AI already runs the monitoring, the simulations, and a lot of the automation behind on-chain risk. Here is what it still can’t do on its own, and why your yield still rests on human judgment.

Here is the short answer. AI already manages real parts of on-chain risk today.

It runs simulations, watches markets every second of every day, and adjusts protocol settings inside limits that people approve.

What it does not do yet is run the whole risk desk on its own. So the honest take on AI risk management in DeFi is this: yes for the hard, repetitive parts, and not yet as a manager you leave alone.

That gap matters more than the hype suggests, so let me walk through what AI does, where it hits a wall, and what measured on-chain risk really looks like.

The quick version

• AI already runs simulation, monitoring, and bounded automation across DeFi.
• Firms like Gauntlet and Chaos Labs provide much of that risk plumbing.
• Automated systems act only within human-set, governance-approved limits.
• The real limits stay: stale data, models that get gamed, and crises no model has seen.
• Over-automation can spread risk, not contain it. AI augments risk management; it does not replace it.

What “managing on-chain risk” actually means

Managing on-chain risk means keeping a protocol solvent and liquid while markets move.

In plain terms, that is four jobs. Set the rules: collateral ratios, borrow caps, interest curves. Watch everything: positions, liquidity, oracle prices, concentration.

Stress-test: model crashes and attacks before they happen. Act fast: change settings the moment conditions change.

It is constant, data-heavy work, and the cost of getting it wrong is bad debt. No wonder teams want to automate it. It is also why a careless version is dangerous.

Where AI already runs the risk desk

AI already handles the simulation, monitoring, and bounded automation behind DeFi risk, and it is now core infrastructure. A few concrete examples, as of late 2026.

Gauntlet builds simulation-driven risk models and has acted as the economic risk brain for major lending markets, tuning parameters like collateral ratios and caps. It now runs some of the largest curated vaults too.

Chaos Labs runs AI-powered economic and security stress tests. Its risk oracles watch indicators in real time and push parameter changes within governance-approved bounds.

Aave’s renewal with the firm described more than 1,100 parameter updates.

AI-curated vaults from firms like Chaos Labs, Gauntlet, and Sentora now sit behind consumer products, rebalancing capital across Aave, Morpho, and Sky inside set risk targets.

The division of labor is simple. The machine does the speed: monitoring, simulation, anomaly detection, bounded adjustments.

The human keeps the judgment: which scenarios to run, which thresholds matter, which bounds to allow. This is real automated risk management in crypto, not a demo.

The part AI still can’t do

AI cannot be trusted to manage on-chain risk unsupervised, and 2026 showed why.

Start with the limits that never go away. It works inside human bounds, so that judgment stays human.

Models depend on data and oracles, and a stale price turns a smart system into a confident fool.

Models can be gamed by adversaries who know which inputs break them. And tail events escape training, because the crisis that matters most is usually the one the model has never seen.

The year drove this home. A string of lending blowups traced back to hardcoded or stale oracles.

In one case a collateral token was marked on-chain near $1.13 while it traded around $0.63 on the open market.

That is not a model being clever. That is automation acting fast on a wrong number.

The uncomfortable twist: automation can add risk. Fast, automated allocation paired with slow or sloppy controls, or a risk curator that itself fails, can spread a problem instead of containing it. Speed cuts both ways.

Governance-safe automation: humans still set the bounds

The norm today is not autonomous AI. It is governance-safe automation: people set the strategy and the limits, and AI acts inside them.

People decide the risk appetite and approve the guardrails. AI monitors, simulates, and makes bounded changes, often with a validation step before anything goes on-chain. And a human brake always exists.

You keep automation’s speed and keep accountability. Handing a model the power to act, with real money and real adversaries on the other side, should not be done blind.

What measured on-chain risk looks like: a Sky case study

The real answer to on-chain risk is not add AI. It is make the risk measurable. Sky is one of the clearer examples of that in stablecoins.

A quick primer, since the names trip people up. USDS is Sky’s fully backed stablecoin and the way into the system. It is the unit of account, not the yield.

sUSDS is what you hold to earn: you supply USDS, receive sUSDS, and it accrues the Sky Savings Rate over time. It is currently the largest rate-bearing stablecoin by supply.

The Sky Savings Rate is the governance-set yield behind sUSDS, funded by revenue from the Sky Agent Network, an independent set of capital allocators who put USDS to work across strategies like onchain loans, Treasury bills, and lending liquidity.

What makes it measured rather than trust me? USDS is over-collateralized and backed by diversified collateral, so the backing is not a single bet.

The yield is risk-adjusted and governance-set, not whatever the market will pay this week.

Sky is the only stablecoin protocol with credit ratings from both S&P Global and Moody’s.

Both are sub-investment grade, S&P at B- and Moody’s at B3, which is the honest part: independent agencies see real credit and concentration risk, and they publish it.

The infrastructure also carries a decade-long track record with no smart-contract exploit, and governance is building a $150 million solvency reserve.

