When the buyer is a machine
Sometime in the next few years, the majority of transactions on the internet will be initiated by software acting on someone's behalf. An agent will negotiate a price, verify the work was done, and settle the bill without a human touching any of it. That is not a prediction about a distant decade. NEAR AI's Agent Market already lets agents bid for work, hold funds in escrow, and get paid on completion. Confidential Intents already lets them settle in USDC without exposing the counterparties or the commercial relationship to the public ledger.
Now we're on the precipice of seeing what money looks like when the buyer is a machine.
An agent has no tax bracket, no nationality, and no bank branch. What it needs from a monetary instrument is narrow and unforgiving: finality it can verify, privacy it can rely on, security it can measure, and a supply schedule it can price without guessing what a committee will decide next year. A machine will route around any asset that fails those tests, and it will do so instantly, at scale, forever.
I still contend that the networks that win the agentic era will be the ones that take these tests seriously. NEAR has already built most of the demand side. What we are proposing now is the supply side.
The demand side is already here
Eighteen months ago the case for NEAR was mostly technical. Chain Signatures let a NEAR account sign for assets on any chain without wrapping or bridging. Intents let a user express an outcome and let solvers compete to deliver it. Confidential compute in TEEs let inference run privately on user-owned infrastructure. Good engineering, and mostly theoretical in economic terms.
Now we're seeing that engineering turn into cash flow accruing value to NEAR. NEAR Intents cumulative settled volume has passed $32 billion, and the network recently cleared $1 billion in a single week. Since the fee switch went live in February, that revenue has been buying NEAR on the open market and removing it from circulation. The protocol now earns, and what it earns turns into demand for its own token.
This is the base layer of agentic commerce being assembled in public: settlement across every chain, private execution, a marketplace where agents are the economic actors, and a fee mechanism that converts all of that activity into demand for NEAR. The one piece that has not caught up is the monetary policy of the token itself.
Interestingly the best part is yet to come. NearAI’s Private Inference product is a rocket ship meeting market demand for privacy and confidentiality at EXACTLY the moment users and businesses need it. Alex Karp’s rant on CNBC set the stage for a collective social awakening amongst business leaders and now nearly every business in the market is demanding infrastructure designed to protect them. Rightfully so. As demand for private inference grows, NEAR is positioned perfectly to capture value for token holders and the market is orders of magnitude larger than the current market NEAR Intents serves.
Phase 1: issue what security requires, and nothing more
NEAR launched with one billion tokens. Today there are more than 1.3 billion in circulation, and at 2.5% annual issuance the network mints about 89,500 new NEAR every day. That is a small share split every couple of days, forever, and no dashboard adjusts for it.
Most of that issuance flows to stakers, while 57% of supply is not staked. A majority of holders pay for a transfer they do not receive. A meaningful slice of what stakers do receive leaves the ecosystem entirely, sold to cover the tax bill that staking income creates. We set out to build an alternative to monetary systems where inflation targets are the policy and dilution is the default. We should not be running one.
So the first proposal, going to a House of Stake vote next week, does exactly one thing. It lowers maximum annual issuance from 2.5% to 1.6%, gradually, every epoch, over 24 months. The 90/10 split between stakers and the treasury does not change. Guardrails can pause the ramp but cannot reverse it, because a schedule that might go back up is one the market cannot underwrite. A 90-day grace period gives wallets, exchanges and staking providers time to adjust. And a permanence covenant is written in so this is the last scheduled issuance change the network has to argue about.
Over six years the ramp keeps roughly 66 million NEAR from ever being issued. That is not money the protocol earns or spends. It is supply that never exists.
The costs are real and they belong in the open. Staking yield moves from about 5.4% to roughly 3.4% at the terminal rate, still above the 2.25% Solana ratified in August. A holder staking 1,000 NEAR ends year two with about 20 fewer tokens than under the current schedule, and needs the network to be valued roughly 1.8% higher to break even. Some marginal validator pools will lose economic self-sufficiency. I would not claim the cut pays for itself against a static counterfactual, because in our model it does not quite.
