Key Insights
- NEAR AI launched a staking-based payment mechanism that converts staked NEAR into recurring AI compute credits.
- Users retain ownership of the underlying stake, which can be withdrawn upon unstaking, while credits are generated from the staked position.
- The model replaces conventional subscription payments with an onchain staking workflow for confidential inference and autonomous agents.
- AI usage denominated through staking increases the amount of NEAR locked while compute is active, creating a utility-driven source of token lockup.
- All 43 models on NEAR AI are accessible through staking, spanning providers including Anthropic, OpenAI, and Google.
- The launch is the first production implementation of NEAR’s proposed “AI money” framework, connecting token staking directly with AI compute access.
Today NEAR is launching Staking for NEAR AI, a way to pay for confidential inference and always-on agents by staking NEAR rather than by putting a card on file. You stake the token and it converts into monthly compute credits that scale with the size of your stake. The capital is not spent but staked, and it returns to your wallet when you unstake. There is no credit card, no cloud account, and no custodian holding your credentials. All 43 models on NEAR AI, from Anthropic, OpenAI, Google, and more, are accessible through this staking mechanism.
This makes NEAR one of the first production systems to pay for confidential inference and autonomous agents through onchain staking.
The Asset That Anchors the Agent Economy
Paying for AI by Staking NEAR
NEAR’s deep dive on the agent economy makes the argument that in an economy where agents are the primary actors, the asset that anchors value and secures the rails becomes a new monetary category, and that NEAR is built to be that asset. Part of that argument is that the classical functions of money change shape when the thing using money is a machine. Store of value stops being a rock you bury and becomes a claim on the throughput of the network it secures.
Today, that new monetary framework comes to life with staking NEAR for AI.
Staking NEAR equates to AI usage, prepaid in a form you can recover. You can adjust the stake as your usage changes, buy one-off credits by locking more, and cancel whenever you want, with the whole lifecycle running onchain. A stake keeps the value in the same token as the rest of your onchain life, doing two jobs at once: paying for the work and holding up the rail the work runs on.
Usage That Tightens Supply
Staking for AI also feeds NEAR's broader token economics, in a way that extends beyond the individual user. Staked NEAR is NEAR taken out of circulation for as long as it is used. A single subscription barely registers against total supply. But payment is not a rare event. It is the thing that happens every time anyone runs inference or stands up an agent, and every instance of it locks tokens away against real usage rather than speculation. As more of the network's AI is paid for this way, more of the supply is committed to work being done, and the value that activity creates routes back toward the people who hold and secure the network instead of pooling at a platform above them. Usage tightening the supply of the asset that secures the usage is the loop NEAR’s AI money thesis was built around.
Stake NEAR Today for Confidential Inference and Agent Hosting
What launches today is one loop, closed: the NEAR you hold and the AI you run, joined without a card in between. Start staking for AI usage at near.ai.
This is what AI sovereignty looks like in practice: You stake some NEAR before you close your laptop. An agent works through the night, private, its credentials never leaving the vault, its inference attested and locked away from the machine that served it. In the morning the work is done. You cancel, unstake, and the NEAR comes back. Nothing you touched ever sat in someone else's custody. The token you held and the intelligence you ran turned out to be the same thing.