Adoption is following. In September 2026, Galaxy Digital, a Nasdaq-listed firm, put $100 million of sUSDS on its balance sheet and approved it as institutional collateral.

And yes, this ties back to AI. Sky’s own 2026 roadmap names AI-driven automation as a growth theme, alongside Sky Agents and tokenized credit.

The through-line is the same: use quantitative, verifiable risk work, not guesswork.

If you want to go beyond the savings rate, Sky Vaults and stUSDS sit further out on the risk curve, and Sky Ecosystem Rewards layers in token incentives.

One honest caveat for savers: the Sky Savings Rate is variable. It has moved from roughly 8.75% in early 2025 to around 3.6% in late September 2026, and governance can change it. Read the live number before you decide anything.

The agent frontier, and why it’s still risky

The next step is AI agents that do not just advise but act, holding permissions and moving funds on-chain, including around tokenized real-world assets. Promising, and early.

Every limit above gets sharper when the model can pull the trigger. An agent that can act can also be manipulated or prompt-injected. A misbehaving agent causes harm directly, not as a suggestion.

Safety research keeps landing on the same answer: strict, often deterministic guardrails and clear permissions, not unsupervised control.

Full autonomy, with your money, is not here, and it is not obviously what you want.

So, can AI actually manage on-chain risk?

Yes for the parts that need speed and scale. Not yet as a manager you can leave alone.

AI already simulates, monitors, and adjusts risk faster than any human team, and it is now core plumbing for major protocols.

But it works inside human-set bounds, it can be wrong or gamed, it struggles with the crises that matter most, and letting it act unsupervised introduces fresh risk.

The grown-up version of on-chain risk is not a black box making the calls. It is measurable risk, verifiable backing, and humans who own the guardrails, with AI doing the heavy lifting in between.

AI can manage parts of on-chain risk as augmentation inside human guardrails. It is not yet a risk manager you can walk away from.

Frequently asked questions

Q. Can AI manage DeFi risk?

A. It already manages important parts, like simulation, monitoring, and bounded parameter changes, but inside human-set guardrails, not as a fully autonomous manager.

Q. What does AI actually do in DeFi risk management?

A. It stress-tests parameters, watches markets in real time, auto-adjusts limits within approved bounds, rebalances curated vaults, and flags anomalies faster than people can.

Q. What are the limits of AI risk management in DeFi?

A. It only acts within human bounds, depends on data and oracles, can be gamed, struggles with novel tail events, and over-automation can itself spread risk.

Q. How does Sky approach risk?

A. Like other protocols, Sky relies on quantitative, increasingly AI-assisted risk modeling to set the parameters behind USDS and the Sky Savings Rate, and its risk is also measured externally with credit ratings from both S&P and Moody’s. That does not make it risk-free.

• • •

On-chain risk is getting more measurable every year. That is the real story, bigger than any single model. If you want to see what governance-set, verifiable yield looks like up close, you can put your stablecoins to work on sky.money and read the live numbers on the dashboard.

AI Risk Management in DeFi: What It Can and Can’t Do was originally published in Coinmonks on Medium, where people are continuing the conversation by highlighting and responding to this story.

Monthly volume went from $306M to $1.22B. Everyone wrote the payments story. Almost nobody wrote the balance story.

Nobody writes a headline about a balance.

We write about the swipe. A coffee bought with USDC in Lagos. A ride paid in USDT in Sao Paulo. Those make good copy.

But a card is not a swipe. A card is a balance that occasionally gets smaller.

And over the last twelve months, the balances sitting behind stablecoin cards quietly turned into real money.

Quick answer, for anyone who searched “stablecoin card spending”: monthly stablecoin card spending hit $1.04 billion in July 2026, roughly triple the $306 million recorded a year earlier, and $1.22 billion by September. Cumulative tracked spending has passed $10.9 billion. Dollar-backed stablecoins fund about 70% of it. The whole category is still only around 0.07% of the global card market.

How Fast Did Stablecoin Card Spending Actually Grow?

Faster than the coverage suggested, and from a base so small that almost nobody treated it as a trend.

The tracked monthly numbers for 2026:

• January: $574M
• April: $768M
• July: $1,038M, the first month above a billion
• September: $1,224M, the third straight month above a billion

Three more figures worth holding onto:

• Cumulative tracked card spending crossed $10.9 billion by late August 2026.
• July alone carried more than 10 million individual card transactions.
• Visa reported more than 160 stablecoin-linked card programmes live or in development as of June 2026.

Concentration matters here too. RedotPay alone generated $401.9 million of September volume, roughly 34% of the entire category, and says it now serves more than eight million users.

Ether.fi’s card did $100.3 million in July purchase volume, up from under $10 million two months after launch in 2025.

Why Did Nobody Call Stablecoin Card Spending a Trend?

Two reasons, and both come down to where people were looking.

First, tripling from a small base does not feel like news. It feels like a chart that was always going to go up and to the right.

Second, the growth is not happening where crypto media usually files copy from.

