Sabitlenmiş Tweet
Pushing the Limits of Zero-Knowledge: Zetta, ZAN, and the Road Ahead for Succinct
There’s been a quiet but significant shift underway in how zero-knowledge proofs are generated. And it’s not only about software anymore.
At the center of this shift is Zetta, a purpose-built FPGA accelerator created for SP1, Succinct’s high-performance zero-knowledge prover. This isn’t just a hardware upgrade. Zetta was built from the ground up to handle the most computationally intense parts of the proving pipeline. It tackles tasks like CosetNTT computations, Merkle tree generation, permutation traces, and FRI folding.
It runs on the AMD Alveo U55C. What really makes it stand out is how modular and programmable it is. That flexibility matters in the long run. Zetta can adapt to new circuits, including those that haven’t even been developed yet. Some might call that future-proof, though perhaps it’s better to say it’s prepared for what’s next. The benchmarks help tell the story. With support from AntChain OpenLabs, Zetta was compared against SP1’s CPU implementation. The performance gains were substantial. Proofs that once took their time now complete much faster.
Still, Zetta on its own doesn’t complete the picture. Hardware needs people who know how to run it at scale. That’s where ZAN enters.
ZAN, a team working closely with AntChain OpenLabs, has joined the Succinct Prover Network as a multi-node cluster prover. If Zetta is the engine, ZAN is the one building out the system that puts it to work. They’ve got practical experience running GPU clusters, tuning hardware, and managing infrastructure in real-world settings. They also played a role in the FPGA acceleration work that brought Zetta to life, which says something about their depth.
Now part of the Prover Network, ZAN will be running proofs, participating in live proof auctions, and delivering production-grade infrastructure to customers who would rather not think about the complexities behind it. It’s not just about adding more machines. It’s about improving how those machines are used.
GPUs are still widely used across zero-knowledge workloads. Even so, FPGAs offer a unique edge in certain scenarios. They can often be more cost-effective than GPUs, especially cloud-based ones that can be expensive to operate. Over time, those cost savings can make a real difference. Speed is important, of course, but there’s also something to be said for accessibility and sustainability.
There’s another layer to all of this that’s worth keeping in mind. Zetta’s design hints at a future where zero-knowledge proof systems are built directly into silicon. ASICs may not be ready just yet, but it’s clear the groundwork is being laid. Zetta isn’t just a one-off project, it’s also a step toward a more integrated, hardware-native approach to ZK performance.
So now, with Zetta raising the bar on performance and ZAN turning those capabilities into usable, production infrastructure, Succinct is moving closer to its broader vision. The team has long talked about building the largest proving cluster in the world. That kind of goal can sound aspirational. Maybe it is. But with partners like ZAN stepping in, and with proven systems like Zetta already in play, it’s beginning to feel more grounded.
There’s still work to be done. Zero-knowledge is a complex space, and progress often comes in small, incremental steps. But something important is unfolding here. The future of zero-knowledge will be shaped not just by better software, but by better hardware too.
And little by little, we’re watching that future come together.
Sources and references:
@pumatheuma
@SuccinctLabs
@mvkise
@0xCRASHOUT
blog.succinct.xyz/zan/
blog.succinct.xyz/sp1-fpga-antch…
Succinct@SuccinctLabs
ZAN joins the Succinct Prover Network as a multi-GPU prover.
English