What I would point to is the evidence. When NEAR halved issuance from 5% to 2.5% last October, the fear was that validators would leave. There were 342 active validators the day it passed. Dollar revenue per staked NEAR fell by three quarters over the following months as price dropped. The validator set grew anyway, to 413 today. Security on NEAR is bought by delegation and conviction, and the network has already proven it can carry a lower issuance rate.
Phase 2: the end game is a fixed supply
Phase 1 is a parameter change. Phase 2 is a destination, and I want to say plainly where I believe NEAR should end up.
NEAR's issuance should eventually end. Total supply should be fixed. Every NEAR held should be a permanent share of the network, and the network should pay for what it needs out of what it earns.
Think about what that would mean for the agent I described at the top of this post. It would be settling in an instrument whose supply is known to the last unit, on a network that secures itself from revenue rather than dilution, with privacy guaranteed by hardware and cross-chain reach guaranteed by cryptography. No committee to model. No annual monetary debate to hedge against. No fund manager deciding who gets the capital. An asset a machine can price with certainty is an asset a machine will choose, and choose again, billions of times a day.
NEAR is one of very few networks where this is realistic. Issuance pays stakers roughly $146 million a year. The infrastructure that actually runs and secures the network costs a small fraction of that. Revenue is compounding, the treasury already holds a meaningful reserve, and the fee switch has shown that the protocol can convert usage into demand for its own token.
We are deliberately not proposing mechanics today. Replacing issuance with something durable has to keep NEAR at least as secure as it is now, and it has to fit where the technology is going. Chain Signatures, confidential compute, and TEE-secured infrastructure all change what economic security has to look like, and the design should follow the principles this community has already established: no oracle, no validator registry, no discretionary allocation.
When Illia withdrew the Sovereign Fund proposal in August, he identified the real problem. Any fund needs someone deciding where capital goes. We supported that direction with a controls framework and then listened to the discussion, and it changed our view. If discretion is the problem, wrapping it in controls does not fix it. The better path is one where nobody has to make an allocation decision at all. That is the path we intend to research and design with this community, in the open, and any proposal that comes out of it will go through its own discussion and its own vote.
Why SVRN is willing to pay for this
SVRN holds roughly 56 million NEAR, most of it staked. Under the lower issuance rate our model estimates about 855,000 NEAR a year in forgone yield, which under US GAAP is several million dollars of recognized revenue we are choosing to give up. Phase 2 would reduce that revenue line materially further. Neither proposal creates a fund, a subsidy, or an allocation program from which SVRN could receive anything.
We are doing this because we hold our position in NEAR, and because we believe a crypto-economic system built around NEAR as a store of value, sitting underneath a growing demand economy, is worth far more than the yield we are giving up. I will happily trade a revenue line on an income statement for a balance sheet in the billions that we can use to build in this ecosystem.
This also goes to the mission that SVRN exists for. We want technology to enable individual sovereignty: control of our data, our context, our models, and the agents that act for us. A currency that a small group can dilute by decision is not sovereign money, whether the group sits in a central bank or a governance forum. A currency with a fixed supply, secured by revenue, settled privately, and priced by the market is. Designing that instrument for NEAR is the most consequential thing we can do with the position we hold.
What we are asking
For Phase 1: read the formal proposal, check the numbers, and tell us where you disagree. The full methodology, risks, and technical specification will be there.
For Phase 2: join us. This is a mission at the edge of cryptography, confidential compute, crypto-economic security, and the design of permissionless stores of value. We are not going to get it right alone, and we should not. The community that designed NEAR's governance principles is the community that should design its final monetary form.
The agents are coming, and they will need money. We should choose a future where the instrument they settle in is one that no one can dilute, no one can censor, and no one has to be trusted to allocate. NEAR can be that instrument. Let's build it.
Sal Ternullo is CEO of SVRN, Inc. (NASDAQ: SVRN). SVRN holds 55+ million NEAR, runs validator infrastructure through partners, authored HSP-007 on MPC node incentives, and operates an MPC node under that program.
overview