StraitsX, a Visa partner that helps other firms launch cards, reported gross transaction value on its rails grew roughly 600% in lower-income markets between March 2025 and February 2026, against about 150% in wealthier ones. Binance said average volume on its Brazil card climbed 80% in a year.

John Timoney of Rain put it plainly at Consensus Miami: stablecoin cards are under 1% of global card spend today, but some Latin American markets could reach double-digit share soon.

And what people are buying is boring in the best way. Food and retail were the largest spending categories on StraitsX’s rails. Groceries, ride-hailing, food delivery. Not trades.

The adoption is real. It is just happening in markets where a dollar balance is a hedge, not a hobby.

Which Dollars Are Funding the Swipe?

Mostly two of them.

• USDC funded 50.8% of July 2026 card volume, up from roughly 48% a year earlier.
• USDT funded 20.3%, up from about 7%.
• The euro-backed EURe, which was as much as 88% of tracked card spending in early 2024, has fallen below 2%.

That last line is the quiet structural change nobody wrote up. The category started as a European experiment and became a dollar product.

What Is a Stablecoin Balance Actually Doing Between Swipes?

Here is the part the payments coverage skipped.

Every card needs a funded balance behind it. You do not spend $1.22 billion a month without holding considerably more than that across millions of wallets, waiting.

So ask the obvious question. What is that money doing on the other 29 days of the month?

For most stablecoin balances, the honest answer is nothing.

• A plain stablecoin carries a fixed $1.00 face value.
• The reserve behind it usually sits in short-term government paper.
• That return accrues to the issuer, not to the holder.
• You hold the dollar. Somebody else holds the yield.

That was a tolerable trade when stablecoins were trading chips. It is a much stranger trade when stablecoins are your grocery money.

How Much Idle Stablecoin Capital Is Sitting Still Right Now?

Total stablecoin supply sat near $302.8 billion in September 2026. Card spending is running around $1.22 billion a month.

Run the division. Card spending is roughly 0.4% of outstanding stablecoin supply each month.

Analysts tracking stablecoin velocity have been blunt about the rest. Most supply is parked rather than circulating, and only a thin slice is held in yield-generating form.

Yield-generating designs are still a thin slice of that base, under $5 billion by most counts, even though they drove more than half of net new stablecoin supply growth in the first quarter of 2026. Demand is clearly not the constraint. Distribution is.

The spending number is the one that gets written up. The 99.6% sitting behind it is the one that either compounds or does not.

What Is the Sky Savings Rate, and Why Should a Spending Balance Care?

This is where Sky Ecosystem fits into the picture.

Sky Ecosystem is a global savings and capital allocation network, and the mechanics are deliberately unglamorous:

• USDS is the base stablecoin and the unit of account. You convert into it 1:1 through Peg Stability Modules, with no fees and no slippage.
• The Sky Agent Network, a set of independent allocators including Spark, Grove, Obex and Osero, competes for access to USDS liquidity under risk limits published in advance.
• Those allocators pay for that access, and their payments contribute to Sky Protocol revenue.
• A share of that revenue funds the Sky Savings Rate, a variable rate set by Sky Governance rather than by borrowing demand in any one market.
• sUSDS is how you hold it. Supply USDS, receive sUSDS, and the rate accrues programmatically. No lockup, no exit fee, redeemable on demand.

Sky Governance moved the Sky Savings Rate up in early October 2026, from 3.60% to 3.80%. It is variable by design, so check the live figure before quoting it anywhere.

sUSDS is currently the largest yield-generating stablecoin in the market. That matters less as a bragging right than as a liquidity fact.

Size is what lets a spending balance move in and out without slippage, which is the whole requirement for money you might need on a Tuesday.

The dollar you spend from and the dollar that accrues do not have to be two different dollars.

Is Institutional Money Already Treating It This Way?

Yes, and it happened three weeks ago.

On 23 September 2026, Galaxy added $100 million of sUSDS to its corporate treasury and approved sUSDS as eligible collateral across its institutional trading business.

Read that as a signal rather than a press release. A trading desk does not put nine figures into a yield-generating stablecoin as an experiment.

It does it because the position stays liquid, the collateral is verifiable onchain, and the balance does something while it waits.

That is the same logic a card balance needs. Just with fewer zeros.

The numbers behind it are public rather than asserted: $14.15B in Total Protocol Collateral against $11.48B in stablecoin supply, all checkable on the dashboards before you believe a word of this post.

What Could Break This Argument?

I would rather name the friction than pretend it is not there.

• The rate is variable. Sky Governance can vote the Sky Savings Rate down as easily as up.
• Card programmes mostly do not settle in yield-generating tokens yet. Today the plumbing runs on USDC and USDT.
• Regulation is live. A US Senate draft proposed banning yield on idle stablecoin holdings outright, and European regulators have been weighing their own restrictions.
• Smart contract risk and governance risk do not disappear because a balance is convenient.
• Cards are a convenience product. Nobody adds a conversion step to buy coffee, which is why this only works if the yield-generating token is the default balance rather than an extra one.

None of that changes the structural point. It means the structural point is contested, which is the only kind worth writing about.

