"I borrow 1 trillion from pension funds X, do <some complex sequence of trades>, and in the end return the trillion to X, all within zero seconds" and "I perform <some complex sequence of trades>, requiring net-zero capital from my end" are equivalent statements.
Rational actors should recognize them as such, and the marginal cost of lending (interest) should converge to zero – in other words, X should concede that they do not provide anything valuable to the trade.
Practically, existing protocols might not be amenable to that insight, and the lending market is apparently too inefficient to support 0% flash loans – at least for now. So the value that X provides is protocol backwards compatibility, I suppose!
Of course there's a need for a lender, even at a microsecond transaction you do need funds to perform arbitrage, it's not just all theoretical funds - there's backing to them.
Else let's all just pretend trade on a real market until it collapses and just make sure they come a-knockin' on some non-existent firm...
Plus my conjecture, from my understanding, is the flash loan terms incentivise fast payback, whilst still retaining some fixed profit.
Akin to a credit card which has an effective 0% rate of you pay back within a certain time frame, but raises to 30% monthly after that.
I'd be happy to provide you with a flash loan at 0.001% fixed rate of profit for me/cost for you for the first 5mins and scare you with a 30% rate calculated every minute after that. For some pre-validated huge sum I know your business can be liable for, of course.
Which allows capital scarce firms to leverage these micro-loans on fast arbitrage opportunity where they should take any transaction that provides anything larger than that 0.001% in transacted profit. And let's them compete with larger firms on optimising pricing.
It’s not a microsecond, it is zero seconds. The entire series of trades either succeeds as a bundle, or does not happen at all.
This is quite different from "normal" arbitrage, which consists of a series of non-atomic trades. There’s various risks here, both on the side of the arbitrageur (offer books can change, making the trade non-profitable halfway through) and by extension for all counterparties (due to the arbitrageur not being able to settle for the promised/borrowed amount), as well as non-zero time of locking up capital. Both have a price.
> lender makes money,
Theoretically, there is no need for a lender.
"I borrow 1 trillion from pension funds X, do <some complex sequence of trades>, and in the end return the trillion to X, all within zero seconds" and "I perform <some complex sequence of trades>, requiring net-zero capital from my end" are equivalent statements.
Rational actors should recognize them as such, and the marginal cost of lending (interest) should converge to zero – in other words, X should concede that they do not provide anything valuable to the trade.
Practically, existing protocols might not be amenable to that insight, and the lending market is apparently too inefficient to support 0% flash loans – at least for now. So the value that X provides is protocol backwards compatibility, I suppose!