NEAR Protocol is betting big on AI agents — but the users haven’t exactly shown up yet
NEAR Protocol is making a serious wager that autonomous AI agents will become a major class of on-chain users, and it is building infrastructure meant to handle machine-speed transactions before the traffic arrives.
- AI agents as on-chain users
- Dynamic resharding in June 2026
- Cross-chain settlement via NEAR Intents
- Privacy tooling and token buybacks
- Big thesis, weak user numbers
The core idea is simple enough: AI agents are software programs that can act on their own, make decisions, and execute tasks like paying for compute, settling payments, labeling data, or even trading assets without a human clicking every button. NEAR believes those agents will need a blockchain built for speed, flexibility, and interoperability — not a network that gets clogged the moment demand spikes.
That is not a silly thesis. In fact, it is one of the more coherent AI-crypto bets in the market right now. The problem is that the network’s usage metrics are telling a far less glamorous story.
Why NEAR thinks AI agents need a different blockchain
NEAR’s pitch starts with a basic reality: conventional blockchains were designed mostly around human users, not software agents firing off bursts of transactions at machine speed. Humans are slow. Bots are not. AI agents can move fast, act in parallel, and create demand spikes that expose weak infrastructure immediately.
That matters because if an agent is supposed to buy compute, move funds, or settle a trade in real time, it cannot sit around waiting for congestion to clear like a shopper trapped in a checkout line with one cashier and a broken scanner. A settlement layer for agents has to be fast, resilient, and able to scale without turning every busy moment into a traffic jam.
NEAR’s answer is dynamic resharding, the centerpiece of its June 2026 upgrade and part of NEAR network release 2.13. NEAR has used sharding since launch, but dynamic resharding takes the concept further by automatically adding capacity when a shard fills up, instead of forcing validators to coordinate that expansion manually.
For readers new to the term, sharding means splitting a blockchain into parallel partitions called shards, so the network can process more activity at once. A useful way to think about dynamic resharding is as adding more checkout lanes when a store gets crowded, except the store is a blockchain and the customers are autonomous agents with money on the line.
NEAR’s co-founder has also leaned into the AI angle with credibility that many crypto teams would kill for, having co-authored the 2017 transformer paper that helped shape modern AI. That does not guarantee NEAR wins, obviously, but it does at least mean the team is not just slapping “AI” on a token and hoping nobody notices the smell.
NEAR Intents: cross-chain settlement for agents
Speed alone is not enough. NEAR is also pushing NEAR Intents, a cross-chain settlement system designed for agent-friendly commerce. The basic idea is that a user or AI agent states the outcome it wants, and specialized participants called solvers figure out the best route to make it happen across blockchains.
That is a cleaner model than forcing every user or bot to manually juggle bridges, swaps, and chain-specific quirks. The agent says what it wants. The network does the plumbing. In that sense, Intents acts like a unified commerce layer where the route matters less than the result.
For autonomous systems, that is a big deal. Machines do not care about chain tribalism, and they do not need a ten-part tutorial to move value from one network to another. They care about execution, cost, reliability, and speed. If NEAR can offer that without making users think like protocol engineers, it has a real shot at becoming useful infrastructure rather than just another ecosystem with a loud mascot and a quiet backend.
NEAR says the Intents system has already handled tens of millions of dollars in fees. That is meaningful, and it suggests the plumbing is not purely theoretical. Still, there is a massive difference between “people are using this” and “this is the default settlement rail for an emerging agent economy.” Crypto has a long and embarrassing history of confusing early activity with durable adoption. A few million in fees can be validation, or it can be a warm-up lap.
Privacy is not optional if AI agents handle money
Another part of NEAR’s pitch is privacy tooling. The network is emphasizing support for confidential treasuries, private multisig, payroll, balance management, and prompt anonymization for AI models. That is not just a nice-to-have feature set; it is probably necessary if AI firms, treasuries, and autonomous systems are going to use blockchain rails at scale.
Public chains are transparent by design, which is great when you want auditability and terrible when you want to keep treasury movements, internal prompts, or compensation flows from becoming public gossip fodder. If a company’s AI stack is handling payments and operational decisions, broadcasting every detail on-chain can turn privacy into a very expensive afterthought.
This is one of the more underappreciated parts of the AI-crypto thesis. The narrative often focuses on speed and automation, but privacy may be just as important. Not every wallet balance, prompt, or payroll transfer should be a public spectator sport.
The tokenomics are built to capture usage
NEAR is also tweaking tokenomics so that usage can feed directly into token demand. That includes lower inflation and a fee mechanism where Intents fees are used to buy NEAR tokens on the open market. In plain English, if activity on the network grows, there is a mechanism that can create demand for the token alongside that usage.
