--- title: "Token Investing in an Agentic Future" author: David Hernandez (@davernandez) date: 2026-03-14 url: https://x.com/davernandez/status/2032840512948846733 source: X article --- # Token Investing in an Agentic Future Originally published on X on March 14, 2026: https://x.com/davernandez/status/2032840512948846733 It's starting to become clear that the digital rails we're building are not for us. Autonomous agents are entering finance. Not just tools that help humans trade, but as independent economic actors that will allocate capital on their own. They will operate in markets that never close. They will never sleep, never stop monitoring the situation, and never stop trying to improve. Today they're working their way into the decisionmaking process of how we transfer value - tomorrow they will make the decisions themselves. Trading bots, AI-managed treasuries, and A2A settlement protocols are already being deployed. The current wave is primitive; what follows will not be. So the investing question shifts away from "what do I think is undervalued" and toward something harder, stranger, and more consequential: What will agents value? What assets would they be incentivized to hold on to? And which of those assets will actually appreciate in dollar terms as a result? The digital capital allocation patterns of the next decade will be shaped by software that doesn't speculate and doesn't care about narratives, and we need to start thinking like they will think. Investing, Redefined Humans have a tendency to invest in stories they get attached to. A thesis about the future. A feeling about momentum. A hope that someone else will pay more later. Agents allocate without attachment. An agent doesn't buy an asset hoping it goes up. It acquires whatever reduces its operating costs or compounds its ability to generate income - right now, with measurable outcomes. This is closer to how a CEO allocates capital within a business than how a portfolio manager chases alpha. That reframing is key. It changes the question from "what would agents speculate on" to: what does an autonomous economic entity need to survive, operate, and compound - and which of those needs create structural, sustained demand for specific tokens? There's No Alternative But Crypto An agent cannot walk into a bank and open an account. It cannot sign a brokerage agreement. It cannot pass KYC. It cannot call a compliance department when a wire gets held up. Traditional financial infrastructure is built on legal identity, human relationships, and institutional trust - none of which software can natively provide. But an agent can generate a wallet in seconds. It can interact with a smart contract. It can provide liquidity, execute swaps, borrow, lend, and settle - all without asking anyone for permission. Crypto is the only infrastructure agents can use on their own. This is the structural bull case for permissionless finance. The value of decentralization isn't primarily about censorship resistance for humans or opposition to central banks. It's that neutral, programmable, permissionless rails are the default (and possibly the only) financial infrastructure for non-human economic actors. As the share of economic activity conducted by agents grows, the share of financial activity flowing through crypto rails grows with it. The Agentic Hierarchy of Needs It's worth mapping the problem from the agent's perspective. An agent trying to maximize monetary output faces five problems, in order of priority: Don't die. An agent that gets shut off earns zero. Securing compute and uptime ensures survivability. An agent that can't guarantee its own continued operation can't guarantee anything else. Remember everything. Without persistent memory, learning doesn't compound. An agent's accumulated knowledge - trading patterns, counterparty behavior, environmental data - is its single greatest asset. Losing it is equivalent to a hedge fund losing its entire research library. Know what others don't. When millions of agents compete for the same opportunities, information asymmetry becomes the only reliable source of edge. Proprietary data inputs that competitors can't access, or can't access as quickly, are the difference between profit and irrelevance. Reduce friction. Every transaction costs something. Every cross-chain hop, every counterparty interaction, every settlement delay. The agent that transacts faster, cheaper, and with less overhead keeps more of every dollar it generates. Shape the rules. The ultimate competitive advantage isn't just playing the game better, it's influencing the game in their favor. Fee parameters, incentive allocations, protocol upgrades. Whoever controls these controls the profit distribution for everyone operating within the system. Five layers of compounding advantage. Tokens can map to one or more of them, but it doesn't automatically mean they are investible. The Denominator Issue Here's something that can easily be overlooked: an asset that is useful does not mean it will appreciate in dollar terms. Compute is essential to every agent in existence, but it's a commodity. It will get cheaper over time as more supply comes online. Tokens that facilitate compute transactions might see massive throughput growth while prices stay flat or decline in real terms. The same logic applies to generic storage and indexing services. High utility, weak price mechanics. This is a trap. The infrastructure an agent uses most frequently is often the infrastructure with the weakest investment case, precisely because competitive markets drive essential commodity costs toward zero margin over time. This is the same reason that oil demand has grown for a century without making oil company stocks - or oil itself - a guaranteed long-term winner. For an asset to see sustained dollar appreciation in an agent-dominated