Thursday Links: sanctions MEV, “stretch,” compute perps, and prediction market sharks

The cost of evading sanction onchain

An academic paper introduces the idea of “sanction-evasion MEV”: the money earned by validators and block builders when a sanctioned entity races to move their stablecoins to safety before an issuer like Circle or Tether can freeze them.

To freeze funds, stablecoin issuers have to get a transaction confirmed on-chain, just like anyone else.

“Whenever a freeze transaction and an evading transfer enter the same ordering pipeline,” the authors explain, “enforcement becomes a priority contest rather than a purely issuer-controlled action.”  

It’s a contest in which the sanction evaders have a distinct advantage.

In one instance cited in the paper, an issuer submitted a freeze transaction 22 minutes before a targeted entity could respond. But the funds — 100,000 USDT — made it to safety because the holder paid a sky-high priority fee to get their transaction accepted immediately. The issuer’s freeze transaction, by contrast, spent three hours languishing in the mempool.

(In other instances, an issuer’s freeze instructions failed simply because they simply ran out of gas.)

These “ordering contests” are rare, but highly profitable for validators and block builders (the “MEV supply chain”) because sanction evaders will pay quite a lot to get illicit funds to safety.

More often, though, the funds are moved well in advance of the freeze. “Despite over $1.5B in stablecoins being frozen in aggregate,” the paper says, “a substantial fraction of sanctioned addresses hold zero or near-zero balances at the time enforcement takes effect.”

It’s a wonder that sanctioned entities use stablecoins at all, knowing they can be frozen. But the paper explains that, too: “Evaders heavily prioritize liquidity over anonymity.” 

Privacy coins like Zcash have been a rare bright spot for crypto in terms of price action lately. But the people who would most benefit from them don’t seem to use them much.

The paper finds that “compliant CEXs are the primary destinations,” for funds subject to sanction, “whereas privacy-enhancing obfuscation services handle less than 1% of the volume.”

In that sense, the $1.5 billion that’s been frozen by stablecoin issuers can be seen as the price that sanction entities are willing to pay for liquidity. 

Strategy stretches the truth

Michael Saylor markets Strategy’s STRC preferred shares as a kind of money market fund, backed by Bitcoin, with a yield of 11.5%, and effectively tax free.

Glenn Cameron says it isn’t any of those things.

In a note for the Bitcoin financial services firm Onramp, Cameron details how Strategy’s claims about “stretch” — as Saylor calls it — are themselves stretching the truth.

Saylor explicitly markets STRC as being “backed” by bitcoin, but STRC’s investors are really just creditors to Strategy. “STRC holders have no direct lien on any Bitcoin,” Cameron explains — in a bankruptcy scenario, they have only a subordinate claim on the bitcoin Strategy holds. 

Yesterday, Saylor said that “BTC capital gains fund STRC dividends.”  In truth, STRC’s dividends are funded by issuing additional shares of STRC. “New investors' capital pays existing investors' dividends,” Cameron explains.

(Generously, he refrains from using the standard term for that sort of arrangement.)

Saylor touts STRC’s “money market-like stability.” But to the extent that’s true, it’s been manufactured by raising the dividend whenever the shares go down — “a defensive adjustment to prevent capital flight,” as Cameron describes it. 

In March, Saylor told the crowd at Blockworks’ Digital Asset Summit that the 11.5% yield on STRC is effectively tax free. For a New York City taxpayer, he said, “it’s like a bank account that yields 23%.”

Cameron, however, explains that “every dollar received reduces your cost basis” — which means the tax bill comes due when you sell. And investors will eventually have to sell because STRC is perpetual preferred equity: “The only exit is the secondary price at whatever price it trades,” Cameron notes.

Even the headline 11.5% yield is misleading because it assumes dividends are always reinvested at the same rate. Without reinvesting, Cameron calculates that over eight years the yield works out to just 8.5%.

Worse yet: “If the holder eventually exits at par after eight years and pays federal long-term capital gains tax of 23.8%... the after-tax CAGR falls to approximately 6.9%.”

6.9% does not seem like a good risk/return for an “unsecured, discretionary yield from a loss-making issuer,” as Cameron describes it.

There are risks here for holders of bitcoin, too.

Cameron estimates that even in a scenario where BTC compounds at a respectable 15% per annum (much better than it’s been doing lately!) there’s still a 9.4% chance that Strategy defaults on its debt.

More concerningly, over an eight-year cycle there would also be a 45.5% probability that Strategy would make a forced sale of bitcoin at least once.

That might be overly optimistic. As of this morning, Polymarket odds suggest there’s a 71% chance that Strategy sells bitcoin this year.

The odds of Strategy selling shot higher after this morning’s post from Saylor: 

That’s quite a change from his previous advice to “never sell.”

Even so, to follow it himself, he’ll have to sell us a lot of STRC. 

Tradable AI compute

BlackRock’s Larry Fink believes the next big thing for investors will be “buying futures of compute.”

By that, he means futures contracts tied to the cost of AI training and inference (which is currently booming). 

This would be useful because the cost of compute is always changing, which makes it difficult for an AI lab like Anthropic to plan their next model.

The latest frontier models cost $100s of millions to train, so this could become a big market.

It might be a crypto market, too: MNX — a crypto-based “AI exchange — is already testing a perpetual future that tracks the cost of renting H100 GPUs (the Nvidia chips typically used to train frontier models).

That might not be enough of a head start if a giant like Blackrock chooses to do the same, I don’t know.

But Enron beat the investment banks to financializing energy, so maybe crypto can beat them to financializing compute, too.

(Without having to cook the books, hopefully.)

Prediction markets professionalized

The Wall Street Journal reports that prediction markets are increasingly attracting professional trading firms, “staffed with dozens of employees that pay millions of dollars for specialty sports and finance data and run trading algorithms” — exactly the types that quickly get banned from traditional sportsbooks 

One good thing about peer-to-peer prediction markets is that, unlike FanDuel or DraftKings, which take the other side of your bets, Kalshi and Polymarket simply facilitate them — so they’re happy for you to win.

The bad thing about prediction markets is that they attract the sharks that have been banned from FanDuel and DraftKings.

This makes winning much harder.

— Byron Gilliam