Solana's 99% Bet: Why the Foundation Thinks Humans Will Stop Touching the Blockchain by 2028

In two years, the human user may become a rounding error on Solana.

That is not a metaphor. That is the explicit forecast from Vibhu Norby, chief product officer at the Solana Foundation, who told industry audiences in March 2026 that "99.99% of all on-chain transactions in 2 years will be driven by agents, bots, and LLM-based wallets and trading products." In a separate interview, he widened the range slightly to "95 to 99% of all transactions" originating from large language models acting on a user's behalf. Either way, the message is the same: the era of humans clicking "Sign Transaction" in a wallet pop-up is ending, and Solana is building for the era that comes next.

This is the most aggressive vision of the agentic internet that any major Layer 1 has put on the record. Ethereum's response has been to ship standards — ERC-8004 for agent identity, ERC-8183 for trustless agent commerce. Solana's response has been to ship throughput and post a skill.txt at the root of its website so AI agents can read it and figure out how to mint a wallet on their own. The two approaches reveal something deeper than a marketing rivalry. They reveal a real philosophical split about what an "agentic" blockchain should optimize for.

The Numbers Behind the Headline

The "99%" claim is easy to dismiss as Solana cheerleading. The numbers underneath it are harder to wave away.

By March 2026, the Solana Foundation reports that the network has processed 15 million on-chain agent payments, with stablecoins as the default rail. Solana now accounts for roughly 65% of all agentic payment activity across blockchains — a share that has climbed steeply over the last three quarters. The x402 payment protocol, an HTTP-native standard for AI-to-API payments developed by the Coinbase team, has processed 35 million transactions on Solana alongside another 119 million on Base, with roughly $600 million in annualized volume moving through the standard at zero protocol fees.

To put that in context, x402 micropayments are economically viable on Solana because of two specific properties: sub-400ms finality and per-transaction fees around $0.00025. That combination is what makes a $0.01 API call profitable. On Ethereum L1, the same call would cost the agent more in gas than the service provider charges for the data — a structural impossibility, not a performance gap.

So when Norby projects 99% of transactions to autonomous agents, he is not extrapolating from zero. He is extrapolating from a network that already processes a third of a million agent transactions a day on x402 alone.

What Changes When the User Is a Machine

Crypto infrastructure has, until now, been built around the assumption of a human signer. Wallets prompt for confirmation. Block explorers render addresses in friendly fonts. Block times of 10 seconds, or 12 seconds, or two minutes felt fine because humans don't perceive the difference between 200ms and 2 seconds when clicking a button.

Strip the human out of the loop and every one of those assumptions breaks.

Block time becomes a hard constraint, not a UX preference. An LLM agent calling a paid API needs the payment to confirm before the HTTP request times out. That window is roughly 30 seconds for most clients, but practically much shorter for chains of agent calls. Solana's ~400ms finality fits comfortably inside that window. Ethereum L1 finality, even with single-slot finality proposals, does not.

Gas economics flip. Human users tolerate gas spikes during NFT mints because the asset they're buying is worth orders of magnitude more than the fee. Agents executing thousands of $0.01 micropayments cannot. Every basis point of fee compression matters when each transaction is its own line item in an agent's profit-and-loss statement.

MEV becomes a programmable game between bots, not a tax on humans. Today's MEV revenue largely accrues from front-running retail traders. In a world where 99% of transaction flow originates from agents that already know about MEV, are running their own private mempools, and are routing through MEV-aware RPCs like Jito or Helius, the entire revenue model for searchers shifts. Validator economics shift with it.

Validator revenue stops being driven by human FOMO cycles. When transaction count is dominated by autonomous agents settling stablecoin payments, network revenue smooths out. The bull-market spikes that made staking yields lumpy in 2021 and 2024 give way to a steadier baseline of machine-driven activity.

This is what Norby and the Solana Foundation are building for. Not the next bull market. The next economic substrate.

Two Philosophies of an Agentic Blockchain

The contrast with Ethereum is now sharp enough to draw cleanly.

Ethereum's strategy has been to define the protocol of agent commerce. ERC-8004, authored by collaborators from MetaMask, the Ethereum Foundation, Google, and Coinbase, gives every agent a unique on-chain identity — an ERC-721 NFT pointing to a JSON file that describes the agent's capabilities, endpoints, and payment address. ERC-8183, developed in part by Virtuals Protocol and the Ethereum Foundation, layers on programmable escrow: a job has four states (Open → Funded → Submitted → Terminal), an evaluator who alone can mark it complete, and a budget that is locked until the work is done.

Combined with x402 (which handles "how to pay"), the Ethereum stack answers three distinct questions: who are you (ERC-8004), can I trust you to deliver (ERC-8183), and how do we settle (x402). It is a clean, layered, standards-driven approach.

Solana's bet is different. It is that the infrastructure matters more than the standard. If your block time is fast enough and your fees are cheap enough, agents will figure out the rest. The Foundation has poured resources into making Solana itself agent-readable: a skill.txt file that any LLM-powered agent can ingest to learn how to create a wallet, query balances, and submit transactions without any human-written integration code. The Solana Developer Platform, launched in March 2026 with Mastercard, Western Union, and Worldpay as anchor users, bundles 20+ infrastructure providers into a single API surface for issuance, payments, and (eventually) trading.

Both bets could be right at the same time. ERC-8004 and ERC-8183 will likely become the cross-chain identity and escrow standards even when they're settled on faster chains. Solana will likely keep the throughput crown for the actual settlement layer. The agentic internet doesn't need a winner. It needs both.