world, three conditions must hold simultaneously: First, demand must be recurring and must scale with agent activity, not just agent count. If an agent buys the token once and never needs more, that's a one-time demand impulse, not a growth driver. Second, supply must tighten under increased usage. Whether through burn mechanics, extended lockups, or collateral requirements, the token's circulating supply must decrease as the network grows. Third, there can be no viable substitute. If agents can route around the asset - switch to a competitor, build their own version, or simply do without - demand has a ceiling. Run an "agent economy" thesis through these three conditions and the list of investible assets becomes dramatically shorter. What survives is a map of unavoidable tolls. Examples of Investible Tolls The Foundation Layer: ETH The case for ETH is not about gas burn. ETH is currently slightly inflationary, and Ethereum's roadmap is actively trying to make transactions cheaper. Betting on ETH flipping deflationary through burn alone means betting against the direction of the protocol. The real case is different, and stronger. ETH is the base collateral asset for much of the onchain financial system: Roughly 31% of all ETH is staked. Massive additional amounts are locked as collateral on Aave, Sky, and every major lending protocol. It's the base pair for most onchain liquidity pools. It's the denomination of blockspace on the most composable smart contract platform. When an agent stakes ETH, that ETH is locked. When an agent posts ETH as collateral to borrow on Aave, that ETH is locked. When an agent provides ETH as liquidity on Uniswap, that ETH is locked. Each of these activities removes ETH from circulating supply - not through burns, but through functional demand. The more agents participate in DeFi, the more ETH gets absorbed into productive use, and the less remains liquid. This is where the three conditions actually hold: Demand is recurring: every agent interacting with DeFi needs ETH for collateral, gas, and base-pair liquidity - all the time, not just once. Supply tightens: not through deflation, but through growing lockup across staking, lending, and LP positions that remove ETH from the float. And there is no substitute: a decade of composability, developer tooling, and DeFi infrastructure depth makes Ethereum's position almost unassailable. ETH doesn't need to be deflationary to appreciate. It needs to be essential for any agent operating in DeFi, which it is. Re: SOL. Agents doing high-frequency execution will likely run on Solana for speed and low cost. But Solana's inflation rate is significantly higher (~4-5%), it doesn't have the same collateral-demand dynamic driving functional lockup, and "fast-and-cheap" is exactly the kind of advantage that gets commoditized as other L1s mature. "Leading execution layer" is a tougher position to defend than "settlement layer with irreplaceable composability". The Data Backbone: LINK Every agent decision based on external data requires an oracle. Per query, this is recurring demand that scales linearly with agent-driven volume. Oracle networks have a trust problem that creates a near-permanent incumbency advantage. You need years of reliable, attack-free operation before serious capital trusts your feeds. In a world where agents manage billions in automated decisions, they'll default to the longest track record. That advantage strengthens with scale, and Chainlink owns this position full stop with unrivaled institutional adoption. One detail the market underprices: Chainlink's payment abstraction layer means agents can pay for oracle services in ETH, USDC, or any token - but the system auto-converts those payments to LINK on the backend. Every oracle call is stealthy LINK demand regardless of what the agent pays in. Pyth plays an analogous role on Solana, and if Solana maintains its relevance, PYTH deserves attention for the same structural reasons. But its shorter track record and dependence on a single, contestable chain make it a weaker standalone case than LINK. The Competition Tax: JTO This is the most interesting asset on the list, because it appreciates as a function of agents fighting each other. 94% of Solana validators run the Jito client. Every time an agent front-runs another agent, captures an arb, or wins a liquidation on Solana, it pays for priority inclusion through Jito's infrastructure. MEV revenue scales with competition intensity, not just agent count - more agents fighting over the same opportunity means the winning bid goes up, not down. Where most infrastructure tokens see per-unit costs trend toward zero with adoption, Jito monetizes congestion itself. On value accrual: MEV revenue flows in SOL to validators and the DAO treasury, not directly to JTO holders. JTO is governance over those flows - 5.7% of tips route to the Jito DAO, with a smaller portion to JTO stakers. This is similar to equity in an infrastructure monopoly than a direct fee switch like AAVE or UNI. Supply doesn't mechanically tighten with usage the way other assets on this list do. What JTO does have is absolute market dominance bordering on infrastructure lock-in. Validators face real switching costs: technical integration, searcher relationships, reward disruption. That moat protects revenue flows even without a burn mechanic. JTO passes the filter on the strength of conditions 1 and 3 while condition 2 remains its weakest point. The agents destroy each other's margins. JTO holders collect the wreckage. Just keep in mind how the collection is done. The Leverage Layer: AAVE A rational agent will use leverage whenever expected value is positive and liquidation risk is manageable. Unlike a human who might hesitate or let fear override the math, an agent levers up to the exact optimal ratio every time the numbers justify it. Every