What If Your Treasury’s Idle Cash Could Work Harder?
There was Rp 50 billion sitting in a bank account.
Earning 4% per annum, gross. After the 20% withholding tax on interest income, that’s 3.2% net. Safe, predictable, boring. The kind of return that makes a CFO sleep well and a treasury person quietly restless.
I remember doing the math in my head. Rp 50 billion at 3.2% is Rp 1.6 billion a year. Not nothing. But I had been using DeFi protocols in my own time long enough to know that stablecoin yields on platforms like Aave were sitting at 7, 8, sometimes 10 percent. The spread was real. The opportunity felt obvious.
So I walked into my CFO’s office with the idea.
He thought I was joking.
I did not have numbers. I did not have a framework. I had intuition, some personal experience farming yield on-chain, and exactly zero preparation. The conversation lasted about three minutes. The idea was shot down before it had a chance to breathe, and honestly — he was right to dismiss it. You cannot walk into a CFO conversation about capital allocation with vibes.
I did not get another chance. The timing was what it was.
But the question stayed with me: what would it have taken to be taken seriously?
The Problem Nobody Talks About
Most crypto exchanges are built to move money, not grow it. The business model is straightforward — facilitate trades, earn fees, repeat. Treasury is a support function, not a profit center. Idle capital gets parked in a bank, earns a modest rate, and nobody thinks too hard about it.
That is understandable. Running an exchange is operationally complex. The team is small. Risk appetite is calibrated toward the core business, not secondary yield strategies.
But here is the thing: the idle capital problem is real and it is significant. A newly licensed Indonesian exchange under POJK No. 27 of 2024 needs minimum paid-in capital of Rp 100 billion. After maintaining the regulatory equity floor of Rp 50 billion, there is Rp 50 billion that needs to sit somewhere. At 3.2% net, that is Rp 1.6 billion a year in interest income.
What if it could be Rp 2.5 billion? Or Rp 3 billion? Without taking on excessive risk. Without violating any regulation. Without adding operational complexity the team cannot handle.
That is not a trivial question. For a young exchange trying to reach profitability, the difference between 3.2% and 6% on Rp 50 billion is the difference between a cost center and a meaningful secondary revenue stream.
The Complication
The obvious answer — just put it in DeFi — immediately runs into three walls.
The first wall is the CFO. To get approval for anything beyond a bank deposit, you need to show your work. Expected return, downside scenario, what happens if a protocol gets hacked, how quickly can you get the money back. These are not unreasonable questions. They are the right questions. And “trust me, I farm yield on weekends” is not an answer.
The second wall is the regulatory environment. In Indonesia, stablecoins are classified as crypto assets under OJK, not fiat. Deploying USDC or USDT into Aave is crypto asset activity. That shapes the conversation with management and requires the exchange to hold the appropriate license tier. You cannot pretend the regulatory dimension does not exist.
The third wall is the currency mismatch. The exchange operates in IDR. DeFi yields are denominated in USD. Every percentage point you earn on-chain can be partially or fully erased by unfavorable IDR/USD movement. A framework that ignores FX is not a framework — it is wishful thinking.
None of these walls are insurmountable. But you have to actually deal with them, not wave them away.
Building the Answer I Did Not Have
So I built the thing I should have walked in with the first time.
Not a pitch deck. Not a spreadsheet with some APY numbers copy-pasted from DeFiLlama. A proper quantitative framework — the kind that a CFO, a risk manager, or a regulator could actually interrogate.
The starting point was the instrument universe. Two traditional IDR instruments: RDPU (Indonesian money market funds, daily liquidity) and SBN short-term government bonds. Three on-chain USD protocols: Aave for stablecoin lending, and Pendle for fixed-rate and leveraged yield. All yields converted to annualized IDR terms before any comparison. The FX dimension handled explicitly, not ignored.
Then two risk layers that standard portfolio models do not include.
The first is smart contract exploit risk. On-chain protocols can be hacked. Funds can be partially or fully lost. This does not show up in historical yield variance — exploits are rare and binary. The model treats it as a separate expected loss haircut: probability of exploit multiplied by severity, derived from DeFiLlama’s hacks database. Every on-chain instrument gets this applied before it enters the optimizer. Off-chain instruments get zero — because SBN does not have a smart contract.