RedotPay projects the category reaches $50 billion a year by 2028. Whether that number lands or not, the balance question gets louder either way.

Stablecoin card spending tripled and the category barely blinked.

My read is that the spending is the visible half. The balance behind it decides whether any of this is actually better than the card already in your pocket.

Two questions, and I will reply to every answer. Would you hold a spending balance in a yield-generating stablecoin, or does the extra moving part kill it for you? And if you used a stablecoin card this year, what did you actually buy with it?

Stablecoin Card Spending Tripled in a Year and Nobody Called It a Trend was originally published in Coinmonks on Medium, where people are continuing the conversation by highlighting and responding to this story.

Onchain markets borrowed the word for years without earning it. In 2026, the three pieces finally showed up in the same place.

Ask a bond desk to price a five-year loan and the first thing they reach for is a curve.

Ask onchain markets to do the same a few years ago and there was nothing to reach for. One floating rate per protocol. No maturities. No term.

The global fixed income market runs about $143 trillion on curves. The entire onchain lending market, roughly $54 billion of it, was running on scattered dots.

That gap started closing this year. Not with an announcement. With plumbing.

What Is a DeFi Yield Curve, and Why Did Onchain Markets Go Years Without One?

A DeFi yield curve is the term structure of onchain rates. What the same dollar earns at one month, three months, six months, a year.

One number is a rate. A set of numbers across time is a curve.

Traditional markets get this handed to them. Governments issue debt at fixed maturities, so the curve draws itself.

Onchain had none of the ingredients:

• No anchor rate that anyone agreed to price off
• No dated instruments to plot against that anchor
• No way to finance a position sitting on the curve

Researchers noticed before marketers did. A 2025 paper called crypto a bondless market and reverse-engineered curves out of options and futures, because nothing native existed.

Building one from scratch means building all three. Sky Ecosystem is one of the few places where all three now sit in the same system.

Why Does Every Yield Curve Need an Anchor Rate Before Anything Else?

Every curve starts at the short end.

In Sky Protocol, that anchor is the Base Rate. It is the wholesale cost of capital that independent allocators pay to borrow USDS, and it is not spat out by a utilization formula.

It is set through Sky Governance, informed by collateral levels and the external rate environment.

Everything else hangs off it:

• The rate paid to sUSDS holders sits a documented spread below the Base Rate
• The rate independent allocators receive sits a narrower spread below it
• Inside that spread, part funds distribution and part is retained by the protocol

That is what a policy anchor looks like. One published decision, with the rest of the stack priced off it.

Sky Governance also moves it in public. In Q2 2026 it trimmed the Sky Savings Rate from 3.75% to 3.60% mid-quarter, calibrated to keep reserve accumulation on pace.

The live level always sits on the financial dashboard, never in a blog post.

How Does the Sky Savings Rate Work as the Floating Point on an Onchain Term Structure?

The Sky Savings Rate is the floating leg of the curve. Hold sUSDS and you hold it.

It is not a farm and it is not a promise. It is a governance-set rate funded from total Sky Protocol revenue: allocator fee payments, vault stability fees, tokenized treasury yield, peg module fees.

Scale matters, because a reference rate nobody holds is not a reference rate:

• sUSDS closed Q2 2026 at $5.52B, up 149% year over year from $2.22B
• Cumulative Sky Savings Rate distributions to holders crossed $250M on June 29, 2026
• Protocol Collateral stood at $14.15B in early October 2026, with USDS supply at $11.48B

Depth is what lets a rate act like a benchmark instead of a quote.

Where Do the Fixed Points on a DeFi Yield Curve Actually Come From?

A floating rate on its own is one dot. You need dated points.

Those came from yield tokenization. Split a yield-bearing stablecoin into a principal token and a yield token, and the principal token trades at a discount that encodes a fixed rate to a specific date. Buy the discount, hold to maturity, and the rate is locked.

sUSDS got its own maturities. The first fixed-rate market on it crossed $50M in TVL in under two weeks.

By late September 2026, the 26 November sUSDS principal token was quoting around 4.83% fixed, with $25M of size filling near 4.66% effective.

Plot that against the floating rate and something new appears. Two points. Different dates. Upward sloping.

The wider market is moving in the same direction:

• Four fixed-rate lending venues launched or expanded between May and September 2026
• Coinbase shipped fixed-rate onchain borrowing to 110 million users in September 2026
• Across assets, principal tokens were paying 5% to 14% fixed to maturity as of 17 September 2026
• Fixed-rate lending is still only about $130M of a $54B onchain lending market, near 0.24%

That last line is the whole opportunity compressed into one number.

What Turns a Yield Curve From a Chart Into a Market?

Charts are easy. Financing is hard.

A curve becomes usable the moment you can borrow against the paper sitting on it. In government bond markets that job belongs to repo. Repo is why the Treasury curve is infrastructure and not a picture.

Onchain got a version of that in September 2026. Galaxy Digital placed $100 million of sUSDS on its own balance sheet and approved sUSDS as eligible collateral across its institutional trading business, which serves more than 1,600 counterparties against an average loan book near $1.4 billion.

