Persistence
Worlds did stay up, and what people built in them is still there. Nothing about the technology failed.
Never the problem.The sequence, dated: what was promised, which assumption failed, what it cost, and what was still standing afterwards.
As at 2026-08-28A narrative built on four assumptions met a market that only supported two of them. Persistent worlds and enforceable digital scarcity worked as advertised. Population and redeemable value did not: the audiences never arrived at the scale priced in, and value earned inside the worlds could rarely leave them. Land repriced by 72–95%, Meta cut its Reality Labs division repeatedly, and the term itself became a fundraising liability by 2023.
| When | What happened | What it meant |
|---|---|---|
| Oct 2021 | Facebook renames itself Meta. | The category acquires a name, a budget and an assumption that hardware would create the audience. |
| Nov 2021 | The record land purchase: 116 Decentraland parcels for $2.43M. | Land is repriced as property. The number is still the category's reference point. |
| Mar 2022 | Metaverse Fashion Week — 70+ brands trade inside Decentraland. | The high-water mark for tenancy, and proof the venue model worked. |
| 2022–23 | Land floors collapse. The Sandbox −95%, Decentraland −89%. | Buyers discover that scarcity without footfall produces no rent. |
| 2023 | "Metaverse" becomes a fundraising liability; companies quietly restate themselves as AI. | The narrative dies well before the underlying platforms do. |
| 2024–26 | Meta cuts Reality Labs repeatedly. Metaverse-linked tokens collapse again from early-2025 peaks. | The loudest occupants leave. Search demand does not. |
Sources: CoinGecko Research, The Block, Dezeen.
Two of the four held up completely. Naming the two that did not is the whole value of the post-mortem.
Worlds did stay up, and what people built in them is still there. Nothing about the technology failed.
Never the problem.On-chain land supply was genuinely limited and genuinely transferable. Ownership worked exactly as designed.
Also never the problem.The audiences priced into land never arrived. Open worlds counted daily users in the thousands while land was valued as though they had millions.
The binding constraint, then and now.Earnings inside a world could rarely be converted to anything outside it, so the only exit was the next buyer.
Turned a correction into a rout.If the failure had been technological, waiting would fix it. It was not — it was a failure to check whether anyone was in the room. That means the same businesses work today, at today's prices, wherever the room is genuinely full. Which is why the question worth asking is not when the metaverse returns, but which revenue models survived.