← All posts

Aave v4 is not a version of Aave v3

It is a different protocol with a different mental model. You no longer pick a lending market — you pick a spoke, and spokes cannot see each other.

Architecture4 min readDeFiLoops

The name invites you to read v4 as v3 with improvements. It is not. It is a separate protocol with a separate deployment, separate liquidity and a different mental model, and the two run alongside each other rather than one replacing the other.

If you carry a v3 habit into it, the thing that will surprise you is not a changed parameter. It is that your positions can no longer see each other.

Spokes

Where v3 gives you one pool per chain, with every asset in it counting toward one health factor, v4 divides lending into spokes. A spoke is a self-contained venue with its own list of assets.

The two we reach:

main
WETH, wstETH, WBTC, USDC, USDT. The general-purpose one
gold
Native XAUt, USDC, USDT. Tokenised gold, which no other venue in this system holds natively

An asset that is not on the spoke’s list is refused rather than quietly falling through to a default — worth knowing because in this design the zeroth entry of a list is a real entry, so “not found” and “the first one” are dangerously easy to confuse in code that assumes otherwise.

Spokes cannot see each other

This is the whole change, in one sentence: a spoke counts only the collateral supplied to itself.

Collateral in main does nothing for a borrow in gold. Two positions in two spokes have two independent health factors, and a position that is comfortable in one does not help a position that is struggling in the other.

The cost lands on you as a user: you now choose a spoke, and the choice sticks.

There is no native way to move between spokes

Nor is there here. A move carries one venue, and no step in this system relocates a position from one spoke to another.

Aave’s own position on this is public — a user asked on the governance forum in July 2026 how to migrate an existing position between venues and was told no automated tool exists yet, one is planned, and in the meantime the route is the manual one: repay, withdraw, supply again, borrow again. Which is exactly the sequence that is a deadlock without borrowed money.

A deposit is not collateral until you say so

The second habit v3 teaches you that v4 breaks. In v4, supplying an asset does not by itself make it back anything — it sits as a deposit until it is explicitly marked as collateral.

So a supply here is three actions, not one: approve the token, supply it, and then mark it as collateral. Miss the third and everything looks correct — your balance is there, the transaction succeeded — while your borrowing power has not moved at all.

Where it actually exists

Worth being exact, because the answer has changed twice this year and will change again.

Ethereum
v4 live since March 2026. This is where we reach it
Avalanche
v4 live since July 2026, the first deployment beyond Ethereum
Base
Under governance discussion as of September 2026
Arbitrum
No v4 proposal. v3 only

For our purposes the consequence is blunt: on two of our three chains, v3 is the only Aave there is. A plan that shifts a position into v4 has somewhere to go on Ethereum and nowhere to go on Base or Arbitrum, and that is a property of where Aave has deployed rather than a limit we chose.

The gold spoke is the interesting one

Native XAUt as lending collateral is not available anywhere else in this system. It makes a shape that was previously awkward straightforward: hold gold, borrow a stablecoin against it, and keep the gold exposure — without selling the metal to raise the cash.