Galaxy called itself one of the first listed companies to hold sUSDS as treasury.

The detail most coverage skipped is the one that counts. Clients who pledge sUSDS keep earning the Sky Savings Rate on the full position for the life of the loan.

Collateral that keeps paying is how a yield curve stops being a chart and starts being a market.

Who Actually Generates the Yield Behind the Sky Savings Rate?

Nobody prints this. Somebody earns it.

The Sky Agent Network is a set of independent capital allocators that borrow USDS from Sky Protocol and put it to work.

Spark operates in lending markets. Grove handles institutional tokenized credit. Keel, Obex and Osero run their own mandates.

They are sovereign businesses, not subsidiaries. They borrow at the Base Rate, keep whatever spread they earn above it, and settle fees back to the protocol on a monthly cycle.

Those payments pool with the rest of protocol revenue, and Sky Governance calibrates the Sky Savings Rate against that total capacity, not any single strategy.

Two scale markers. Grove alone provides Galaxy with a $500 million warehouse facility.

And Sky Protocol holds anchor positions in tokenized funds from BlackRock and Janus Henderson.

No single counterparty sets the front of the curve. That is the design, not a side effect.

Do the Numbers Support a Stablecoin Benchmark Rate That Holds?

A rate only works as a benchmark if the thing paying it keeps paying.

From Sky Frontier Foundation reporting:

• Q1 2026: $123.79M in Gross Protocol Revenue and $46.04M in Net Protocol Surplus
• Q2 2026: $107.35M in Gross Protocol Revenue and $33.29M in Net Protocol Surplus, a fifth consecutive positive quarter
• First half 2026: $231.66M in Gross Protocol Revenue, with $55.18M remitted to Sky Reserves at a 43.5% net margin
• Protocol Collateral up 45.5% year over year at Q2 close

A Grayscale analysis in June 2026 ranked Sky Protocol fourth among the highest revenue onchain applications globally.

Every figure above can be checked. The financial record sits on one dashboard, the onchain state on another.

A benchmark rate that cannot be audited is just marketing with a decimal point.

Is the Sky Savings Rate Becoming DeFi’s Reference Rate, or Not Yet?

This is where people oversell.

A reference rate is never declared. It gets adopted. The test is whether others price off it without being asked.

Evidence they are:

• Fixed-rate venues now quote dated maturities against it
• A listed company holds the yield-bearing stablecoin as treasury and lends against it
• Independent allocators price their own economics off the same anchor

Evidence they are not, yet:

• No standard curve publisher exists for onchain rates
• Liquidity thins out badly past twelve months
• Most onchain credit is still floating by default, priced off utilization

Both lists are true at once. Anyone showing you only one is selling something.

What Still Has to Be Built Before the Onchain Yield Curve Is Finished?

Three gaps, plainly stated.

• Depth. $130M of fixed-rate lending is a rounding error next to $143 trillion of global fixed income.
• Tenor. The curve runs short. Multi-year onchain paper is still rare and thinly traded.
• Standardization. Different venues, different conventions, no agreed daily snapshot anyone can cite.

None of it is unsolvable, and all of it is unglamorous. Curves get built by plumbing, then everyone pretends they were always there.

If you want the inputs instead of the narrative, start with USDS, then sUSDS, then read the dashboards.

Now the part worth arguing about. If an onchain curve really does get built, what should the anchor be: a governance-set rate, a market-clearing rate, or something pinned to off-chain policy? Each one breaks in a different way. I have a view. I would rather hear yours first.

What a DeFi Yield Curve Looks Like When You Build One From Scratch was originally published in Coinmonks on Medium, where people are continuing the conversation by highlighting and responding to this story.

Today is the last day of a $750 million experiment. Here is what it proved about where stablecoin yield actually comes from.

Tonight, something quietly ends.

From October 1, 2026, Ethena stops paying ENA token rewards to anyone holding USDe or sUSDe.

The protocol announced it on September 26. Incentives tied to USDe growth had already fallen roughly 85% since the first airdrop in 2024. Now they go to zero, permanently.

Here is the price tag. More than $750 million in rewards distributed since launch.

And here is what happened to the thing those rewards were buying.

USDe peaked near $15 billion in October 2025. By late August 2026 it had dropped below $5 billion. A contraction of more than 65% in under a year.

This is not a scandal. Nobody got hurt. The peg held.

It is a receipt. And it is the clearest lesson onchain dollars have handed us in years.

For scale: in September 2025, USDe alone held roughly 4.5% of the entire stablecoin market at a $14.4 billion market cap. A year on, it is a fraction of that.

Why did the USDe supply decline happen this fast?

Three things stacked up at once.

• The subsidy switched off. Token rewards were the growth engine. Capital that arrives to farm rewards leaves when the rewards shrink. Ethena said so itself.
• Funding rates cooled. USDe earns from a delta-neutral basis trade, long spot and short perpetual futures. That pays well when leveraged traders crowd the long side. It pays very little when they do not.
• October 2025 forced the unwind. The crash triggered a deleveraging event. Supply fell from above $14 billion to roughly $8.5 billion within weeks, then kept sliding.