That is the dream every protocol wants: real utility translating into real token value instead of price action driven mostly by vibes, leverage, and people pretending a chart is a business model.
The line NEAR likes to push is that “the token is the currency of agents.” If AI agents are actually transacting, then a token with direct usage and value-capture mechanics could matter a lot. But that only works if the agents show up in meaningful numbers. A token demand model built for an economy that has not arrived yet is still a model, not proof.
The ugly part: users are falling, not rising
Here is the part that keeps this from becoming a victory lap: network usage has reportedly dropped sharply. Daily active users reportedly fell from nearly 3 million earlier in 2026 to a small fraction of that later on. That is a serious red flag.
Strong infrastructure and a strong narrative are useful, but they do not automatically create adoption. If anything, falling usage can be a warning that the market is getting ahead of itself. Crypto loves to front-run reality, and AI has become the latest excuse to do that with a straight face.
NEAR’s price rally has also been helped by the AI narrative and short squeezes. A short squeeze happens when traders betting against a token are forced to buy back in as price rises, which can accelerate the move higher regardless of underlying fundamentals. In other words, a violent price run is not always a sign of deep conviction; sometimes it is just leverage getting dragged through the shredder.
That does not mean NEAR is fake. It means the market may be rewarding the story faster than the network is proving it. Those are not the same thing, and crypto investors who confuse them usually learn that lesson the hard way.
What NEAR gets right — and what still needs proving
To NEAR’s credit, this is not a gimmick. The network’s architecture is aimed at a genuinely plausible future: one where AI agents transact, settle, and coordinate at scale across multiple chains. Dynamic resharding, Intents, privacy features, and tokenomics all fit together in a way that feels deliberately engineered rather than lazily narrated.
That matters. A lot of projects in crypto hear “AI” and start hallucinating product-market fit. NEAR at least seems to understand the technical shape of the problem. It is trying to build the rails before the traffic becomes unavoidable.
But that still leaves a brutal question: where is the traffic?
The thesis is coherent, but falling active users show the agent economy has not arrived yet. The bet is placed and the infrastructure is built; the economy it is built for has yet to show up. That is not failure. It is just the reality of building ahead of demand, which can look brilliant in hindsight and painfully overengineered in the meantime.
Compared with broader blockchain competition — including Ethereum, Solana, and other chains chasing scalability, interoperability, and AI-related use cases — NEAR is trying to carve out a specific lane: settlement for autonomous agents. That niche is real, and if AI agents become meaningful economic actors, the network could be well positioned. But niche positioning only matters if the niche becomes big enough to matter.
For now, NEAR looks like a technically serious blockchain making a credible bet on decentralized AI infrastructure, machine-speed transactions, and cross-chain settlement. It is one of the cleaner examples of the AI-crypto thesis with actual engineering behind it. Still, a good thesis and a good chart are not the same thing. One can survive reality; the other often gets mugged by it.
Key questions and takeaways
What is NEAR betting on?
NEAR is betting that autonomous AI agents will become major blockchain users and will need fast, reliable settlement infrastructure for payments, compute, trading, and data workflows.
Why are AI agents important for blockchain?
AI agents can transact at machine speed and in bursts, which makes them a much tougher infrastructure test than ordinary human users clicking through wallets one at a time.
What does dynamic resharding do?
Dynamic resharding automatically expands network capacity when demand spikes, reducing the need for manual validator coordination and helping the chain handle heavy traffic.
How does NEAR Intents work?
Users or agents specify the outcome they want, and solvers find the best cross-chain route to complete the transaction. It is designed to make settlement simpler and more automated.
Why do the privacy tools matter?
AI systems and treasuries often need confidentiality. Private multisig, confidential treasuries, and prompt anonymization help keep sensitive operations from becoming public clutter.
How does NEAR’s token capture value?
NEAR has lower inflation and a fee mechanism where Intents fees are used to buy NEAR tokens on the open market, tying usage more directly to token demand.
What is the biggest risk?
The biggest risk is that the AI-agent thesis stays ahead of actual adoption. If daily active users keep falling and Intents growth stalls, the narrative can outrun reality for only so long.
Is NEAR a serious project?
Yes. The architecture and product direction look credible. The question is not whether NEAR is building something real — it is whether enough users, agents, and fees will arrive to justify the bet.
“The future of crypto is autonomous AI agents transacting at machine speed.”
“Dynamic resharding removes the human bottleneck entirely.”
“NEAR is positioning itself as the infrastructure built to absorb that load.”
“The thesis is coherent, but falling active users show the agent economy has not arrived yet.”
