Blast and Abstract announced wind-downs four days apart after deciding their chains could not pay their way. Six months after this site named five struggling chains, Somnia is processing 1.95 million transactions a day for 357 active addresses and $358 in fees.
Two Ethereum layer 2 networks announced within four days of each other that they are switching themselves off. Blast said on October 2 that the cost of running its chain had overtaken what the chain earned, and on October 6 Igloo said it would close Abstract, the network it built for the Pudgy Penguins audience, on December 15.
Neither decision followed a hack or a regulator. Both came down to the same arithmetic: a chain has to earn more from its users than it costs to run, or someone has to keep paying the difference. Blast earned $1,793 across the whole of September, CoinDesk reported, and DefiLlama put its fees at $124 a day when it announced. Igloo's chief executive, Luca Netz, put the cost of Abstract at "tens of millions of dollars" over about two years and said the company "could no longer justify taking from the Pudgy Penguins business."
In April this site named five chains whose funding had far outrun their use: Somnia, Blast, Berachain, Sei and Monad. Six months later one of them is closing. This piece returns to the same five, using the same data provider, so the list was not picked with hindsight. Current figures are DefiLlama readings taken on October 7, 2026. April figures are the ones this site published on April 7.
Somnia: plenty of transactions, very few people
Somnia shows the widest gap between what was promised and what is being used. The chain was "built by Improbable under contract for the Somnia Foundation," according to Improbable, the London company founded and run by Herman Narula. When the mainnet launched in September 2025, Improbable described "over $270 million in committed investment from Improbable, MSquared, and the Foundation." That is a commitment, not a sum known to have been spent, and an earlier announcement in February 2025 put it at "up to" $270 million from Improbable and MSquared alone. Neither announcement says what form the capital takes.
The pitch was capacity. Improbable advertised "1M+ TPS, with sub-second end-to-end latency" and said the testnet had processed 80 million transactions in a single day. Narula, Improbable's founder and chief executive, called the launch "an inflection point for real-world use cases of blockchains."
Improbable has been here before. The London company became one of British tech's best-funded names after SoftBank led a $502 million round in 2017 that reportedly valued it at more than $1 billion, raised to build SpatialOS, a platform for third parties to build "vast virtual and simulated worlds." It has since sold its games services business, backed MSquared and its open standards for avatars and assets in the metaverse, and arrived at Somnia. Improbable did report its first annual profit, £11 million on £66 million of revenue for 2023, so the company is not the failure here. Its bet on virtual worlds is the part that keeps not paying off. Across three versions of the idea, a games platform, then the metaverse, then a blockchain, the pitch has barely changed: build the infrastructure for millions of people to share a virtual space, and the people will come. On Somnia's numbers, 357 of them have.
On October 7, DefiLlama showed Somnia processing 1.95 million transactions in 24 hours from 357 active addresses, or about 5,460 transactions for every active address. The chain collected $358 in fees that day, of which $179 counted as revenue, and held $2.93 million in value locked. DEX volume was $203,445, down 66% on the week. Fifteen new addresses appeared.
Somnia sold itself on how many transactions it could carry, and by that measure it is busy. By the measures that need a person at the other end, it is nearly empty. Its 357 active addresses are the fewest of any live chain in this group: Sei had more than four times as many, Berachain thirteen times and Monad thirty-six times. A transaction count that large from that few addresses says very little about how many people use the chain, and nothing read for this article identifies what is generating it.
The numbers have improved since April, and that should be said plainly. Fees were $12 a day then, value locked was $1.37 million and DEX volume was $14,372, and all three have risen. The base was close to zero, though. Annualizing a single day's reading, $358 a day comes to about $131,000 a year, set against a capital commitment of more than $270 million. SOMI traded at $0.20, for a market value of $31.59 million.
The rest of April's list
Berachain raised $142 million at a $1.5 billion valuation, as this site reported in April. Its value locked has fallen from $91 million then to $37.04 million, and it collected $36 in fees on October 7, all of it counted as revenue. It has far more users than Somnia, with 4,738 active addresses and 63,798 transactions, and $47.42 million of stablecoins on the chain, although that balance fell 29% over the week.
Sei, which raised $85 million, held $34.68 million in value locked against $40.5 million in April and took $44 in fees. It is not quiet in the way Somnia is. DEX volume was $6.79 million and stablecoins on the chain were worth $306.75 million. What it does not do is turn that activity into fee revenue.
Monad is the exception. In April this site called it too early to judge. Since then its value locked has grown from $362 million to $1.024 billion and its daily fees from $6,311 to $16,633, from 12,818 active addresses. That is still modest beside the $3 billion valuation at which it raised $248 million, and its DEX volume fell 69% on the week to $32.95 million. It is nonetheless the one chain on the list that has grown toward its funding rather than away from it.
What the fee lines show, and what they cannot
On fee income alone, Berachain and Sei each took in less on October 7 than Blast was taking a day when it said the economics no longer made sense: $36 and $44, against Blast's $124. Somnia's $358 is higher than all three.
That comparison has limits that matter. Blast is a layer 2 that pays to publish its data to Ethereum. Berachain and Sei are layer 1 networks that reward their validators largely with newly issued tokens rather than out of fees, so a thin fee line costs them in a different way. None of these chains publishes what it costs to run, which is why Blast's statement that "the ongoing costs of maintaining Blast exceed the revenue generated by the L2" is the only direct account of a cost line here. Revenue can be compared from outside. Whether a chain can live on it is something only its operator knows.
Thin activity is not a shutdown notice either. Somnia, Berachain and Sei have announced no plan to stop, and chains have run for years on less. What Blast and Abstract show is what happens when whoever covers the shortfall stops: in Abstract's case a consumer brand deciding to stop subsidizing a chain, and in Blast's a project concluding that it saw no "credible path to making the chain economically sustainable."
Netz said Igloo had considered launching a token to keep Abstract going and decided against it. "A token only works if there is something driving demand to it," he wrote. Every chain on April's list has one.