CoinGecko put USDe at $5.9 billion at the Q1 2026 close. By September 28, 2026, readings sat near $4.9 billion.

One more detail worth holding onto. A governance proposal approved in September ties future ENA buybacks to USDe supply climbing back above $7.5 billion. From around $4.9 billion, that is roughly 53% of growth still to find.

Ethena kept every promise it made about the mechanism. The mechanism simply did what that kind of mechanism always does.

What happens to stablecoin yield when funding rates go quiet?

It compresses. Fast.

sUSDe paid above 35% at its Q1 2024 peak. By mid-September 2026 it sat near 5%.

Now put that next to the risk-free floor. The effective federal funds rate was 3.63% on September 11, 2026.

So a strategy carrying derivatives exposure, exchange counterparty exposure and hedge execution risk was paying roughly 1.4 points over cash.

A yield that depends on other people staying leveraged is a yield you are renting, not one you own.

That is the real story inside the usde supply decline. The model did not break. It was always cyclical. The cycle finally showed up on the chart.

Where does a durable yield-bearing stablecoin get its rate?

This is where the two designs part ways.

A synthetic dollar sources yield from one market condition. A capital allocation network sources it from many.

USDS is the fully backed stablecoin at the base of Sky Protocol. It is the unit of account and the credit facility for allocators. Holding it pays nothing on its own.

Supply USDS and you receive sUSDS, the yield-generating stablecoin that carries the Sky Savings Rate.

Value accrues against USDS continuously. No lockups, no cooldown, convertible back on demand.

The Sky Savings Rate is governance-set, not market-set. That single distinction is the whole argument.

Here is the actual path the money takes:

• Sky Protocol makes USDS liquidity available under risk parameters set in public
• Independent allocators in the Sky Agent Network draw on that liquidity
• They deploy it across credit, lending markets, treasury strategies and tokenized funds
• Fees flow back as Gross Protocol Revenue
• Sky Governance sets the rate from that revenue base
• sUSDS holders capture it automatically

Six steps. Not one of them needs a perpetual futures funding rate to stay positive.

Is diversified collateral more durable? Check the 2026 tape.

Fair challenge. So here are the numbers, published by the Sky Frontier Foundation.

• Q1 2026: $123.79 million in Gross Protocol Revenue and $46.04 million in Net Protocol Surplus. Record quarter.
• Q2 2026: $107.35 million in Gross Protocol Revenue and $33.29 million in Net Protocol Surplus. Fifth consecutive positive quarter.
• Protocol Collateral: up 45.5% year over year to $12.32 billion at the Q2 close.
• sUSDS: up 149% year over year to $5.52 billion at the Q2 close.

Category context matters here too. Yield-bearing stablecoins peaked near $22.7 billion in March 2026 before native-yield supply contracted roughly 15% in Q2.

sUSDS came through that stretch as the largest single yield-bearing stablecoin in the market.

Now the part most brands would quietly skip.

In the August 2026 update, Gross Protocol Revenue came in 11.47% below August 2025. Protocol Collateral closed the month at $11.10 billion, stablecoin supply at $9.53 billion.

So no. Diversification is not immunity. Rates fall, revenue follows.

The difference is the shape of the fall. One model shed two thirds of its supply.

The other posted five straight positive quarters and still grew Protocol Collateral 18.2% year over year.

Who is allocating into USDS and sUSDS right now?

Names, not adjectives. As of September 1, 2026, Sky Agents held roughly:

• $1.23 billion with Janus Henderson
• $618.32 million with BlackRock
• $304 million with Galaxy
• $239.60 million in PayPal USD
• $220 million with Anchorage
• $103.11 million with Securitize

Galaxy is the mover here. Exposure sat at $27 million at the Q2 close and reached about $304 million by September 1.

Then on September 23, Galaxy Digital added $100 million of sUSDS to its own corporate treasury and approved sUSDS as eligible collateral across an institutional business carrying a $1.4 billion average loan book.

Token Terminal data as of August 21 placed Sky Protocol at $4.6 billion in tokenized funds. Largest single issuer in a $34.4 billion category, ahead of Securitize, Circle, Franklin Templeton and Ondo Finance.

Across the full Sky product suite, $7.1 billion was reported on August 19, with $4.92 billion of it sitting in the Sky Savings Rate.

A synthetic dollar’s collateral is a hedge. This collateral is a list you can read.

Five questions to ask before you trust any stablecoin yield

Steal this checklist. Run it on every token in the category, including this one.

• What pays the yield? Name the source out loud. If the honest answer is “market conditions,” you are holding a cyclical asset and should size it that way.
• Who sets the rate, and how? Market-set rates move with sentiment. Governance-set rates move on a vote you can read before it happens.
• Is the backing verifiable right now? Not quarterly. Right now. Sky’s public dashboards publish collateral, revenue and supply continuously.
• How fast is the exit? Lockups and cooldowns are risk. sUSDS converts back to USDS on demand, with no lock-up period and no exit fee.
• What does the issuer publish on a bad month? This is the real tell. Anybody can publish a record quarter.

