NEAR is staking its future on a very specific idea: that the next big wave in crypto will not be humans clicking buttons, but autonomous AI agents trading, paying, and coordinating with each other at machine speed – and that these agents will demand a blockchain purpose‑built for their needs.
At the center of this strategy sits a major June upgrade, “dynamic resharding,” which turns NEAR into a chain that can automatically scale its capacity up and down as demand changes. Around this core, the project is building tools for cross‑chain settlement, privacy, and intent‑based transactions, while branding NEAR as “the currency of agents” and “a unified commerce layer” for AI.
The story is ambitious and internally coherent. The technology exists and is shipping. Yet there is also a hard, inconvenient number: a gap between NEAR’s bold narrative and its current on‑chain usage, which anyone evaluating the project needs to confront.
Below is a structured breakdown of NEAR’s thesis, how its tech stack fits that vision, how the token is designed to capture value, and what’s missing today.
—
The core thesis: an economy of AI agents on-chain
NEAR’s entire roadmap flows from a clear and narrow view of the future: most blockchain activity will eventually be driven not by individual users, but by autonomous software agents.
In this vision, an “AI agent” is a program capable of acting independently toward some objective. It might:
– Buy and sell compute resources for model training or inference
– Pay for access to proprietary datasets or APIs
– Execute arbitrage and liquidity strategies across exchanges
– Negotiate and settle micro‑tasks like data labeling or content moderation
– Coordinate with other agents to fulfill complex workflows
Crucially, these actions happen at machine pace. An opportunity might trigger thousands or hundreds of thousands of transactions in seconds as swarms of agents race to capture value or complete tasks.
That traffic pattern is nothing like today’s retail‑driven crypto usage, where even busy periods are constrained by human attention spans and interaction speeds. Humans open wallets, sign transactions, and submit them sporadically; agents can hammer a network continuously and unpredictably.
NEAR’s leadership argues that existing blockchains – even scalable ones – were not architected with this pattern in mind. They can handle steady growth and some bursts, but they are not optimized to withstand repeated, massive spikes driven by machine actors without degrading user experience or making fees explode.
The bet, then, is that there will be a sizable future market where AI agents:
– Need trustless settlement rails
– Need neutral, programmable coordination environments
– Need a way to pay and be paid reliably across many services and chains
…and that a chain built around those requirements will become a key piece of digital infrastructure.
Adding weight to the story, NEAR’s co‑founder previously co‑authored the research that introduced the transformer architecture, the same family of models that underpins modern large language models. That background makes NEAR’s AI‑native framing less like opportunistic marketing and more like a continuation of a long‑running interest in scalable machine intelligence.
—
Why AI agents might need a specialized blockchain
A reasonable objection is: why can’t AI agents simply use whatever blockchain humans already use?
In principle, they can. In practice, NEAR argues that three constraints become critical at machine scale:
1. Unpredictable demand spikes
Agents don’t log in for an hour after work. They react the instant conditions change. When a profitable arbitrage, a favorable pricing model, or a data‑buying campaign appears, thousands of agents could rush to act in the same block interval. Traditional chains experience congestion under such conditions: fees spike, confirmation times lengthen, and the network becomes unreliable for low‑value tasks.
2. Fine‑grained, frequent, low‑value transactions
Many agent‑to‑agent interactions will be tiny – micro‑payments for single API calls, per‑row data labeling, incremental compute usage. The base chain must be cheap and responsive enough that settling these on‑chain makes economic sense, even at high frequency.
3. Cross‑domain activity and composability
An agent may need to buy compute from one ecosystem, trade on another, and store encrypted results elsewhere, all in a single flow. This demands not just raw throughput, but also robust cross‑chain connectivity and abstraction so the agent doesn’t “care” which underlying chain it touches.
From NEAR’s perspective, building for this world means:
– Scaling elastically with demand, not just pushing a static TPS ceiling higher
– Keeping transaction costs predictable and low enough for machine‑to‑machine micro‑commerce
– Providing developer tooling that makes it straightforward to script and orchestrate AI agents’ economic behaviors
Dynamic resharding is NEAR’s main answer to the scaling side of this challenge.
