$5.2 billion in venture funding. Six mainnets live. Combined daily fees: $360.
I pulled those numbers from DeFiLlama at 3 a.m. while tracing the transaction logs of Berachain, Celestia, Scroll, Eclipse, Sonic, and Manta. The math doesn't lie. These projects raised enough capital to buy a small country, yet their chains produce less economic activity than a single coffee shop in Bangalore. This isn't a crash. It's a silent extinction event.
The Context: the narrative inflation cycle
Between 2021 and 2024, crypto was drowning in liquidity. VCs were desperate to deploy capital into any project that promised “the next L1” or “the ultimate scaling solution.” The story was always the same: a new consensus mechanism, a modular architecture, a zero-knowledge circuit design that would revolutionise the industry. Teams hired from Stanford and MIT, wrote elegant whitepapers, and raised nine-figure rounds. Berachain alone secured $755 million. Celestia raised $155 million. Scroll, $190 million. Eclipse, $117 million. Sonic, $170 million. Manta, $130 million. The total exceeds $5 billion.
But by July 2026, most of these chains are ghost towns. The daily transaction fees on Scroll are $24. On Eclipse, the TVL is $1.15 million – barely enough to cover a single DevOps salary. Berachain's BERA token has dropped 98% from its peak. Celestia's TIA is down 98% as well. Manta's TVL fell from $650 million to $4 million after its airdrop. The narrative that drove these valuations has evaporated, replaced by AI and real-world asset tokenisation hype.
The Core: a forensic ledger of failure
I spent last week decompiling the on-chain activity of each project. Here is what the data reveals.
Berachain – The Liquidity Proof That Proved Nothing
Launched in early 2025, Berachain promised a novel Proof-of-Liquidity consensus where validators stake liquidity tokens instead of native coin. The idea was elegant: align validator incentives with DeFi health. The reality? In May 2025, the chain suffered a vulnerability inherited from a Balancer hack on Ethereum. Validators had to pause the network. After that, user interest collapsed. The annual report from Berachain's foundation admitted “a decrease in narrative heat and a shrinking total addressable market.” Today, the chain processes maybe a few thousand transactions per day. The BERA token trades at a fraction of its launch price. Brevan Howard Digital, a lead investor, negotiated a clause that gave them a one-year risk-free refund right – essentially a put option that protects VCs while retail holds the bag.
Celestia – The Data Availability Layer That Nobody Needs
Celestia was the modular blockchain darling. The thesis: separate consensus from data availability, let rollups post data to Celestia for cheap. It worked technically – the mainnet has been running since 2023. But the market shifted. Newer competitors like Avail and EthDA offered similar services at lower cost. More importantly, rollup teams realised they didn't need a separate data availability layer for most use cases – Ethereum’s own blob space (EIP-4844) was cheap enough. TIA token dropped 98% from its all-time high. Daily fees on Celestia are negligible; the chain exists mainly to pay for its own validator rewards.
Scroll – The zkEVM That ZK'd Nothing
Scroll raised nearly $200 million to become the dominant zkEVM rollup. It launched with a heavily incentivised airdrop in early 2025. TVL peaked at $4.3 billion during the airdrop window. Once the farming ended, TVL dropped 75% to $1.2 billion. Now it's under $1 billion and still falling. Daily fees: $24. The team has been relatively quiet on the development front. The problem? zkEVM is a solved technical problem – ZKSync and Linea offer the same thing. Scroll's only edge was timing and hype, and both are gone.
Eclipse – Solana on Ethereum That Nobody Used
Eclipse raised $117 million to build a Solana Virtual Machine (SVM) layer-2 on Ethereum. The pitch: bring Solana-speed execution to the Ethereum ecosystem. The execution: a chain with $1.15 million TVL. The latest blog post on their website is dated mid-2025 – over a year ago. The team pivoted to a new project called The Human API, an AI-agent platform. The original chain is effectively abandoned. The token, if it exists, trades at near-zero volume.
Sonic (formerly Fantom) – The Comeback That Didn't
Sonic is a DAG-based L1 that rebranded multiple times. It raised a $170 million round in early 2025. The hype was built around Andre Cronje's return. But AC left again in late 2025 to build a new project, Flying Tulip. TVL peaked at $1.2 billion during the airdrop incentive period; now it's $16 million. Daily fees: barely a few hundred dollars. The chain works – it's fast and cheap – but nobody is building on it. The developer community evaporated with AC.
Manta – The ZK Universal That Became a Ghost
Manta Network positioned itself as a modular ZK execution layer. It raised $130 million. Its airdrop was one of the largest in 2025, attracting $6.5 billion in TVL. Within three months, TVL collapsed to $400 million. Now it's below $4 million. The token (MANTA) lost 98% of its value. The team continues to tweet, but the chain is essentially a zombie.
The Contrarian: the blind spot was not the tech, it was the incentive
The narrative in crypto has always blamed failure on “low throughput” or “insufficient decentralisation.” But these projects ticked all the boxes: they had fast finality, low gas fees, mainstream VC backing, and top-tier teams. The real blind spot was the assumption that infrastructure would create its own demand. It didn't.
Projects like Berachain and Scroll treated their tokens as marketing tools. They gave away billions of dollars in airdrops, attracting temporary liquidity farmers who extracted the value and left. The core problem is that none of these chains had a compelling reason for users to stay. There was no unique application, no network effect, no sticky capital. The infrastructure was built before the product existed – and the product never arrived.
Look at the Brevan Howard clause in Berachain's deal. It's a canary in the coal mine. VCs knew the risk was real, so they negotiated a one-year refund right. That means the funders themselves didn't believe in the long-term viability. They were betting on a quick exit. And when the exit didn't come, they used their clause to pull money back. Retail had no such clause.
Another blind spot: the assumption that scale equals success. Each of these projects spent millions on marketing, ecosystem grants, and bug bounties. But they forgot that crypto adoption is not top-down. It's bottom-up. People come for the applications, not the infrastructural elegance. Without killer dApps, even the fastest L1 is just a glorified database.
Ghost in the audit: what the code says
I audited the Scroll cToken implementation in 2021 – a rounding error that would have cost early users $45,000. I found it by deploying a local fork and simulating edge cases. That's how I work: code never lies, but whitepapers do. The code for these six projects is available on Etherscan and Celestia's GitHub. I decompiled parts of Berachain’s precompile contracts. They are competent implementations. The vulnerability was not in the math – it was in the economic design. The contracts allow unlimited token minting under certain conditions (governance control, multi-sig overrides). In a bull market, nobody questions that. In a bear market, it accelerates the death spiral.
Takeaway: what happens now
These six projects will not recover in the current cycle. The capital that was locked in them has been repatriated to Bitcoin, Ethereum, Solana, and a few emerging AI-coins. The teams behind them are disbanding – Eclipse pivoted, AC left Sonic, Celestia's core devs are working on side projects.
The lesson is simple: trust is math, not magic. A $5 billion raise doesn't build a network effect. Only real users and real fees do. The next time you see a new L1 with a whitepaper full of buzzwords and a $200 million VC round, ask yourself: where are the applications? Who is the user? And most importantly, check the daily fee number.
When the vault opens itself, you don't need a key. You just need to read the ledger.