What the synthetic dollar contraction actually teaches

Here is the part nobody says out loud.

Ethena built something genuinely new, and it is now diversifying its reserves into institutional lending, real-world assets and tokenized equities.

That is not a retreat. That is the whole category converging on the same answer.

Diversified backing. Verifiable collateral. A rate that survives a quiet market.

The wider tape agrees. Total stablecoin supply sat near $302.8 billion in September 2026, up about 14% year over year, and it held close to record highs straight through a downturn that took 20% off total crypto market cap in Q1 alone.

Dollars onchain are not going anywhere. The only real question is which ones pay you, and out of what.

Regulators are tightening around who may pay yield, too. The GENIUS Act bars payment stablecoin issuers from paying holders directly, and in September the European System of Central Banks proposed tighter limits on stablecoin rewards. Structure is quietly becoming the differentiator, not marketing.

USDe’s chart did not prove synthetic dollars were a bad idea. It proved that where a yield comes from matters more than what the yield says today.

The subsidy era of onchain dollars ends tonight. What comes next gets funded by revenue, or it does not get funded at all.

Your turn. Would you rather hold a stablecoin paying 8% sourced from one strategy, or 4% sourced from twenty? Drop your answer below and tell me what would change your mind.

From $15 Billion to $6 Billion: What the Synthetic Dollar Retreat Teaches was originally published in Coinmonks on Medium, where people are continuing the conversation by highlighting and responding to this story.

Fees are not revenue. Revenue is not surplus. The onchain revenue rankings look very different once you check the second number.

Over a trailing year, Uniswap collected roughly $892 million in fees. It kept about $12 million.

Hyperliquid collected $1.05 billion. It kept $943 million.

Same industry. Same twelve months. Two completely different businesses.

That gap is the whole story of onchain revenue in 2026. It is also the reason most “highest revenue DeFi protocols” lists fall apart the moment you look past the headline number.

So let’s rank them properly. Then let’s look at the one protocol that publishes a quarterly income statement like a real company, and argue about whether it is paying out too little.

Which Are the Highest Revenue DeFi Protocols Right Now?

Short answer first, because you came here for a ranking.

By protocol revenue actually retained over a trailing year, per DefiLlama:

• Hyperliquid — around $943M
• Sky Protocol — around $209M
• Jupiter — around $134M
• Aave — around $127M
• Uniswap — around $12M

Two caveats before anyone starts typing in the comments.

• Third-party methodology is not the same as a protocol’s own accounting. Sky Protocol’s own monthly settlement cycles put its annualised gross run rate at $429.4M entering the second half of 2026.
• Several very large protocols read as near-zero revenue because their fees flow to suppliers rather than the treasury. Morpho and EigenCloud both look like that. Huge, but structurally different animals.

For scale, DefiLlama tracked $93.9 billion locked across all of DeFi on 21 September 2026.

The entire industry’s retained revenue still fits inside what a single mid-sized bank books in a quarter. Worth remembering before anyone declares victory.

Fees or Protocol Revenue: Which Number Actually Tells You Who Makes Money?

Three words get used interchangeably and they mean three different things.

• Fees are what users pay. That is throughput, not income.
• Protocol revenue is what is left after liquidity providers, validators and incentives are paid.
• Surplus is what survives after the protocol covers its own operating costs.

Most rankings stop at the first one. That is how a protocol gets called a cash machine while it hands 98% of its take straight back out the door.

The number worth tracking is the revenue capture ratio: revenue divided by fees.

• Hyperliquid: around 90%
• Aerodrome: around 77%
• Sky Protocol: around 53%
• Jupiter: around 30%
• Aave: around 13%
• Uniswap: around 1.4%

A low capture ratio can be a deliberate growth choice. It can also be a business that never learned how to keep anything.

Where Does Sky Protocol’s Revenue Actually Come From?

This is where the model gets genuinely unusual, because it does not look like a trading venue at all.

Sky Protocol issues USDS and makes it available as wholesale liquidity. Independent capital allocators, together called the Sky Agent Network, borrow that liquidity at a governance-set Base Rate and put it to work.

• Spark runs lending markets and is the largest by activity
• Grove handles institutional tokenized credit
• Osero operates a neobank
• Obex incubates new allocators

Each agent keeps the spread it earns above the Base Rate. That spread is their business.

Their Base Rate payments then settle back to the protocol every month, alongside vault stability fees, real-world asset yield and Peg Stability Module fees.

So the revenue line is not one product with one demand curve. It is a portfolio of independent allocators competing for the same liquidity.

For scale: Total Protocol Collateral stood at $14.15B, against $11.48B in stablecoin supply, when I checked the dashboard on skyeco.com this week.

How Much Has Sky Protocol Made in 2026 So Far?

Published quarter by quarter by Sky Frontier Foundation, which is unusual in itself.

