Terra: Staking Rewards Special Comment
The initial autopsy suggests an avoidable disaster.
We have many, many thoughts on Terra that we will share in greater depth in our regular newsletter next week.
Tl;dr: On an industry level, UST grew too big, too fast. Terra’s indifference to the growing systemic risk of Anchor, combined with the off-chain liquidity and 4pool migration the system presented a weak point which cost of thousands retail investors their life’s savings. This has been a disaster for our industry that will set it back years.
Among the few positives:
Proof of Stake networks are resilient and functional even amid the worst scenario imaginable. Proof of Stake is here to stay.
The Tendermint consensus worked amazingly well, all considered, under the worst stress test imaginable.
But it was also, in our view, an avoidable disaster. A disaster we aren’t doomed to repeat, so long as we rigorously study, learn, and self-critique, down to the finest details of this debacle.
Let’s reflect, rebuild, and re-emerge stronger.
Our more detailed takeaways, based on what we know today:
Anchor, in its initial conception, was no “Ponzi.” It was the equivalent of Jeff Bezos selling a convertible bond in AMZN in February 2000, a low-probability future dream at a very high cost (assuming the dream worked out). It was a way for Terra to simultaneously rent capital at scale, onboard new users into its protocol, and help Terra scale its broader network. Bets on growth that don’t pan out are always termed “Ponzis” by journalists and other professional critics who never actually built anything.
By December 2021, with the launch of the Degenbox, Anchor had strayed very far from its original mission (as a clever user marketing hook to onboard new DeFi users), when Daniele Sestagalli took advantage of Anchor’s ambitious aUST composability to “loop” a couple hundred million dollars of deposits into $1.2B of UST at a 100%+ annual yield. This was ridiculously parasitic. Kwon’s implicit support of this was a massive red flag that showed no regard whatsoever for sustainable growth.
The crypto-crash of January/February 2022 killed off the Degenbox, reducing stress on Anchor and offering Terra a golden opportunity to cut deposit rates to more sustainable levels and punish looping. Kwon, however, wanted stablecoin market dominance at any cost, and saw Anchor’s unsustainably high rates as a worthwhile price to get there.
During 1Q22, Anchor turned into an obvious systemic risk for the Terra protocol. Bigger and bigger sums of UST were parked there, doing nothing else within the broader ecosystem. That wasn’t “user onboarding for real-world use cases.” It was parasitism that would run for the exits at the first sign of systemic stress. It was pointed out, debated, and discussed ad nauseam. Everyone agreed that it was going to be a problem sooner or very sooner.
Less discussed, and even less defensible, was how Anchor’s high deposit rates also sucked capital out from more productive projects on Terra and other integrated protocols, like Avalanche. Who needs to take big risks funding binary-but-maybe-productive assets when you can just get paid 23% per year for doing nothing? This was a major complaint by other Terra protocols, the Avalanche team, and many other parties who only wished Terra the best. Kwon never cared.
UST deposits put to real world use (eg, lending to a small business for 3 months, or being locked up in a smart contract as duration collateral against some other digital asset) can’t just run for the exits when something goes wrong. Unproductive UST deposits on Anchor, by contrast, had every flexibility to run for the exits — and run they did.
By March, governance proposals on Anchor emerged to start to cut deposit rates. All of the proposals were way too conservative. I was told on good authority that the Anchor team wanted significantly deeper cuts than what was approved: a 150bps per month cut to a 15% deposit rate floor. Kwon single-handedly vetoed them.
Many, probably a large majority, of “independent protocols on Terra” employees were always either direct employees of Terraform Labs, or dependent on being in good standing with TFL to survive on Terra. The centralization of the network was far more acute than widely understood. Everyone ultimately depended on Kwon.
To finance the unproductive Anchor deposits, Terra had to dump larger amounts of LUNA onto the market. Terra was, in effect, shorting its own governance token to subsidize the growing systemic risk of idle Anchor deposits.
Terra simultaneously decided that it should move from a zero-collateralization model to a diversified-collateralization model. This required more LUNA to be sold, to then be exchanged for the preferred diversified collateral (BTC). This was a very debatable move, as it radically centralized the network’s solvency (management of the reserve) at the expense of LUNA holders. However, it made sense to most people up to a marginal point.
To create corresponding "LUNA burn” to offset this second wave of LUNA issuance, Do Kwon made his most critical mistake: listing UST on centralized, off-chain venues (FTX, Binance, etc). This created huge pools of “FTX-wrapped UST” and “Binance-wrapped UST”, trading at its own price, not subject to Terra’s exit tax failsafe mechanism.
What was Terra’s exit tax failsafe mechanism?
Terra foresaw the risk of capital flight events, and thus implemented a soft capital control system, whereby any capital fleeing the system exceeding around $100M per day would face a rapidly increasing “swap fee” (exit tax).
