43 AI models on NEAR AI, including models from Anthropic, OpenAI and Google. The tokens stay locked, not spent, and become available again after unstaking, the protocol said.

Users decide how much to stake. Larger positions generate more compute points, scaling usage without switching payment models. Someone who increases their stake gets additional one-time credits; someone who reduces it gets fewer credits the next month. Withdrawing entirely means unstaking.

NEAR described the rollout as one of the first production systems to pay for confidential AI inference and always-on agents through onchain staking. “The NEAR you hold and the AI you run, joined without a card in between,” the company wrote.

Every supported model is available through the staking mechanism, the protocol said. Users can switch providers without rethinking how they pay for inference or agent hosting.

Staking as a recoverable expense

Unlike a traditional subscription where funds are spent each billing cycle, the staked tokens themselves are not consumed while the service runs. The capital is not spent but staked, and it returns to the wallet when unstaked, NEAR said. The company framed the design as a way to keep control of assets and credentials while interacting with AI services. Confidential inference and hosted agents run without requiring users to hand over payment information to third parties.

Token economics effect

NEAR argued that staking for AI turns the token into a recoverable payment instrument rather than a consumable expense. The locked assets stay out of circulation while supporting AI workloads, the protocol noted. A single subscription would have little effect on overall supply, but repeated usage across developers and applications could commit more tokens to active computing instead of free trading.

The announcement did not estimate how much supply could eventually become locked through AI payments. It did not provide adoption forecasts.

Broader staking push

The new payment model builds on earlier efforts to expand staking beyond validator rewards. In February 2025, Nomura-backed Laser Digital introduced the Laser Digital NEAR Adoption Fund for institutional investors. The fund uses TruStake, an institutional staking solution from TruFin, allowing participants to earn staking rewards while supporting network consensus.

Laser Digital Chief Executive Officer Jez Mohideen said at the time that the fund combined exposure to artificial intelligence and digital assets with staking income. The product was available to eligible institutional and professional investors in selected jurisdictions outside the U.S.

NEAR also changed its monetary policy later that year. On Oct. 30, 2025, the protocol activated a network upgrade reducing annual token inflation from about 5% to roughly 2.4%, cutting yearly token issuance by nearly 60 million NEAR. The update also lowered expected staking yields from around 9% to approximately 4.5%, assuming roughly half of the circulating supply remained staked.

The new payment feature introduces another role for staking in the NEAR ecosystem by linking token deposits directly to AI computing, rather than relying only on validator participation or investment products. NEAR described the system as an example of “AI sovereignty,” where users can stake tokens, let private AI agents run without exposing credentials, and later recover the same tokens after unstaking.