At the global launch of the Midnight mainnet in 2026, Charles Hoskinson announced to the audience with an impactful PPT:
The Midnight native ZK-rollup scalability solution will achieve over 10,000 TPS, reduce gas fees by 99%, and completely solve the industry's pain points related to public chain congestion and high costs, making Midnight the world's first Layer 1 public chain that balances privacy, security, and high performance.
The white paper dedicates an entire chapter to outlining the technical blueprint of this scalability solution: recursive compression proofs based on Halo2, shared security with the mainnet, EVM compatibility, second-level finality, and trustless cross-Rollup interoperability, which has been positioned by the official as the core engine for Midnight's large-scale commercial use and a key factor in attracting DeFi developers. Many KOLs in the community have also jumped on the bandwagon, claiming that Midnight will surpass Solana and Aptos to become the benchmark for the next generation of high-performance public chains with this solution.
As a developer who has been struggling in the DeFi field for 5 years, I am particularly sensitive to the scaling capabilities of public chains. Previously, while working on DEX products on Ethereum, I was repeatedly caught by high Gas fees and network congestion, with Gas fees for a single exchange transaction during bull market peaks reaching tens of dollars, and user complaints never ceased. So when I heard about Midnight's scaling promotion, I was immediately interested—if it could really achieve over 10,000 TPS while balancing privacy and security, then our DEX could completely migrate there, providing users with a disruptive experience.
In the past two weeks, based on the Rollup SDK provided by the officials, I built a complete testing environment and deployed a simplified version of the spot DEX contract, initiating a total of 1,200 transfer transactions and 800 token exchange operations, completing three rounds of full-load stress tests, recording each TPS, Gas cost, finality delay, proof generation time, and DUST consumption, while cross-referencing GitHub code submissions, official documents, and feedback from community developers.
After testing, my conclusion is chillingly cold:
The current ZK-rollup scaling solution of Midnight is purely more about publicity than implementation, with the actual TPS being less than 1/10 of what was advertised, and the reduction in Gas fees far below the promised levels, with finality in seconds completely failing; the core recursive compression proof has not been implemented at all, achieving even demo-level usability is out of reach, let alone supporting large-scale commercial use.
This article does not blow smoke or blacken, but only talks about my genuine testing experience, actual test data, and thoughts from a developer's perspective, restoring the most authentic appearance of this scaling solution.
First, TPS is overstated by 10 times, advertised as over 10,000, but the actual maximum is less than 800.
The core promise of the official scaling solution is a peak TPS of over 10,000, claiming to support high-frequency trading for millions of users simultaneously, completely solving the congestion problem of public chains.
But my three rounds of stress test results directly refuted this promise.
I used 10 testing addresses to initiate 1,000 concurrent transfer transactions at the same time, simulating the peak exchange scenario of the DEX, and the final test result is:
• Peak TPS: 782
• Average TPS: 560
• Transaction completion rate: 82%
• 18% of the remaining transactions either queued for over 20 minutes before going on-chain or timed out directly
In other words, the actual peak TPS is less than 1/10 of what the officials advertised, making it no different in essence from ordinary Layer 1 public chains. The so-called "10,000 TPS" is completely paper data.
After in-depth analysis of the logs, I found that there are two core reasons why TPS cannot go up:
First, the transaction sequencer is completely centralized. Currently, there is only one sequencer node operated by the official, with no entry for any third-party nodes. When concurrency increases, the node's processing capacity directly hits a bottleneck, leading to overflow in the transaction queue.
Second, the core recursive compression proof has not been implemented at all. The white paper repeatedly emphasizes that using Halo2's recursive proof can compress proofs for thousands of transactions into one, greatly increasing TPS. However, in practice, each transaction must generate ZK proofs separately, which cannot be compressed in batches, and the time taken for proof generation directly bottlenecks the transaction processing speed.
A friend deeply involved in ZK scaling technology directly stated after reading my test report:
"Without recursive compression, you cannot talk about ZK-rollup scaling. Midnight's current solution is just a side chain with ZK proofs, not a true Rollup. It’s surprising if TPS can go up at all."
Second, the promise of reduced Gas fees has failed, with actual costs being only 30% lower than the mainnet, far below the advertised 99%.
The other major core advantage of the official promotion is that Gas fees are reduced by 99%, making small transactions and high-frequency operations possible, completely addressing the pain point of high costs on public chains.
But my actual test data differs from the promotion by a full two orders of magnitude.
I conducted a cost comparison test with the same operations on both the mainnet and Rollup, and the results are as follows:
• Single ordinary transfer: The mainnet averages 1.1 DUST consumed, Rollup averages 0.8 DUST consumed, costing only 27% less
• Single token exchange transaction: The mainnet averages 3.2 DUST consumed, Rollup averages 2.3 DUST consumed, costing only 28% less
• Contract deployment cost: The mainnet averages 120 DUST consumed, Rollup averages 94 DUST consumed, costing only 22% less
The so-called "99% reduction in Gas fees" is completely empty talk, with actual costs being less than 30% lower than the mainnet, which makes no essential difference to ordinary users and developers.
