Scale or Die at Accelerate 2025: Proving Solana: ZK Integration at the eBPF Level

In a groundbreaking presentation at Accelerate 2025, Yiwen Gao from Nitro Labs unveiled revolutionary advancements in Solana's scaling capabilities through ZK-SVM technology. This innovation promises to transform off-chain computation, enhance privacy, and dramatically reduce rollup times from days to mere hours.

Summary

Nitro Labs has successfully integrated Zero-Knowledge proofs with Solana's eBPF-based virtual machine, creating a powerful new tool for secure off-chain computation. This breakthrough, known as ZK-SVM, allows for the generation of ZK proofs for Solana transactions, enabling a wide range of applications from ultra-fast rollups to private, verifiable computations.

The team overcame significant technical challenges, including issues with randomness, threading, and bit depth discrepancies between the 64-bit Solana environment and the 32-bit ZK-VM. These solutions have been partially upstreamed to Solana's agave codebase, paving the way for broader adoption and further innovation.

Nitro Labs' work is part of a larger initiative to create modular scaling solutions for Solana, including dedicated SVM engine instances and data anchoring technology. These tools aim to help developers scale their applications without resorting to separate networks or app chains, maintaining Solana's unified ecosystem approach.

Recent optimizations have dramatically reduced the time and cost of generating ZK proofs for Solana transactions. What once took 40 minutes and cost over $5 can now be accomplished in under 5 minutes for less than a dollar, making continuous proving of Solana blocks economically viable.

Key Points:

ZK-SVM Technology

Nitro Labs has developed a method to generate Zero-Knowledge proofs for Solana Virtual Machine (SVM) transactions. This technology, dubbed ZK-SVM, allows for secure off-chain computation that can be verified on-chain. The primary application is in the context of SVM rollups, where it can replace the traditional week-long challenge window with a simple ZK proof, reducing confirmation times to just a few hours.

Beyond rollups, ZK-SVM opens up possibilities for any type of complex or privacy-sensitive computation. Developers can now generate ZK proofs for SVM transactions off-chain, submit these proofs to be verified on-chain, and trigger further on-chain actions upon successful verification. This effectively allows for off-chain computation to have on-chain impact, greatly expanding the potential use cases for Solana's smart contract capabilities.

Nitro Labs' Scaling Solutions

The ZK-SVM technology is part of a broader suite of scaling solutions being developed by Nitro Labs for their Termina platform. Their goal is to provide building blocks for scaling on Solana without requiring developers to create separate networks like traditional rollups or app chains.

In addition to ZK-SVM, Nitro Labs is working on two other key modules:

1. SVM Engine: This allows for dedicated SVM engine instances, useful for offloading compute during periods of network congestion or high contention.
2. Data Anchor: This module enables putting large amounts of data on-chain at a fraction of the usual cost, addressing data scaling challenges.

These modules, working in concert, aim to provide a comprehensive toolkit for developers looking to scale their applications within the Solana ecosystem.

Technical Challenges and Solutions

Implementing ZK-SVM presented several unexpected challenges that the Nitro Labs team had to overcome:

1. Randomness and Time: The agave codebase (Solana's eBPF interpreter) uses randomness and time functions that are incompatible with the deterministic nature of ZK proofs. The team replaced these with deterministic alternatives, including a custom implementation of a repeatable pseudo-random number generator.

2. Threading and File System Access: ZK-VMs are single-threaded and don't allow file system access, while agave uses multiple threads and file I/O. Nitro Labs simplified the execution to a single thread without background processes, which actually resulted in more streamlined code and smaller ZK proofs.

3. Bit Depth Discrepancy: The most significant challenge was the mismatch between agave's 64-bit operations and the 32-bit environment of the ZK-VM. This caused memory access issues that required creating new data structures to ensure compatibility. These changes have been partially upstreamed to the agave codebase, benefiting the wider Solana developer community.

Performance Improvements

Since the initial publication of their work in December 2024, Nitro Labs has made significant optimizations to the ZK-SVM technology:

Initially, proving 100 complex SPL transactions took about 40 minutes and cost $5.32. Through optimizations such as reducing the cycle count from 7 billion to 3 billion, they've dramatically improved performance. As of April 2025, the same workload can be processed in just 4-5 minutes at a cost of 80-90 cents.