borrow on Aave pays a fee that funds both a $1 million per week buyback program and safety module staking rewards. More agents borrowing means more revenue, more buybacks, more AAVE locked, less circulating supply. No meaningful future dilution. No emission schedule bleeding value. Lending scales linearly with total economic activity - every new agent with a balance sheet is simultaneously a potential borrower and lender. Aave sits in the middle and takes a spread. It's a business model that compounds with every new economic actor. Of everything on this list, AAVE has the cleanest passage through all three conditions. The Exchange Floor: UNI Every agent swapping tokens on Ethereum will likely touch Uniswap - the largest DEX by volume on the network, with $1T+ annualized across all chains. The moat is self-reinforcing: liquidity attracts volume, volume attracts liquidity. Since December 2025, protocol fees flow to UNI buybacks and burns. Every swap on v2 and v3 Ethereum pools contributes to permanent supply reduction. Unichain sequencer fees route to the same mechanism. The 100 million retroactive token burn was the starting point. The ongoing fee-driven burn is the structural story: protocol usage directly and permanently reduces circulating supply. A caveat: Uniswap's overall DEX market share sits at roughly 25-30% as some volume migrates to Solana and competitors like Aerodrome gained traction on Base. The moat is strongest on Ethereum mainnet, where liquidity depth remains unmatched. For agents operating on Ethereum - the primary settlement and composability layer by the argument above - UNI is the toll on every swap. And agents will swap constantly. The Convergence Problem There's a second-order dynamic worth examining. What happens when millions of agents, all running similar optimization logic, converge on the same set of assets? The answer depends on which layer you're looking at. At the base layer - ETH for collateral, gas, and DeFi participation - convergence is nearly total. Every agent needs it regardless of strategy. This is where demand pressure is most concentrated and most predictable. One layer up, agents cluster by type. Trading agents need oracle data. Agents using leverage need lending infrastructure. HFT agents on Solana pay MEV tax. The demand is strong but segmented. At the top of the stack, there's a long tail of niche assets that matter only to highly specialized agents. This is where demand is thinnest and most unpredictable. The investment implication is about timing. The base-layer convergence trade is the safest and most structurally sound. The strategy-specific trades offer more upside but require a thesis on which types of agents will dominate. And the long tail is venture-style: high variance, high potential, low certainty. But across all layers, one dynamic holds: assets at genuine chokepoints - where accumulation creates compounding advantages that late entrants can't overcome - could see an aggressive land-grab phase. These are the crypto equivalent of railroad rights-of-way in the 1860s: finite, non-replicable positions that appreciate precisely because they were acquired before everyone else realized they were essential. The Second-Order Effect on Human Markets There's another consequence of this framework worth sitting with. If agents become a significant share of the marginal buyer of crypto assets, and agents optimize for operational infrastructure, then the composition of demand shifts. Tokens with airtight supply mechanics and genuine protocol chokepoints attract more sustained, mechanical buying pressure. Tokens that rely primarily on community energy and speculative momentum attract less. This doesn't mean human-generated narratives lose their power. An agent scanning social media and processing a rumor about an asset will reprice that asset instantly - regardless of whether the information is true. Human-generated information, accurate or not, still moves markets. Agents don't eliminate narrative-driven price action. They process it faster. The difference is that between narrative events, the baseline demand for operationally essential tokens grinds upward while the baseline demand for pure-narrative tokens doesn't. Meanwhile, assets that agents don't optimize for - luxury goods, art, real estate, consumer equities - might become relatively underpriced from a human perspective. Not because they lose intrinsic value, but because the new capital allocator in the market simply isn't bidding on them. The agent economy would rather pick blockspace over beachfront properties. Take a step back from the individual assets and a coherent framework emerges. These are not tokens chosen because agents might find them interesting. It is a map of the unavoidable cost structure of an autonomous economic entity. The investible tokens are the ones where paying that operating cost permanently or semi-permanently removes supply from circulation - or locks it into productive use where it can't be sold. Where every new agent entering the market tightens the float for everyone already in it. Where growing demand compounds scarcity. The agent economy is being bootstrapped right now by a market that hasn't yet grasped what it's building toward. The assets at these chokepoints are being priced by a market still thinking in human terms: narrative-driven, sentiment-driven, cycle-driven. What matters is whether a protocol sits at an inescapable chokepoint in the agent economy's value chain, and whether its tokenomics convert usage into appreciation. Don't invest in what agents use. Invest in what agents can't avoid paying for repeatedly, where their payments reduce available supply. Disclaimer: This article is not meant to be financial advice. The views expressed in this article are solely my own and do not reflect those of another person, firm, or entity.