—
Dynamic resharding: automatic scaling in and out
NEAR has been a sharded blockchain from early on: instead of every validator processing every transaction, the network is split into multiple “shards,” each handling a portion of the load. This allows parallel execution and greater total throughput.
Static sharding, however, has its own issues. If the shard count is fixed, then:
– You may over‑provision: many shards sit half‑empty in quiet periods, wasting resources.
– Or under‑provision: too few shards during surges cause congestion and fee spikes.
Dynamic resharding is meant to solve this by allowing the network to automatically:
– Split hot shards when load gets too high, distributing accounts and contracts across more shards.
– Merge underutilized shards when demand falls, reducing overhead and keeping validator duties efficient.
Conceptually, it turns NEAR into an “auto‑scaling” blockchain: the capacity should expand and contract in response to actual usage, without requiring manual governance proposals or risky, infrequent hard forks.
For AI agents, this is important because it addresses exactly their most challenging trait: volatile load. A sudden swarm of trading or compute‑buying agents can be absorbed by the system spawning more shards to handle the spike, instead of bottlenecking the entire network.
Technically, dynamic resharding is complex. It involves:
– Efficiently reassigning state (accounts, contracts, storage) between shards
– Maintaining consistency and security guarantees while splits and merges happen
– Ensuring developers and users don’t have to constantly worry about which shard they’re on
The June 2026 upgrade that introduces this behavior is therefore more than a performance tweak; it is the backbone of NEAR’s claim that it can be the execution and settlement hub for machine‑driven commerce.
—
Beyond scaling: the other pieces of NEAR’s “agent commerce” stack
Scaling alone does not make a chain attractive to AI agents. NEAR is also assembling additional components so that an agent‑centric economy can actually function:
1. Chain abstraction and cross‑chain settlement
Agents will not live exclusively on a single blockchain. NEAR is working on infrastructure that lets users and developers route actions across multiple chains while abstracting away the complexity. In an ideal flow, an agent expresses an intent – “re-balance this portfolio,” “procure X units of compute at the best rate” – and NEAR’s middleware handles the routing and settlement, even if it spans different networks.
2. NEAR Intents and higher‑level UX
Intent‑based systems let users or agents specify desired outcomes rather than individual transactions. This aligns well with how AI agents operate: they care about goals and constraints, not the step‑by‑step transaction sequence. NEAR is baking such abstractions into its roadmap to make it easier for agents to plug into on‑chain commerce.
3. Privacy and data handling
AI workloads sometimes involve sensitive data or proprietary models. NEAR is exploring privacy tooling so that agents can transact, access data, and prove certain properties without fully exposing everything publicly. For a real agent economy, some level of confidentiality is likely to be mandatory.
4. Developer tooling for agent integration
NEAR’s view is that many agents will be built and run by developers who are not crypto‑native. That implies the need for robust SDKs, straightforward wallets, and integration patterns that make it easy to connect agents to NEAR for payments, settlement, and coordination, without asking every AI engineer to become a DeFi expert.
Taken together, these elements aim to turn NEAR from “a fast L1” into a vertically integrated environment for agent commerce: scale, abstraction, privacy, and ease of use.
—
Tokenomics: tying network usage to the value of NEAR
For NEAR to function as “the currency of agents,” its token must not only be used within the network, but also be structured so that increasing activity translates into economic value for holders and network participants.
Key design points include:
– Gas and fees in NEAR
All on‑chain operations – transactions, contract calls, storage – are paid in NEAR. If AI agents begin submitting massive volumes of transactions, their activity drives sustained demand for the token as a medium of payment.
– Burn mechanisms
A portion of transaction fees is burned, removing NEAR from circulation. If network usage grows significantly, this burning can offset issuance or even result in net deflation, strengthening the link between utilization and scarcity.
– Staking and security
Validators stake NEAR to secure the network, and delegators can participate by staking through them. In a high‑activity future, rewards funded by transaction fees and inflation must balance security costs with sustainable economics.
