What happened
NEAR AI posted on X on Oct. 21 that cumulative deposits into its staking-for-inference product crossed 1 million NEAR. Crypto.
News reported the figure the same afternoon, citing the team's own disclosure. The mechanic is straightforward. A user stakes NEAR through the service, the position accrues staking rewards at the network rate, and those rewards are redirected to cover inference calls on NEAR AI's models.
The principal stays with the user. There is no token swap, no wrapped derivative, no new asset issued at the point of deposit, based on how the team has described the flow to date. One million NEAR is the stake-side figure, not a dollar figure, and the team did not break out how many unique wallets contributed.
Why it matters
Most AI-and-crypto crossovers so far have leaned on token issuance or point systems to bootstrap usage. This one tries something narrower. It uses the yield a holder was already going to earn and spends it on compute.
That changes the pitch to the user. You are not asked to sell NEAR, lock it into a new contract class, or trust a bridge. You are asked to let the staking rewards you were accruing anyway pay for inference you were going to buy anyway.
The downside is boring but real. If inference demand from stakers is thin, the headline staked number flatters the business. One million NEAR earning a mid-single-digit staking rate is a modest annual compute budget, and it is split across however many users signed up.
The number to watch next quarter is not stake. It is inference calls per staked NEAR.