Onchain Banking Explained: Inside the Surgecast Episode With infiniFi's Rob Montgomery

1. The retail yield gap is a duration problem before it is an access problem, and duration is the one variable most DeFi depositors refuse to trade.
2. infiniFi splits depositors into a liquid senior tranche (siUSD) and a locked junior tranche (liUSD) that takes first loss in exchange for a higher rate.
3. The yield comes from maturity transformation alone, with no protocol leverage and no emissions, which caps the upside and removes the usual blow-up mechanism.
4. infiniFi Prime sells the ladder to neobanks as segregated instances holding only tokenised assets, because the intermediary owns the distribution DeFi rails never captured.
5. The GENIUS Act bars stablecoin issuers from paying yield, pushing the earn account off the issuer and onto asset-manager structures like Prime.

• What Does Onchain Banking Actually Mean in 2026?
• Why Has the Retail Yield Gap Lasted This Long?
• How Does a Deposit Move Through infiniFi?
• Where Does the Yield Come From, and What Breaks It?
• What Is infiniFi Prime, and Why Build It for Neobanks?
• What Does Prime Give a Neobank That It Cannot Build?
• Is the Earn Account the Real Neobank Battleground?
• What Are the Challenges and Limitations?
• What Are the Regulatory Considerations?
• What Did Rob Say About the Q4 Token Launch?
• How Surgence Labs Does This
• FAQs

Rob does not dress the model up. Asked whether infiniFi is a bank rebuilt from first principles, he agreed, then described the oldest trade in finance.

“What we’re doing is we’re just replicating what the traditional banking system does, which is borrow short, lend long asset management.”

Deposits arrive with no fixed term. The bank lends them out across durations, holds back enough to cover expected withdrawals, and keeps the difference. The part Rob objects to is the split.

“Meanwhile, they’re paying you pennies on the dollar and making a massive amount of profit since they’re lending that at like seven, 8 percent. And maybe they’re paying you 10 bips if you’re lucky. They’re just pocketing that spread.”

infiniFi runs the identical trade and hands the majority of the spread back. Depositors pick their own duration, which tells the system how much term risk the book can carry.

The mechanics are boring. The distribution of the profit is the product.

Ruthie asked why retail has never reached institutional yields that have been visible onchain for years.

Rob rejected the access framing. “Beyond just the access of like institutional yields, having institutional procedures associated with them, I’d say the main thing is really the duration component.”

Real-world asset products already publish good rates. They come with lockups, and lockups are what the market will not accept. “The dramatic majority of people who are in DeFi, they want to be able to get their capital like that. They need very fast access to it.”

That preference has a price, and depositors pay it daily without seeing the invoice. Three-month Treasury bills yielded 3.91 percent on August 31, 2026. infiniFi benchmarks its own blended return at 7.4 percent against Aave at 4 percent and Ethena at 6 percent. The delta is a minority of depositors agreeing to wait 30, 60 or 90 days on behalf of everyone else.

A useful heuristic: if a yield product cannot tell you which depositors are waiting, it is not doing maturity transformation. It is taking a different risk and calling it duration.

Sell liquidity you never priced and the first redemption queue prices it for you.

Deposit USDC, the system mints iUSD, then routes you into one of two tranches.

Stake into siUSD and your capital sits fully liquid for four to eight hours until the automated rebalancer fires. Roughly 75 percent then spreads across the ladder, with the rest held as reserve. Lock into liUSD and you choose a redemption window between one week and thirteen weeks, with a longer option near 120 days in development. Longer lock, higher rate.

At recording, Rob put siUSD near 6.5 percent and liUSD between 6.7 percent and roughly 9.5 percent at thirteen weeks.

These are not two flavours of the same risk. They are a waterfall.

Rob was direct about who pays when something breaks. “They take first loss in the event of any loss on any of the underlying assets, which so far has not occurred.”

The sizing is what to check. The locked tranche runs around $20 million, and infiniFi sizes most individual positions below that figure, so any single asset could go to zero without touching senior holders. One position sits below full coverage, and Rob named it himself rather than leaving it in the docs.

Concentration risk does not vanish because a tranche is labelled junior. It gets a name and a number.

This question separates durable products from the ones that show up in post-mortems. Rob answered it in six words.

“It just comes from a maturity transformation. That’s it.”

No leverage on the protocol’s own book. No emissions subsidising the headline rate. Illiquid positions pay more than liquid ones, and as long as the reserve covers redemptions at a reasonable frequency, the mismatch is a feature.

Then the honest part. “We’ve had, I think, four or five bank runs at this point in our life. We’ve never had a moment where we offered anything less than the same block liquidity.”