– Agent‑centric financial flows
Longer term, NEAR’s team has framed NEAR not just as gas, but as the default unit in which agents denominate their interactions: paying for compute, datasets, and other services. If this narrative takes hold, NEAR could gain additional monetary premium as a unit of account in agent economies.
If the AI‑agent thesis proves out, NEAR’s token design is positioned to capture a share of that growth. If it does not, much of the token’s value case reverts to more typical L1 dynamics: user growth, developer traction, and ecosystem strength.
—
The complicating number: usage vs. narrative
Against this polished story stands a basic metric: actual on‑chain usage.
By its own admission, NEAR today does not yet host the kind of agent‑driven, high‑frequency economy it is designing for. Transaction counts, while respectable, are nowhere near the stress levels dynamic resharding is built to accommodate. AI‑specific agents and workloads are still in their infancy across the entire crypto sector, not just on NEAR.
This gap between narrative and reality is the “number that complicates the story.” It raises several questions:
– Are AI agents really going to move enough activity on‑chain to justify a dedicated architecture?
– If they do, will NEAR be the platform they choose, rather than other high‑throughput chains or off‑chain coordination layers?
– How long can NEAR invest in infrastructure for a future use case before that future materially arrives?
NEAR’s defenders argue that infrastructure must be built ahead of demand. If you wait for agents to overwhelm existing blockchains before building an AI‑native network, you are already too late. They frame today’s relatively modest activity as a normal phase for a platform investing in long‑term technical differentiation.
Skeptics counter that the AI‑agent use case is still speculative and that NEAR’s current adoption, relative to its ambitions, suggests a risk of over‑engineering for a market that may not emerge at the required scale.
Both views hinge on time horizons and risk tolerance.
—
How to weigh NEAR’s AI‑agent bet
Evaluating NEAR’s strategy means weighing several factors:
1. Technical coherence
The dynamic resharding upgrade, sharded architecture, and focus on elastic capacity genuinely align with the challenges posed by machine‑driven workloads. On a technical level, the “blockchain built for agents” claim is well grounded.
2. Differentiation in a crowded L1 market
Many chains are courting generic “AI + crypto” narratives. NEAR stands out by focusing on a specific vertical – agent commerce and settlement – and by shipping concrete protocol‑level changes to support it, rather than just integrating AI models into user interfaces.
3. Market timing and adoption risk
The biggest uncertainty is not whether NEAR can scale, but whether a large, economically significant agent economy will emerge on‑chain within a timeframe that rewards NEAR’s investment. This is a bet on both technology and behavior: that agents will prefer on‑chain coordination to more centralized options.
4. Ecosystem building
The raw protocol is only part of the equation. Success requires real projects building agent frameworks, markets for compute and data, intent routers, and DeFi venues that agents actually use. The pace and quality of that ecosystem growth will be crucial.
5. Comparative advantage in AI
NEAR’s AI pedigree and early positioning may help attract AI‑focused builders. But other platforms are also racing to integrate AI agents and off‑chain compute networks. NEAR must show that its design confers tangible advantages – lower costs, less congestion, better developer experience – in real deployments.
—
Is NEAR a good investment?
Whether NEAR is attractive as an investment depends heavily on your belief in the AI‑agent thesis and your tolerance for speculative infrastructure bets.
Consider:
– If you believe that autonomous AI agents will become major economic actors, that they will transact heavily on‑chain, and that elastic scaling and unified settlement will be crucial, then NEAR’s roadmap directly targets that scenario. In that case, current usage metrics might be less important than the quality of the underlying design and the team’s ability to deliver.
– If you are skeptical that agents will need blockchains for most of their activities, or you think off‑chain or hybrid systems will dominate, then NEAR’s specialized focus might look like a niche play, with limited upside beyond being “another L1.”
Risk‑aware analysis should also factor in:
– The current disconnect between narrative and on‑chain traction
– Competition from other high‑performance chains and rollup ecosystems
– Regulatory uncertainties around AI and autonomous economic agents
– NEAR’s treasury, runway, and ability to keep funding R&D and ecosystem incentives
NEAR is not a purely “AI narrative coin”; it is a functioning network with real technical depth. But the future it is optimized for has not yet undeniably arrived, making any allocation a bet on that future materializing.