I scoured the GitHub code repository and finally found the core reason why costs cannot come down: the cost of generating ZK proofs for Rollup is extremely high, and this portion of the cost is entirely passed on to transaction users. The official claim of "batch compression reducing costs" has not materialized due to the lack of recursive proofs, resulting in each transaction having to pay separately for proof generation, thus costs cannot be reduced.
What’s even more outrageous is that transactions failing on Rollup still incur a full charge of DUST, with no fees refunded. During two weeks of testing, I lost 127 DUST due to transaction timeout failures, and my appeals were ignored.
Third, the promise of finality in seconds has failed, with an average delay of 12 seconds, and the longest exceeding 1 minute.
The official promotion clearly states:
"Rollup transactions achieve finality in seconds, with confirmations completed within 1 second, without waiting for multiple block confirmations, delivering an experience comparable to centralized exchanges."
But the actual test results show that the finality delay of transactions does not meet the advertised standards.
I compiled the finality confirmation times for 2,000 transactions, and the results are as follows:
• Average finality delay: 12.3 seconds
• Minimum delay during low peak: 4.7 seconds
• Maximum delay during peak period: 67 seconds
• More than 30% of transactions have a finality delay exceeding 20 seconds
The so-called "finality in seconds" has not been achieved at all, and is even slower than the transaction confirmation speed of the mainnet.
After my analysis, I found that the core reason for the delay is: the Rollup transaction proofs must be submitted to the Midnight mainnet for verification, and only after passing verification is the transaction considered final. The block time of the mainnet itself is 12 seconds, and coupled with the time required for proof generation, data submission, and on-chain verification, achieving finality delays of less than 1 second is fundamentally impossible.
The officials deliberately concealed the verification process of the mainnet in their promotion, failing to mention the true situation of finality delays, painting a picture for developers and users that is fundamentally unachievable.
Fourth, the completely centralized architecture has completely failed the trustless security promise.
The white paper repeatedly emphasizes that the Midnight ZK-rollup solution shares security with the mainnet, is completely trustless, has decentralized sequencing and proof generation, and eliminates single points of failure.
However, in practice, the centralization level of this solution is even higher than that of most consortium chains.
First, the sequencer is completely centralized. Currently, there is only one sequencer node operated by the official, with no entry for any third-party nodes, no staking mechanism, and no penalty rules. If the official node goes down or is attacked, the entire Rollup network will directly collapse, and all transactions will be interrupted.
Second, the proof generation is completely centralized. All ZK proofs for transactions must be generated by the official proving nodes; users cannot generate proofs on their own, nor can they verify the authenticity of the proofs, and must unconditionally trust the official nodes.
Third, the contract upgrade rights are completely centralized. The core contracts of Rollup have all administrator rights held by the official multi-signature address, with no time locks, no community governance voting mechanisms, allowing the officials to modify contract rules, upgrade contract logic, and even freeze user assets at any time, with users having no checks and balances.
This means that the current Rollup solution has not only failed to achieve "shared security with the mainnet," but has instead introduced a higher risk of centralization; the so-called "trustless, decentralized" is entirely empty talk.
Fifth, my final conclusion: the scaling solution is still in the prototype stage, and developers should not blindly migrate.
After two weeks of in-depth testing, I completely abandoned the plan to migrate the DEX to Midnight, deleted all testing environments, and continued to iterate products on Ethereum Layer 2.
Midnight's ZK-rollup scaling solution is not just a few bugs; there are fatal problems throughout the entire link, from core technology to architectural design: the advertised peak TPS of 10,000 is actually less than 800; the promise of a 99% reduction in Gas fees is actually less than 30%; finality in seconds has completely failed; the core recursive compression proof has not been implemented; the architecture is completely centralized with extremely high security risks.
It does not possess any conditions for production environment usage at all, and even testing during the development phase is very difficult. Using it to develop products will only provide users with a very poor experience, bringing uncontrollable risks to the project.
My clear suggestion is:
Before Q4 2026, all DeFi developers and project parties should not attempt to develop products based on Midnight's ZK-rollup solution, as performance, cost, and security risks are completely uncontrollable.
I still recognize Midnight's long-term vision for privacy and compliance, and I believe ZK-rollup is the ultimate direction for public chain scaling. But I refuse to pay for false publicity, refuse to bear the trial and error costs of a half-baked technical solution, and will not gamble the future of the project and the assets of users on an unverified concept.
I hope that IOG can completely restructure this scaling solution in the upcoming versions and fulfill their promises: to achieve recursive compression proofs, increase TPS, reduce Gas costs, open up decentralized sequencer and prover networks, and improve security mechanisms, truly providing developers with a usable, trustworthy, high-performance scaling solution.
Otherwise, no matter how grand the high-performance narrative is, it cannot support a public chain that cannot even handle core scaling capabilities.
#Midnight $NIGHT @MidnightNetwork #night