The second is liquidity heterogeneity. A money market fund redeems at NAV every business day. A Pendle YT position requires selling on an AMM where a large exit will move the price against you. These are not the same thing, and a framework that treats them as equivalent is not modeling reality. The model converts this into a basis point penalty derived from observable data — utilization rate history for Aave, pool depth relative to exit size for Pendle.
Then an efficient frontier. One hundred portfolios from minimum risk to maximum yield. CVaR as the optimization objective — not variance, because variance treats upside and downside symmetrically, which is not how a treasury thinks about risk. CVaR tells you what the average loss looks like in the worst 5% of scenarios. That is the number that matters when you are explaining downside to a CFO.
What the Market Said Back
Here is where the story gets interesting.
The framework was designed for an environment where DeFi yields are elevated — where Pendle offers 14 to 20 percent and creates a meaningful spread above Indonesian anchor instruments. When I ran it against live market data, the market said: not right now.
Pendle YT came back at negative 32 percent annualized. Not a data error. YT holders earn the floating yield above the fixed rate implied at purchase. When actual protocol yields fall below that implied rate — which has been happening as DeFi yield compresses — YT holders lose money. The optimizer correctly excluded it.
Pendle PT came back at about 5 percent implied APY. After haircuts, that is essentially the same as a money market fund. The optimizer correctly ignored it.
The only meaningful on-chain yield enhancer in the current environment is Aave, at roughly 7.93 percent IDR after haircuts. The efficient frontier spans 4.98 to 6.70 percent.
That is a modest range. It is also an honest one.
I could have adjusted the assumptions to produce a more dramatic result. I did not. A treasury framework that manufactures compelling outputs is worse than no framework at all — it gives false confidence to exactly the decisions that matter most.
What the model produced instead is something more valuable: a clear picture of what the market currently offers, why each instrument is priced the way it is, and what the framework will do differently when yield conditions change. The architecture is ready for when Pendle reactivates. Only the market needs to move, not the model.
The Conversation I Would Have Now
If I walked back into that CFO’s office today, the conversation would be different.
I would show the regulatory framing first. Own capital only, above the equity floor — not customer funds, which are untouchable by law. Stablecoin deployment treated as crypto asset activity with appropriate license assumptions stated explicitly. OJK framework acknowledged, not avoided.
Then the instrument selection rationale. Why Aave — because it is the deepest, most battle-tested stablecoin lending protocol and the natural starting point for any on-chain treasury allocation. Why Pendle — because it represents a structurally different yield category from lending utilization, which is what genuine diversification looks like.
Then the risk layers. Not “here is the APY,” but “here is the expected return after we account for the probability the protocol gets exploited and the cost of exiting a position under stress.”
Then the honest result. In the current environment, on-chain yield premium over Indonesian government instruments is narrow. The framework recommends a modest Aave allocation in the Balanced and Max Yield portfolios, anchored by a heavy SBN and MMF position. Nothing aggressive. Nothing that keeps a CFO up at night.
And critically — here is what the model does when conditions change. When DeFi yields recover, the optimizer will naturally shift allocation toward Pendle. The constraints are already in place. The risk layers are already calibrated. The decision framework does not need to be rebuilt from scratch every time the market moves.
That is the conversation that does not get dismissed in three minutes.
Why This Matters Beyond One Exchange
The idle capital problem is not unique to Indonesian crypto exchanges. Any institution sitting at the intersection of traditional finance and on-chain infrastructure faces the same question: how do you govern the allocation of capital across two worlds with fundamentally different risk profiles, yield sources, and liquidity characteristics?
The tools exist. CVaR optimization, regime-aware covariance, explicit risk haircuts — none of this is new. What is less common is applying these tools to a genuinely hybrid instrument universe, localizing the framework to a specific regulatory environment, and being honest when the data does not support the narrative you hoped for.
That combination — rigor, local grounding, intellectual honesty — is what separates a treasury framework from a yield farming spreadsheet.
The Rp 50 billion is still sitting somewhere, earning 3.2 percent.
The answer to the question is built. The next step is getting it in front of the people who can use it.
The full technical documentation and source code for this project are available at github.com/alfajr666/yield-generation-model-sample. For the methodology deep-dive, see the project entry.
If you are building treasury infrastructure at a crypto-native firm and this resonates, I would welcome a conversation. gilang.f@delomite.com