—
Additional considerations for the AI‑agent settlement thesis
To assess NEAR’s chances in more depth, it helps to zoom out beyond the project itself:
– What exactly needs to be on-chain?
Not all agent behavior benefits from blockchain settlement. Many interactions – like rapid‑fire inference calls – may be better handled off‑chain with periodic, batched settlement. NEAR’s architecture is more compelling if you expect meaningful value and state transitions (not just logging) to land on‑chain.
– Will agents prioritize latency or trustlessness?
For ultra‑low‑latency trading or micro‑optimization, even a fast L1 might be too slow compared to centralized venues. NEAR’s role may be strongest where trustlessness, auditability, and composability matter more than absolute speed.
– Standardization of agent frameworks
If a few dominant agent frameworks emerge and choose particular chains as their default settlement layers, those choices could shape the landscape quickly. NEAR’s outreach to AI developers and the ease of integrating with mainstream tools will matter as much as TPS numbers.
– Economic incentives for early adopters
Getting the first wave of serious AI‑agent projects onto NEAR may depend on targeted incentives, grants, and co‑development partnerships. Builders will go where they see not just tech, but also capital, support, and distribution.
—
Where NEAR stands today
At this point, NEAR represents one of the clearer, more technically grounded attempts to define what “AI x crypto” could actually mean beyond buzzwords. It is:
– Making a specific, falsifiable bet: that an on‑chain agent economy will emerge and require elastic, sharded infrastructure.
– Shipping protocol‑level upgrades, especially dynamic resharding, that align with that bet.
– Building complementary tools around chain abstraction, intents, and privacy to support real AI‑native workflows.
The missing ingredient is proof at scale: real AI agents driving sustained, economically meaningful activity on NEAR itself.
Until that happens, NEAR will sit in an in‑between state: more concrete than pure narrative plays, but still ahead of the evidence needed to fully validate its vision.
—
Frequently asked questions
What is NEAR betting on with the AI‑agent thesis?
NEAR is betting that autonomous AI agents will become major users of blockchains, transacting at machine speed and volume. It aims to be the primary settlement and coordination layer for these agents by providing elastic scaling, low fees, and tools tailored to their workflows.
What is dynamic resharding?
Dynamic resharding is NEAR’s mechanism for automatically adjusting the number of shards (parallel execution units) based on demand. Shards can split when overloaded and merge when underused, allowing the network to scale capacity up or down without manual intervention.
Why would AI agents need a special blockchain?
Agents can cause sudden, massive transaction spikes and rely on frequent, low‑value interactions. A specialized chain like NEAR is designed to handle these volatile loads with predictable fees and performance, while offering abstractions that make cross‑chain, goal‑oriented behavior easier to implement.
How does NEAR’s token capture value from this?
NEAR is used to pay transaction fees and for staking. If agent activity grows, demand for NEAR as gas increases, and fee burning can reduce supply. Over time, if agents use NEAR as their default unit for paying for compute, data, and services, the token could also gain monetary utility in agent economies.
What is the main problem with NEAR’s story?
The primary issue is the gap between the ambitious AI‑agent narrative and current on‑chain reality. The agent‑driven economy NEAR is built for is still largely hypothetical. Usage today does not yet validate the scale of the bet, making it a forward‑looking, high‑conviction strategy rather than a response to existing demand.
Is NEAR a good investment?
NEAR may be compelling if you strongly believe in on‑chain AI agents and are comfortable backing infrastructure ahead of clear adoption. It is riskier if you doubt that agents will use blockchains at scale or think that generic L1s and off‑chain systems will be sufficient. As with any crypto asset, it should be viewed through the lens of high volatility and technology‑driven uncertainty.
—
NEAR’s attempt to become the settlement layer for AI agents is one of the more focused and technically ambitious plays in the intersection of AI and crypto. Whether that bet pays off will depend less on the elegance of dynamic resharding and more on a harder question: will AI agents, in meaningful numbers, actually choose to live and transact on-chain?