Moon Pursuit’s Ahuja says AI agents will push demand toward onchain financial rails
He points to stablecoins, wallets, smart contracts and identity, as TradFi tests crypto-adjacent settlement via Lynq and derivatives licensing.
Moon Pursuit Capital founder Utkarsh Ahuja argued that an economy of autonomous AI agents will need blockchain-based financial infrastructure to move value programmatically at scale. He framed stablecoins, wallets, smart contracts and identity as the rails most likely to see early demand, alongside quiet institutional settlement and derivatives signals.
Key Takeaways
- Moon Pursuit Capital’s Utkarsh Ahuja laid out a thesis that autonomous AI agents will require blockchain and digital assets to transact programmatically at scale.
- Stablecoins were singled out as “particularly important” because they pair smart-contract automation with a familiar unit of account.
- Ahuja warned that “AI” branding and token issuance are not a business model, and that not every AI company needs a token.
- Institutional plumbing signals cited alongside the thesis included Goldman Sachs routing a roughly $100 billion Treasury fund via Lynq without tokenization and a Cboe/S&P licensing extension that leaves room for tokenized options.
Why Ahuja Thinks AI Agents Need Programmable Money, Not Just Better Payments
Ahuja’s claim is mechanical: if AI agents become numerous and autonomous, the limiting factor stops being model quality and becomes transaction throughput. An AI agent, in his framing, is software that can take actions on its own within human-set parameters, including negotiating with other agents, buying compute, paying for data, and executing transactions.
That workload looks less like a human clicking “send” and more like continuous machine-to-machine commerce. Ahuja argues legacy payment infrastructure was built for people and institutions initiating transactions, not for “potentially millions of autonomous software agents conducting low-value transactions continuously across borders.” The consequence is friction: approvals, batching, intermediaries, and settlement timelines that are tolerable for humans become failure points when software is trying to transact at internet cadence.
Blockchains, by contrast, make value programmable in the same way APIs made information programmable. Ahuja’s mechanism is that an agent can hold a wallet, call a smart contract (code on a blockchain that executes when conditions are met), and transfer a stablecoin without the manual intervention layers that sit between most traditional payment flows and final settlement.
Where the “Machine Economy” Could Show Up First: Stablecoins, Wallets, Identity and Settlement
Ahuja’s investable map is not “AI tokens win,” it is “rails win.” He explicitly cautioned against narrative premiums: “That does not mean every AI company needs a token, nor does it mean attaching “AI” to a crypto project suddenly creates value.” The filter he proposes is whether a project solves a concrete automation-driven constraint: payments, settlement, identity, cybersecurity, custody, and rails that connect traditional and digital markets.
Stablecoins sit closest to the default primitive in that stack because they keep the unit of account familiar while still being software-native. Ahuja called them out directly: “Stablecoins are particularly important here because they provide a bridge between blockchain's programmability and a familiar unit of account.” If agents are going to transact continuously, the argument goes, they need something that behaves like cash, clears quickly, and can be embedded into contract logic.
Identity and provenance are the other early pressure points in his thesis. Identity, here, is not a login screen. It is the ability for markets to determine “who or what is behind a transaction” and what an agent is authorized to do. Provenance is the verifiable record of where data or assets came from and who controlled them, which becomes harder when AI systems consume data, generate IP, and execute trades. Ahuja’s claim is not that blockchains solve the whole identity stack, but that verifiable shared records make them a “natural part of the infrastructure stack” when counterparties need auditability.
The capital-markets angle is where the thesis stops sounding like payments commentary and starts sounding like market structure. Tokenization, in Ahuja’s framing, is already pulling traditional assets onto blockchain rails, stablecoins have already proven meaningful activity can run onchain, and AI is automating decision-making. If those trends converge, an agent that can analyze markets but cannot efficiently hold, exchange, or settle assets becomes operationally capped.
TradFi’s Quiet Plumbing Moves: Goldman via Lynq and Cboe/S&P’s Tokenized-Options Optionality
Ahuja paired the agent-rails thesis with institutional signals that point in the same direction: settlement first, tokenization later. He described Goldman Sachs’ FTIXX Treasury fund as being routed at roughly $100 billion scale to institutional crypto trading firms through Lynq, a settlement network used by digital-asset companies, without tokenizing the fund itself. The important detail is the “without tokenization” clause. It suggests incumbents can adopt crypto-adjacent plumbing for cash-and-collateral movement even when the asset wrapper stays traditional.
He also flagged a licensing extension between Cboe Global Markets and S&P Dow Jones Indices that “left room” to explore tokenized options products. That is not a product launch, but it is a permissioning step that matters in derivatives, where index licensing is part of the distribution pipeline. If tokenized options ever ship, the gating items will be filings, market structure, and counterparties, not a whitepaper.
The market tape in the same note framed the setup as medium-horizon rather than an immediate catalyst. Bitcoin fell to $82,500 on Monday and recovered above $84,000 overnight as yields steadied after the 10-year Treasury yield briefly moved to 5.2%. Spot bitcoin ETFs took in $30 million on Tuesday after $2.84 billion over six sessions the prior week.
Ahuja’s chart snapshot also leaned into tokenization as a live, if still concentrated, flow. Tokenized-equity DEX flow was described as consolidating into bStocks and Robinhood while Backpack’s share dropped to about 2%. BP’s token price was cited as climbing to about $1.35, framed as the market paying up for “the one licensed, tradable token proxy for the theme.”
The AI agents need blockchain financial rails Milestones Ahead
The near-term validation path for Ahuja’s thesis is not another agent demo. It is usage data on the rails he named, and the policy and institutional steps that make those rails usable at scale.
Stablecoin adoption and policy milestones matter most because issuance and redemption mechanics determine whether stablecoins can function as settlement cash for automated flows. If stablecoins become easier to mint and redeem against regulated banking rails, onchain settlement becomes less of a closed loop.
The Goldman-Lynq datapoint needs primary-source confirmation and more structure to be tradable as a signal. Timing, counterparties, and whether the routing expands beyond the cited roughly $100 billion scale would clarify whether this is a one-off operational channel or the start of a repeatable institutional workflow.
On tokenized options, the follow-through would be visible in filings, pilots, or named exchange partners. A licensing agreement that leaves room to explore is optionality, not commitment.
Finally, the tokenized-equity flow concentration is worth tracking as a market-structure tell. If bStocks/Robinhood dominance persists while Backpack stays near the cited ~2% share, liquidity and distribution may be consolidating around a small set of venues, and BP’s price and volume behavior becomes a proxy for how much the market is willing to pay for regulated access to the theme.
My Take: The Trade Is Shifting From “AI Winners” to Transaction Rails—But Proof Will Be in Usage Data
The threshold that matters is whether autonomous agents actually become numerous enough that payments and settlement turn into the bottleneck. If that happens, the scarce asset is not “intelligence,” it is throughput and permissioning: wallets that can be controlled safely by software, stablecoins that can be redeemed reliably, and settlement paths that clear without a human in the loop.
This still reads more like a positioning theme than a catalyst trade because the institutional signals cited are plumbing moves and licensing optionality, not shipped tokenized products. If stablecoin settlement volumes and institutional routing through networks like Lynq start compounding in a way that can be verified outside commentary, the rails thesis stops being narrative and starts being measurable infrastructure demand.
