Solana's rent cut is 90% off for new accounts and a 10x premium frozen into the old ones

Step three does most of the work. The rate walks 6,960 to 6,333 to 5,080 to 2,575 to 1,322 to 696 lamports per byte across five separately gated activations ^[4], and the 5,080-to-2,575 gate alone removes 2,505 lamports per byte, roughly 40 percent of the entire reduction ^[20]. By the time the fourth gate lands, the rate is already down to 19 percent of where it started, leaving the final flag to deliver the last fifth ^[21].

The asymmetry is in two sentences of the proposal. The change only relaxes the constraint: accounts are permitted to hold less, never pushed down to the new minimum ^[8]. And CloseAccount refunds whatever lamports the account actually holds, not what the current rent formula would ask for ^[9]. So an account funded before activation refunds about 0.002 SOL when closed, and its post-activation equivalent refunds about 0.0002 SOL ^[10], a ratio of exactly ten to one because that is the ratio of the two per-byte rates ^[22].

Run that through the numbers in the writeup. One memecoin season commonly leaves 50 to 200 dead token accounts behind ^[12], which is where the reported 0.1 to 0.4 SOL of dead ATA balance in an active wallet comes from ^[13]. The same wallet, assembled after full activation, would be sitting on 0.01 to 0.04 SOL instead. The gap, 0.09 to 0.36 SOL per wallet, is over-funding that no future wallet will ever accumulate ^[23]. Developers carry the larger version of this: an interrupted deploy strands a buffer account holding rent for the whole binary, often 1 to 5 SOL per failed attempt ^[15], all of it priced at the old rate.

Nothing on the network closes these for you ^[11]. Tens of millions of empty accounts have already been swept by people who noticed, and far more are still open ^[14], and the sweep needs no custodian, only a signed instruction with the refund pointed where you like ^[17].

The second-order effect is a re-ranking of what a mint costs. Metaplex charges a fixed 0.0015 SOL, which becomes the largest line item on a Core NFT mint once the rent underneath it collapses ^[19]. Against the old ATA deposit that fee was 0.75 times the rent; against the new one it is 7.5 times ^[24]. Compression takes the same hit in reverse: the rent advantage of compressed accounts falls from roughly 400x to roughly 40x ^[18], which is the identical factor of ten ^[25].

One caveat on provenance. This all comes from a single writeup published on dev.to and originally on xroot.dev, which says it checked the proposal text rather than the coverage ^[26]; the late-August-2026 start for Agave 4.2 mainnet feature activations is that author's timing ^[5].