He described a February drawdown where a lender pulled out of Morpho markets after taking bad debt elsewhere, forcing borrowers to delever. infiniFi uses no leverage, but plenty of iUSD holders did. His summary: “60 mil drawdown in 48 hours. No illiquidity.”

He expects that record to break. When a one-week queue finally appears, the token starts behaving like a principal token, tradeable at a discount to whoever will hold it to maturity.

The security choices are legible. No centralised minting queue, so there is no queue to drain. A peg stability module instead, giving up the mint and redeem spread that funds competitors. The rebalancer is an off-chain key that can only move funds between whitelisted farms, and adding a farm takes a timelock.

“I could literally give you the private keys right now and you could shuffle money around and annoy me, but you couldn’t actually do anything that would allow you to financially profit off of those operations.”

Add a four-of-seven multisig on separate signing devices, an audit on every release including single-line changes, and formal verification of the core by Certora. Cover is available on Nexus Mutual, and Rob claims the premiums price cheaper than Morpho’s. “We’re going to have to go pay for an audit. We are that aggressive.”

Run this scorecard on any onchain yield product before capital moves.

Every line a team cannot answer is a line their depositors answer later, at a worse moment.

Prime came out of a question infiniFi kept getting from money that wanted in but would not touch the rails.

“They don’t want the DeFi smart contract risk. They don’t want curators,” Rob said. “We don’t want smart contract risk. We want known asset managers.”

Prime answers both objections with one structure. It is a segregated instance of infiniFi whose exposure is limited to tokenised assets from a named, regulated manager. Rob’s worked example pairs a treasury-backed stablecoin with Fasanara’s flagship fund, which Midas tokenised as mF-ONE in June 2025. The counterparty becomes a name a risk committee recognises. The partner takes the senior tranche. infiniFi Core takes the junior and harvests the risk premium. He put the headline rate on record: 6.5 percent fixed on stablecoin deposits.

The asset class this opens has been closed by structure rather than by rule. The Financial Stability Board sized private credit at between $1.5 trillion and $2 trillion in May 2026. Rob’s explanation for the closure is unglamorous and convincing.

“Most asset managers do not want to build retail access mechanisms. They don’t have distribution to retail. They don’t want to build distribution to retail because it would take them longer than they are comfortable with.”

Not every fund qualifies. Rob drew the line at gating, which caps how much can be redeemed in a period, and revocation, which suspends redemption entirely. Funds with either are a poor fit. Shorter-dated asset-backed strategies without them, trade receivables and credit card receivables among them, map onto a 30, 60 and 90-day ladder cleanly.

The route to market is the sharper choice. Rob has been in crypto since 2013 and calls himself a libertarian who loves the ideology, then concedes what building taught him. “Most people are not interested in maintaining their own private keys. They’re not interested in dealing with MetaMask or understanding what is a scam and what isn’t a scam.”

His conclusion: “I think a lot of us are now believing that the way that retail is truly going to arrive is through an intermediary.”

That is a distribution thesis, not a technology thesis, and it sits behind every crypto ETF filing and every payments integration shipped this cycle. It also changes the acquisition maths, because winning users inside an existing app is a different motion from winning them on open rails.

Build for the wallet and you compete for people who already have one. Build for the app and you inherit its user base.

Ruthie asked Rob to argue against himself: what stops a well-funded fintech assembling this in-house? His answer was a list, and the list is the moat.

The backstop improves as the network grows. If depositors want out of one Prime instance at a 1 percent fee, Core can cash them at 99 cents with a one-week wait, an arbitrage worth roughly 52 percent annualised. The same reserve serves every instance, since “it’s very unlikely that that capital is going to be needed in two places at once.”

More instances, deeper backstop, better terms for each. A flywheel, not a feature list.

The concession Rob did not volunteer, and which any integrating team should force: outsourcing portfolio construction outsources the credit judgment, not the reputational liability. When an underlying position marks down, the neobank’s users will not write to the asset manager.

Neobanks currently compete on cashback and card rewards. Ruthie asked whether the earn account becomes the fight that matters.

Rob said yes, and framed it as margin rather than marketing. “The Earn account is where their margins truly ramp up.” Cashback is a cost centre dressed as a benefit, and beating a rival by ten basis points of it improves nothing.

Sourcing yield changes the arithmetic. If competitors pay 5 percent and Prime delivers 6.5 percent, the neobank does not have to pass all of it through. “They can give them 5.5, take 1 percent, their margin just increased dramatically.” His compact version: “You can compress your competitors without damaging your margins.”

The conservative route has a visible ceiling. Treasury-backed products pay around 3.5 percent by Rob’s estimate, close to the 3.91 percent the three-month bill actually paid at the end of August 2026. That is the rate a neobank reaches without duration, roughly three points below what the ladder produces.

