The Billion-Dollar Bottleneck: Why Your Crypto Transactions Are So Expensive | Opinion

High fees on crypto transactions are often attributed to a complex process known as zero-knowledge (ZK) proving. This article explores the reasons behind these expenses and proposes solutions to make transactions cheaper and more efficient.

Key Takeaways

The high fees associated with ZK-rollups are primarily due to the intensive computing required for generating zero-knowledge proofs.
Over 60-70% of fees on ZK-layer 2s (L2s) arise from this proving process, creating a centralized market dominated by a few providers.
Solutions lie in developing specialized hardware and open markets for proving to significantly reduce costs.
Cheaper proving can open the door to a new era of decentralized applications with efficient, low-cost transactions.

To solve the bottleneck problem, we need a global bazaar for proofs, where anyone contributes, everyone verifies, and the open web is secure.

Understanding the Cost of Your Crypto Transactions

Imagine trying to transfer $50 from Ethereum to a ZK-rollup, like zkSync, and facing a transaction fee between $0.15 and $0.50. This is frustrating, especially when alternative rollups, like optimistic rollups, often cost three times less—plus, recent upgrades like Dencun reduced data costs by over 90%. The real issue here isn’t network congestion but rather the zero-knowledge proof itself.

Creating a zero-knowledge proof for batches of transactions is a demanding task. It requires performing trillions of mathematical operations—mainly elliptic curve multiplications—on hardware that isn’t optimized for this function. According to L2Beat, around 60-70% of all fees on ZK-layer 2s are attributed to this proving process.

This situation has given rise to a centralized market known as "prover farms," estimated to be valued at over $97 million. This represents a bottleneck worth billions, highlighting how a centralized and resource-intensive system hampers the potential of ZK scaling.

Breakdown of Transaction Fees

When you conduct a ZK-rollup transaction, there are various costs involved. While L2 execution is almost negligible and data posting on the mainnet is now affordable due to blobs, the proving is where expenses truly mount. Generating a single proof for a batch of 4,000 transactions can take between two to five minutes on sophisticated GPUs, costing between $0.04 and $0.17 solely in cloud computing fees. When demand is high, this can escalate to over ten cents per transaction.

It's important to note that ZK proving differs from the matrix math used in artificial intelligence. It depends on multi-scalar multiplications (MSMs) and number-theoretical transforms (NTTs) over elliptic curves—operations that require a lot of GPU power. Many GPUs remain idle during this process, as tests show that approximately 80% of their cycles are wasted when generating these proofs.

For instance, a report from zkSync Era revealed that proving accounted for 65% of their total revenue of $2.3 million in Q1 2025. This translates to $1.5 million being funneled to centralized systems, reinforcing the issue of centralization in the ZK space.

Optimistic rollups, while offering lower costs, come with considerable drawbacks. They impose a minimum seven-day withdrawal delay for transaction verification and trust in validators. For numerous real-world applications requiring speed, such as DeFi, these delays are simply too cumbersome.

To truly enable scalable, affordable, and trustworthy transactions, we must rethink how proof generation is undertaken rather than avoiding it altogether.

The Problem of Centralization

More than 90% of ZK-rollups rely on a small group of "prover-as-a-service" companies, creating significant vulnerability and a single point of failure. For example, in March 2025, Blast operators temporarily froze transactions from attackers, impacting 12,000 innocent users for 48 hours. Centralized provers can see plaintext batches, take excess blockchain value (MEV), and can refuse transactions.

This reliance on a few dominant players leads to systemic weaknesses. Even though top provers claim 99.2% uptime, a single outage can cause significant disruptions. In fact, Starknet lost 20% of its throughput due to one node failure in Q2 2025. Furthermore, these providers are capturing 80% of profits, leaving rollups financially strained, while relying on Ethereum's first layer for security. As prominent figure Vitalik Buterin mentioned at Devcon 2024, relying on external proving essentially negates the benefits of decentralization.

A Path Forward

To address this pressing issue, a two-part strategy is necessary. Firstly, we need specialized hardware tailored specifically for zero-knowledge tasks, and secondly, we require a marketplace for these proving services.

Current GPUs are like race cars on a conventional road; they excel in AI tasks but struggle with the elliptic curve operations used in ZK proofs. This is where domain-specific ASICs play a crucial role. Winning participants in the 2025 ZPrize competition managed to achieve STARK proof generation speeds of 6-8 seconds on FPGAs, which is significantly more efficient than GPUs. Cysic has developed ASICs that can produce proofs for large circuits in under two seconds while consuming 50 times less energy. This is reminiscent of the transition in Bitcoin mining from CPUs to specialized ASICs, only this time it's focused on verifying truths rather than hashes.

Secondly, we must innovate how we approach the proving marketplace. Moving beyond the software-as-a-service (SaaS) model, we need to establish a transparent marketplace for proving resources. For example, in Succinct’s 2025 mainnet, rollups can auction proving jobs to specialized rigs that can respond in seconds. This approach has led to a 40% reduction in costs, fairer profit distribution, and resistance to censorship.

Unlocking New Opportunities

If we can reduce the costs of proving, we could see transaction fees plummeting to below $0.01, with 15,000 transactions per second achievable, according to the zkSync Atlas roadmap. This could pave the way for pay-per-pixel NFTs, dynamic gaming economies, and AI-driven trades without concerns about front-running. Layer 2 solutions would transform into neutral infrastructure instead of monopolistic proving domains.

This would encourage developers to concentrate on optimizing virtual machines rather than managing extensive server setups, catalyzing significant innovation. The total value locked (TVL) in ZK technologies could increase from $28 billion in 2025 to over $100 billion. The potential impact is systemic.

When proving becomes cost-effective and decentralized, the entire value proposition will crystallize. Users will experience better speed and lower costs, developers will have reliable platforms without needing to trust a limited number of operators, and Ethereum will achieve scalable solutions that uphold security without compromising on essential principles.

Conclusion: The Challenge Ahead

The main hurdle to widespread adoption of ZK technology isn't only related to cryptocurrency or gaming but rather the underlying economics of engineering. Current centralized infrastructure is hindering the scalability needed to tap into the benefits of ZK. The battle is not over consensus but rather about creating a robust hardware and marketplace ecosystem for verifiable computations. By deploying decentralized and accelerated proving solutions, we can offer affordable, trustless transactions for a broader audience.

We seek to move beyond centralized provers to establish a global bazaar for proofs where everyone contributes, everyone verifies, and the integrity of the open web is secured.