• Q1 2026 — $123.79M Gross Protocol Revenue and $46.04M Net Protocol Surplus. The strongest quarter on record. DL News covered it here.
• Q2 2026 — $107.35M Gross Protocol Revenue, up 10.5% year on year. Net Protocol Revenue of $40.09M, up 25.1%. Net revenue margin widened from 33.0% to 37.3%.
• Five consecutive quarters in Net Protocol Surplus.
• Full-year 2025 landed at $338M. The published 2026 outlook is $611.5M.

One line deserves a pause. Operating expenses in June 2026 came in at $161K, against $9.89M in June 2025. That is not a rounding difference.

It is an organisational consolidation that turned a modest revenue increase into a much larger surplus increase.

Revenue growth gets the headline. Cost discipline did the heavy lifting.

Who Actually Gets Paid When a DeFi Protocol Makes Money?

Revenue is a vanity metric if you cannot trace where it lands. Here you can.

The Sky Savings Rate gets funded first. sUSDS holders are paid before anyone else in the stack.

What remains runs through the Stage 2 framework:

• 50% to the Surplus Buffer
• 22.5% to SKY buybacks
• 22.5% to USDS staking rewards
• 5% to SKY buy and burn

Sky Reserves closed Q2 at $82.4M against a $150M Solvency Reserve target. Roughly 55% of the way there. Until that floor is met, reserve building outranks distributions.

Sky Governance made that call in March 2026, after S&P Global assigned Sky Protocol a B- issuer credit rating in August 2025, the first credit rating a major agency has given a DeFi protocol.

Why Is the Sky Savings Rate Only 3.6% When Revenue Is at a Record?

This is the question I see most often, and it is a fair one.

The Sky Savings Rate is a dial, not an output. Sky Governance sets it by onchain vote. It moved to 3.6% on 3 September 2026.

• Turn the dial up and current sUSDS holders get paid more today
• Turn it down and more surplus is retained, reserves fill faster, the protocol gets harder to break

Right now the dial is pointed at resilience rather than payout. You can argue that is too conservative, and plenty of people do.

What you cannot argue is that the tradeoff is hidden. Every change is proposed and voted onchain before it takes effect.

There is one wrinkle most coverage skipped. In July 2026, Sky Governance cut the Sky Spread from 0.1% to zero through the weekly Atlas Edit cycle, ratified onchain on 23 July.

Tiny number, real signal. More of the Base Rate now reaches savers instead of being retained as protocol margin.

Demand has not exactly collapsed in response. sUSDS supply reached $5.52B at the end of Q2 2026, up 149% year on year.

A published, governance-set rate is quietly one of the most useful things in onchain finance. It gives a treasury an actual discount rate for operating cash.

What Does Galaxy’s $100M sUSDS Position Tell Us About Onchain Revenue?

On 23 September 2026, Galaxy Digital (Nasdaq: GLXY) added $100 million of sUSDS to its corporate treasury and approved sUSDS as eligible collateral across its institutional trading business. Full announcement here.

The detail underneath the headline is the interesting part:

• Galaxy’s institutional loan book averages $1.4 billion
• It serves more than 1,600 trading counterparties
• Clients who post sUSDS against a loan keep accruing the Sky Savings Rate on the full position for as long as the loan runs
• Grove already provides Galaxy a $500 million warehouse lending facility

That is collateral doing two jobs at once, which is the thing traditional cash management cannot do.

Zoom out and the trend line is the same everywhere. Onchain real-world assets, excluding stablecoins, passed $33 billion in July 2026, roughly four times their level in early 2025, per rwa.xyz. Revenue rankings follow that kind of capital flow, with a lag.

How Do You Verify Any of These Revenue Numbers Yourself?

Please do not take a ranking on trust. Mine included.

• info.skyeco.com — live TVL, collateral composition, and the current state of USDS and sUSDS
• financial.skyeco.com — the balance sheet, Gross and Net Protocol Revenue, Protocol Surplus and Sky Reserves
• insights.skyeco.com — the quarterly reports and monthly financial updates published by Sky Frontier Foundation
• DefiLlama’s Sky page — an independent cross-check with a different methodology

If two sources disagree, that is usually a methodology difference rather than a scandal. Read both.

So Who Actually Makes Money Onchain?

If “makes money” means fee throughput, Hyperliquid wins and it is not close.

If it means retained protocol revenue with a published income statement, audited security reviews and a credit rating behind it, the field narrows very fast.

And here is the thing I would actually like to argue about in the comments.

Should a protocol build reserves to a hard floor before paying holders in full? Or pay holders now and raise reserves later?

Sky Governance chose the first, and capped distributions until reserves hit $150M. Ethena, Hyperliquid and Aave have each answered it differently.

Which side are you on? And which protocol do you think will top the onchain revenue rankings twelve months from now?

Drop it below. I read every response.

Highest Revenue DeFi Protocols in 2026: Who Actually Makes Money Onchain? was originally published in Coinmonks on Medium, where people are continuing the conversation by highlighting and responding to this story.