Asked which neobank he backs, Rob picked Revolut without much hesitation. “It’s probably still Revolut.” The numbers support the instinct: Revolut reached 75 million customers and a $75 billion valuation by July 2026, and has since launched a euro stablecoin.

Win the earn account and you win the deposit. Lose it and cashback is all you have left to spend.

Maturity transformation has one failure mode, and every bank in history has met it. Redemptions arrive faster than the ladder returns capital.

infiniFi answers with a reserve that pushes from 25 percent toward 35 percent when conditions turn, one-week positions called first, and Core standing behind the structure. Good answer. Not immunity, and Rob said so when he described the future queue as a matter of when.

The rate range is capped too. Stripping out leverage and emissions removes the mechanism that produces both 40 percent yields and total losses. Anyone hunting the top of the market will find higher numbers elsewhere, attached to risks that are harder to name.

Prime carries a different limitation. Its safety rests on the quality of the underlying manager and the honesty of the tokenised fund’s marks. Removing smart contract risk does not remove credit risk. It relocates it somewhere the reader can no longer audit onchain.

• Reading a headline APY without checking which tranche it belongs to.
• Assuming a locked position is safer because it pays more.
• Treating “no leverage at the protocol level” as “no leverage in the system”.
• Sizing a position against the protocol’s TVL rather than against the junior tranche.
• Accepting an aggregate coverage ratio when the risk sits in one asset.
• Buying a fixed rate without reading the redemption terms of the fund behind it.
• Marketing on the rate and burying the waterfall in the docs.

That last one is where projects create legal problems for themselves. It is a marketing workflow issue with regulatory consequences, rather than a regulatory issue with marketing side effects.

The most consequential fact for this category never came up in the episode, and it explains the entire Prime thesis.

Section 4(a)(11) of the GENIUS Act, enacted July 18, 2025, bars permitted payment stablecoin issuers from paying holders “any form of interest or yield (whether in cash, tokens, or other consideration) solely in connection with the holding, use, or retention of such payment stablecoin.” The OCC’s proposed implementing rule goes further, creating a rebuttable presumption of violation where an issuer arranges for affiliates or related third parties to pay that yield.

The consequence is structural. A neobank cannot source its earn rate from the issuer of the stablecoin its users hold. The yield has to come from a separate arrangement with a separate economic source, documented well enough to survive that presumption. An asset-manager structure that takes deposits, buys tokenised funds and pays a return on the portfolio is one of the few shapes that fits.

Rob felt this edge in real time. Asked about token utility, he described the long-term design, stopped himself, and said “that sounds too much like Revshare. Got to be careful with this.” Twice more he qualified an answer by referring to legal review.

That instinct belongs in the marketing brief rather than only in the term sheet. Under MiCA, marketing communications must be fair, clear and not misleading, which makes a fixed rate quoted without its loss waterfall a compliance exposure regardless of how accurate the number is.

Publish the rate and the waterfall in the same asset, or a regulator eventually asks why you did not.

infiniFi announced a Q4 2026 TGE, and Rob was deliberately careful on specifics.

The initial design will be conventional. “It will have effects that are similar to what is the current standard in token design,” he said, adding that his long-term vision goes beyond that. The interesting version is not the launch version, because the narrative is too complex to lead with. Teams planning their own token launch usually learn that lesson later and more expensively.

The long-term design is worth understanding. Rob wants to apply ve(3,3) mechanics, proven on Aerodrome and Velodrome for allocating capital across liquidity pools, to asset selection in a yield system.

“No one has tried to do anything like that with a yield system. They’ve done it with LP systems, but no one’s done it with a yield-bearing RWA-like system.”

The addition that makes it more than a copy is accountability. “What happens if you put up your capital and vote on this and it loses money? You should be the first guy to get slashed.”

Voters carrying the downside of their own allocation is a different incentive structure from most governance tokens, where voting is free and the consequences land on somebody else. Whether it survives legal review is the open question.

For depositors positioning ahead of the event, Rob pointed at the points program, which is capital-weighted and running at elevated rates in what he called overtime.

The build timeline, in his own dates:

His wish list for that final row includes BlackRock, Robinhood, Wise and Revolut. Ambitious, and he named them on the record anyway.

A token that ships on schedule is a milestone. A ladder that survives its first real queue is the product.

We work with protocols and exchanges at exactly this transition, where a product built for onchain depositors has to be sold to institutions, fintechs and neobanks that read risk differently.

The work is usually three things. Rewriting the category story so it lands in a risk committee rather than a Telegram group. Building the proof surface, meaning the audits, coverage ratios, stress history and disclosures a partner’s diligence team asks for before they ask about the rate. Then running distribution into the intermediaries that hold the users, using the same playbook behind crypto exchange marketing.