This tax revenue would accrue to LUNA stakers, the critical loyalists of the network, and would be denominated in UST, which under the protocol’s design would always be much more stable than the price of LUNA — thus giving LUNA stakers a rapidly increasing source of stable revenue at a time of rapidly increasing systemic stress. Again, risk symmetry creatively solved a previously intractable problem, with a robustness that dealt with very significant edge-cases.
It’s thus no coincidence that the crash in UST started on off-chain venues.
Terra had, by this point, raised around $3.5B of non-LUNA reserves. Seemingly $1.5B of it was squandered defending Terra’s off-chain peg when the bank run started on Sunday night.
Instead of working as an “on-chain UST redeemer of last resort” (ie, buying back UST at X% discount to par, after exiters were already being hit heavily with exit taxes), half of Terra’s reserve was blown as the buyer of first resort covering up an off-chain attack surface which Terraform Labs’s growth-at-all-costs decisions had created.
After the initial $1.5B wasn’t enough to save the off-chain pegs, the idle Anchor deposits exited en masse. This happened on Monday and Tuesday, crippling the on-chain peg to .73 and the off-chain pegs to lows of .60 or so.
By now, LUNA was trading in the $20s, but it wasn’t dead yet. The peg gradually healed to around $.90 as markets awaited the “plan” of TFL and its loyal institutionalists (Jump, Delphi, etc).
Kwon’s long-awaited plan turned out to be yet another off-chain bailout at punitive terms, dumping $1.5B of LUNA at a 50% discount to current prices and a one-year lockup. It was leaked by multiple parties and never denied by TFL, so the market justifiably interpreted it as fact.
This proposal was a disaster: it didn’t address any sources of systemic risk (Anchor etc), it guaranteed that all retail LUNA buyers were being treated as suckers, and it broke the back of incremental LUNA buying at the exact moment when the protocol needed more funds to recapitalize.
Therefore, UST’s peg broke and LUNA’s issuance exploded at the same time. LUNA went from $20 to $4 very quickly, while the rate of UST “bad debt” retirement slowed significantly, because LUNA’s price was falling much faster at this point than UST was being burnt. This was when LUNA/UST was effectively doomed, because the price of LUNA was falling / supply of LUNA was growing much faster than the protocol could “burn” UST.
Buyers had no reason to rescue the protocol while such a dilutive plan was being discussed.
UST/LUNA is currently bankrupt because its liabilities ($11.5b UST outstanding) vastly exceeds its assets (unknown, presumed to be $2-3b of BTC). No amount of LUNA buying will change that fact.
The only path to protocol solvency lies in the protocol buying back UST at the rate at which assets equal liabilities, which in this case would be $.10-.20 of UST.
Retail investors should not under any circumstances buy any more LUNA, because:
Terra has “temporarily, for a few days” (indefinitely) suspended validator delegation. This means that TFL will probably use the excuse of an institutional attack (never remotely proven) to do a staker snapshot pre-attack / LUNA token burn to unwind an arbitrary amount of blockchain time. LUNA bought after whenever that snapshot takes place will be worthless.
The above decision will enrage Terra decentralization loyalists, destroy Terra’s legitimacy, and destroy long-term faith in the UST/LUNA burn mechanism. It represents massive systemic risk. It will preserve TFL’s control over the protocol but destroy pretty much everything else, in my opinion. Do Kwon would be a king of ashes, perpetuating a hoax to justify remaining in control.
Terra may be weighing a much less stupid alternative:
Terra said it wanted to be much closer to the FRAX model going forward, i.e., 50-75% collateralized in non-LUNA tokens. For $4 billion of UST in remaining circulation, Terra would need $2-3B of assets. So Terra would have to somehow retire $8.5 billion or so of UST (there are still roughly $11.5B outstanding, per Smartstake) with $1B of BTC, and then get $1B USD of external bailout at 1:1 UST, to have post-bailout assets of $2B of non-LUNA assets, $1B of LUNA assets, and $3B-ish of liabilities (67% non-LUNA collateralization).
This latter path is the only sustainable path forward. It will also require TFL to give up control of the project.
Kwon’s prioritization of unsustainable growth over real-world use beyond any rational point (by keeping Anchor rates as high for as long as he did) and prioritization of marginal UST growth over on-chain failsafes (by listing UST on off-chain venues) were fatal, protocol-destroying decisions for which TFL must pay, and which the industry must NEVER forget.
Cults of personality ALWAYS become central points of failure in the end.
DAOs and protocols which emphasize toxic maximalism (gloating, punching down on skeptical outsiders, generally redefining-upwards the socially acceptable level of arrogance) in the face of objective criticism should not be trusted. Terra’s structure always had bigger tail risks than most, which they knew better than everyone else. If any protocol was obliged to respect criticism, not punch down on skeptical outsiders, and generally behave in a humble manner, it was Terra.