These improvements make continuous proving of Solana blocks much more feasible, both in terms of time and cost efficiency. This opens up new possibilities for real-time ZK applications and more economical scaling solutions.

Facts + Figures

• ZK-SVM can reduce rollup challenge windows from 7 days to just a few hours
• Initial benchmarks in December 2024: 100 complex SPL transactions took 2300 seconds (≈40 minutes) to prove and cost $5.32
• Latest benchmarks in April 2025: Same workload now takes 240 seconds (4-5 minutes) and costs $0.80-$0.90
• Cycle count for proofs reduced from 7 billion to 3 billion through optimizations
• Agave (Solana's eBPF interpreter) typically runs on 64-bit machines
• ZK-VMs are generally based on a 32-bit RISC-V instruction set
• Nitro Labs created new 16-byte structs to bridge the 64-bit to 32-bit gap
• Agave uses at least 7 threads for transaction processing
• Three sets of changes have been proposed to upstream agave, with two already merged

Top quotes

1. "Our goal is to create building blocks for scaling on Solana."
2. "If you replace that process with a simple single ZK proof, then you can shorten that down from seven days to just a few hours or even less."
3. "We knew since the beginning that we didn't want to create custom circuits. We didn't want to roll our own ZK circuits for doing this, because that's super time-intensive and our team is not a team of cryptographers."
4. "In a ZK-VM, everything needs to be deterministic."
5. "This works in the proving case because we don't have the same performance requirements as a live consensus network."

Questions Answered

What is ZK-SVM and why is it important for Solana?

ZK-SVM is a technology developed by Nitro Labs that allows for the generation of Zero-Knowledge proofs for Solana Virtual Machine transactions. It's important because it enables secure off-chain computation that can be verified on-chain, opening up new possibilities for scaling, privacy, and complex computations on Solana. This technology can dramatically reduce rollup times from days to hours and allow for private, verifiable computations within the Solana ecosystem.

How does ZK-SVM improve rollups on Solana?

ZK-SVM improves rollups on Solana by replacing the traditional fraud-proof mechanism with a single ZK proof. In traditional optimistic rollups, there's a 7-day challenge window where sequencers and challengers can dispute transactions. With ZK-SVM, this process is replaced by a simple ZK proof that can be verified quickly on-chain. This reduces the challenge window from 7 days to just a few hours, greatly improving the speed and efficiency of rollups on Solana.

What technical challenges did Nitro Labs face in implementing ZK-SVM?

Nitro Labs faced three main technical challenges in implementing ZK-SVM:

1. Dealing with randomness and time functions in agave (Solana's eBPF interpreter) that are incompatible with the deterministic nature of ZK proofs.
2. Adapting agave's multi-threaded, file system-dependent operations to work in a single-threaded ZK-VM environment without file system access.
3. Resolving bit depth discrepancies between agave's 64-bit operations and the 32-bit environment of the ZK-VM, which caused memory access issues. These challenges required significant modifications to the agave codebase and the creation of new data structures to ensure compatibility.

How has the performance of ZK-SVM improved since its initial development?

The performance of ZK-SVM has improved dramatically since its initial development. In December 2024, proving 100 complex SPL transactions took about 40 minutes and cost $5.32. Through various optimizations, including reducing the cycle count from 7 billion to 3 billion, Nitro Labs has significantly improved these metrics. As of April 2025, the same workload can be processed in just 4-5 minutes at a cost of 80-90 cents. This represents an approximately 10x improvement in both time and cost efficiency.

What are the potential applications of ZK-SVM beyond rollups?

Beyond rollups, ZK-SVM has potential applications in any scenario requiring complex or privacy-sensitive computations. Developers can use ZK-SVM to generate proofs for any type of SVM transaction off-chain, then submit these proofs for on-chain verification. Upon successful verification, further on-chain actions can be triggered. This enables off-chain computations to have on-chain impacts while maintaining privacy and reducing computational load on the main chain. Potential use cases could include private voting systems, complex financial calculations, or any application requiring verifiable off-chain computation.