Surgence Labs has run go-to-market for Binance, Animoca Brands, Polygon, Falcon Finance, Incentiv, Rails, Biptap, SendIt, Black Mirror and Boost.

The yield gap between institutions and everyone else is not a conspiracy of gatekeepers. It is the price of liquidity, charged quietly to depositors who will not wait and collected by the small group who will.

infiniFi’s contribution is to make that trade explicit, name who takes the first loss, then sell the resulting ladder to the intermediaries that own the users. Size the reserve correctly and the structure compounds with every instance. Size it wrong and the first genuine queue writes the case study.

Rob has fifteen months, four or five stress events, and no loss on an underlying position. That is a track record, not a guarantee, and he was careful to say so.

Make sure to watch the full Surgecast episode with Rob Montgomery for the complete conversation, including the private credit fund types that fit the mandate and the ones that never will.

Onchain banking is the practice of pooling stablecoin deposits, laddering them across fixed maturities, and passing the resulting spread to depositors rather than shareholders. It reproduces the borrow-short, lend-long mechanics of a commercial bank using smart contracts and tokenised assets. The difference from a bank is where the profit lands.

infiniFi is an Ethereum protocol that recreates fractional reserve banking onchain, holding roughly $48 million in TVL as of September 2026. Depositors mint iUSD with stablecoins, then choose a liquid or locked tranche. It launched in June 2025 and is built by infiniFi Labs, co-founded by Rob Montgomery.

siUSD is the liquid senior tranche, redeemable in the same block and paying a blended ladder return near 6.5 percent at the time of recording. liUSD is the locked junior tranche, paying between 6.7 percent and roughly 9.5 percent depending on a term of one to thirteen weeks. liUSD absorbs the entire first loss before siUSD is touched.

Maturity transformation and nothing else, according to Rob Montgomery. The protocol deploys most deposits into longer-dated positions paying more than liquid ones, holds a reserve to service redemptions, and passes the difference to depositors. There is no protocol-level leverage and no emissions subsidising the headline rate.

Rob describes four to five bank-run events across fifteen months, none producing worse than same-block liquidity. The largest was a February drawdown of roughly $60 million in 48 hours, triggered by a delevering cascade in Morpho markets. He expects a redemption queue eventually, at which point the token would trade at a discount on the secondary market.

infiniFi Prime is a segregated instance of the protocol built for neobanks and consumer fintech apps, offering 6.5 percent fixed on stablecoin deposits. Exposure is limited to tokenised assets from a named regulated asset manager, removing direct DeFi smart contract risk. The partner takes the senior tranche while infiniFi Core supplies the junior.

Because the intermediaries own the distribution. Rob’s position is that most retail users will never manage private keys, and will arrive through apps like Robinhood and Coinbase that handle custody for them. Selling to the app inherits its user base rather than competing for wallet-native users.

The duration ladder built on live redemption data, junior capital it would otherwise have to source, portfolio construction handled by a regulated manager, and a liquidity backstop no single balance sheet can provide. What it gives up is control of the underlying rate, fixed at the Prime level.

Only for the right funds. Rob excludes any fund with gating, which caps redemptions in a period, or revocation, which suspends them entirely. Shorter-dated asset-backed strategies such as trade receivables and credit card receivables map onto a 30 to 90-day ladder, which is what makes a liquid wrapper viable.

Yes. Section 4(a)(11) bars permitted payment stablecoin issuers from paying holders interest or yield in connection with holding the stablecoin, and the OCC’s proposed rule creates a rebuttable presumption of violation where affiliates or related third parties pay it instead. Yield has to come from a separate economic arrangement, which is part of why structures like Prime exist.

infiniFi announced a token generation event for Q4 2026. The initial design follows current market standards, with Rob describing a longer-term ambition to apply ve(3,3) voting to asset selection, including slashing voters whose allocations lose money. Details remain subject to legal review.

Ask for the yield source in one sentence, the protocol-level leverage position, the loss waterfall with tranche sizing, per-asset coverage ratios, and dated stress history with redemption outcomes. Any line the team cannot answer is a line their depositors answer later, at a worse moment.

• How to Launch a Token: The Complete Go-to-Market Guide
• Onchain AI Agents: Inside the Surgecast Episode With KeeperHub’s Simon
• Agentic L1 Explained: Inside the Surgecast Episode With Somnia’s Paul
• How Crypto ETFs Are Changing Investor Access to Digital Assets
• Accepting Crypto Payments: How Businesses Are Integrating Digital Currency in 2026
• Crypto Market Cycles Explained: How to Time Your Project’s Growth Strategy