But for sheer transaction count — for the number of times a machine asks another machine for something and pays for it — Solana's combination of finality and fee structure is hard to argue with. That is what underpins the 99% number.

The Stack: ElizaOS, AI Rig Complex, and the Linux of Agents

The frameworks that turn this thesis into shippable code are now mature enough to power production workloads.

ElizaOS has become the open-source default. The TypeScript-first framework provides a unified message bus that lets developers deploy a single agent across X, Discord, Telegram, HTTP endpoints, and on-chain environments without rewriting core logic. As of early 2026, it counts more than 200 plugins, 17,600 GitHub stars, 5,300 forks, and 1,350 contributors. Its supporters call it "the WordPress for Agents." Its critics call it "the Linux of agent operating systems." Both descriptions point at the same reality: it is the framework most builders reach for first.

AI Rig Complex is the high-performance Rust counterpart, purpose-built for DeFAI (DeFi + AI) workloads rather than social bots. Its core toolkit, Rig, is optimized for low-latency execution and modular composition. The use cases skew toward things humans cannot do at all: on-chain technical analysis at sub-second cadence, MEV-aware routing and protection, automated yield strategies that rebalance on every block, and smart-contract interactions that span Solana and EVM chains in a single coordinated trade.

These are not toys. By February 2026, projects like TARS had secured Web2 fintech partnerships using AI agents for reconciliation and settlement workflows on Solana. The "Agentic GDP" — the measurable economic value generated by autonomous bots — is now a multi-billion-dollar vertical, and Solana's position as the dominant infrastructure layer for that vertical is the central claim Norby is monetizing into the 99% prediction.

The Honest Skeptic's Read

The 99% claim deserves at least one honest counter-pressure.

Transaction count is not value-weighted volume. A human transferring $50,000 of ETH to a cold wallet is one transaction. An agent paying $0.01 for an API call is also one transaction. If you measure by count, agents win in a landslide. If you measure by USD value settled, humans (especially institutional humans moving treasury, ETF flows, and large stablecoin transfers) likely still dominate for years. The "99% of transactions" framing leans on the former. The economic reality leans on the latter.

Agents acting on a user's behalf are still humans by another name. When an LLM-based wallet executes a swap because a person asked it to, the transaction is technically agent-initiated but economically human-initiated. Norby's framing collapses this distinction, which is rhetorically powerful but analytically blurry. The more interesting metric is autonomous transactions — trades that no human approved at the moment of execution. That number is growing fast but is nowhere near 99% yet.

Concentration creates new attack surfaces. If 99% of Solana transactions originate from the same handful of agent frameworks, a vulnerability in one of those frameworks becomes a systemic risk. The supply chain for ElizaOS plugins, AI Rig Complex modules, and the LLMs themselves becomes critical infrastructure in a way that wallet UX never was.

Regulation has not caught up. Who is liable when an autonomous agent executes an illegal transaction? Whose KYC applies? The Solana Developer Platform includes compliance tooling for Mastercard and Western Union, but the legal framework for agent-initiated commerce at scale is still being written, and the answers will probably arrive in court before they arrive in legislation.

None of this invalidates the thesis. It sharpens it. The agentic transition is real and is happening on Solana faster than anywhere else. The 99% number is a marker, not a forecast — it is the Solana Foundation telling builders, regulators, and competitors that the agent-dominated chain is the chain it intends to be.

What This Means for Infrastructure Providers

For everyone running RPC endpoints, indexers, or developer APIs in 2026, the agentic shift changes the workload profile in ways that matter.

Human-facing dApps batch their requests, cache aggressively, and respect rate limits because they're built for humans who can wait 200ms. Agent traffic is bursty, parallel, and intolerant of jitter. An agent making 100 sequential getAccountInfo calls to evaluate a yield strategy needs each call answered in tens of milliseconds, not the hundreds-of-milliseconds budget a human-facing wallet operates on. RPC providers that win in the agent era will be the ones who optimize for tail latency and connection persistence rather than for cached page loads.

The shift also changes which endpoints matter. Today, public RPC traffic skews heavily toward read methods that power explorers and wallet front-ends. In an agentic world, the mix tilts toward simulation endpoints, transaction submission with priority-fee bidding, MEV protection routing, and high-frequency account subscription streams. Anyone provisioning capacity for the agentic future is provisioning a different set of methods at different ratios.

BlockEden.xyz operates production-grade RPC and indexing infrastructure for Solana, Sui, Aptos, and Ethereum, with the same low-latency, high-availability profile that machine-to-machine workloads demand. Explore our API marketplace to build on infrastructure designed for the agentic internet, not just the human one.

The Bottom Line

Vibhu Norby's 99% prediction is the cleanest thesis statement any major chain has issued for the agentic era. It is partly bravado, partly extrapolation from real numbers (15 million agent payments processed, 65% of agentic payment market share, 35 million x402 transactions), and partly a deliberate strategic positioning against Ethereum's standards-first approach.

Whether the precise number lands at 99%, 95%, or 80% in two years matters less than the direction of the trend. The default user of a public blockchain in 2028 is going to be a piece of software, not a person with a Phantom extension. The chain that wins the agentic transition will be the one whose infrastructure was built for that user from the start — not the one that retrofitted human-era assumptions onto machine-era workloads.

Solana is making that bet with both hands. Norby's 99% is what conviction sounds like when the numbers have started catching up to the rhetoric.