# **Ethproofs 2025 Review & 2026 Roadmap for Real-Time Proving**

**December 6, 2025**

## **2025 in Review: Building the Real-Time Proving Ecosystem**

It’s been an extraordinary nine months since we launched Ethproofs on Jan 28, 2025. What started as a simple dashboard to track zk proofs for Ethereum L1 blocks has become a central hub for the zkVM ecosystem. We’ve onboarded 7 zkVMs and 15 provers, verified ~200k blocks, and watched latency fall 5× while costs dropped 15×. Your feedback and daily proving runs have shaped everything. Thank you.

As we push toward the North-star goals outlined in the [zkEVM team’s EF blog post](https://blog.ethereum.org/2025/07/10/realtime-proving)(≤10s P99 latency, ≤$100k CAPEX, ≤10kW power, 128-bit security, ≤300KiB proofs, fully OSS) the speed of progress this year has made one thing obvious: **Ethproofs must rise with the ecosystem**. As proving becomes faster, cheaper, more distributed, and more reproducible, the platform needs to evolve as well—simplifying where things have grown complex, strengthening where the ecosystem now depends on us, and raising the bar on clarity and comparability.

Before outlining what’s ahead, it’s worth revisiting what the ecosystem achieved in 2025—because those milestones illuminate exactly where Ethproofs should go next.

---

## **2025 Highlights: A Year of Breakthroughs (Top Ten)**

### **1. 100% Reproducibility Across All Provers**

Every prover is now fully reproducible, with direct links to open-source repos or verified binaries. This is foundational for transparency and credibility.

### **2. All Verifiers Are Dual-Licensed Open Source**

Every verifier is now MIT + Apache 2.0. This removes ambiguity, broadens reuse, and reinforces platform neutrality.

### **3. Multi-GPU & Real-Time Proving Became First-Class**

Ethproofs expanded from single-GPU 1:100 proving to full support for multi-GPU and 1:1 every-block proving. Four teams now hit **sub-10s P99** using clusters under **$100k CAPEX** and **10kW power**—directly aligned with zkEVM guidance.

### **4. Single-GPU Provers Got Fast**

Single-GPU proving dropped from **16 minutes to under 60 seconds** for most teams—one of the year’s biggest decentralization wins.

### **5. Proof Costs Collapsed**

Costs fell from **$1.69** to **under a penny** per proof.

### **6. A Growing Mesh of Always-On Provers**

Multiple teams now run always-on provers submitting proofs for every block—early evidence of decentralized real-time proving.

### **7. Hosting the EF Client-Side Proving Benchmarks**

Ethproofs became home to the EF’s **Client-Side Proving (CSP)** benchmarks, publishing quarterly metrics for local proving across circuits and zkVM stacks. This collaboration is a template for future neutral, public benchmarking efforts.

### **8. zkLighthouse: Validating Blocks via Ethproofs Proofs**

A modified Lighthouse client demonstrated live validation of Ethereum blocks using zk proofs fetched from Ethproofs, verifying multiple proofs per block in milliseconds. This experiment previews a future where **mandatory proving** is part of fork choice and consensus.

### **9. Building the Ethproofs Community**

The community flourished:

* Telegram grew to **600+ members**
* **Six** major community calls drew ~250 participants each
* **Ethproofs Day** at Devconnect brought **600+ attendees** for demos, talks, and live proving

### **10. Ethproofs Formally Becomes an EF Team**

The EF established a clear, coordinated structure around zkEVMs and real-time proving. Within this, Ethproofs was recognized as an EF team and now serves as the interface between public research and private proving. Alignment across zkVMs, prover operators, and EF workstreams allowed the ecosystem to move together toward mandatory proving.

---

These accomplishments show the zkEVM and prover communities converging on a shared goal: **decentralized, verifiable, real-time proving at commodity scale.**
To support that, Ethproofs’ platform and metrics are evolving in 2026.

---

## **2026 Roadmap: Scaling, Simplifying, and Preparing for Mandatory Proving**

### **Standardizing Cost via a 5090-Based Model**

Our current cost-per-proof system relies on stale cloud pricing, unverifiable assumptions about operator choices, and backend complexity. It also nudges provers toward centralizing behaviors.

Going forward, all cost will be normalized through a **single, fixed hourly rate for an RTX 5090**. This provides a consistent, verifiable baseline without depending on cloud dynamics.

---

### **Why 5090s Are the Ethproofs Unit of Measure**

By mid-2025, nearly every prover—across zkVMs, cloud-hosted and bare metal alike converged on the RTX 5090 or clusters of them. Once optimizing for real-time proving, **5090s emerged as the globally available, consumer-grade, high-parallelism sweet spot** the zkEVM team anticipated.

**Benefits:**

* excellent balance of cost, availability, parallelism, bandwidth, and power
* aligned with Ethereum’s decentralization ethos
* strong performance for proving workloads
* accessible to home and office provers

**Compared to alternatives:**

* data-center GPUs (H100, A100, MI300) are expensive, gated, and misaligned with decentralization
* earlier consumer GPUs can’t meet sub-10s P99 goals

Since 90%+ of clusters are already 5090-based, Ethproofs will embrace this reality. The 5090 becomes the **unit of measure** for proving capacity, akin to gas for EVM compute. Future hardware (including CPU-based provers like Jolt) will be added as they mature.

---

### **A Temporary Constraint with Long-Term Benefits**

The 5090 standard is a pragmatic simplification, not dogma. When new consumer hardware or CPU-based proving reaches parity, we will expand the profiles.

Starting with one hardware baseline delivers:

* cleaner metrics
* consistent comparisons
* a better user experience
* support for user-led, permissionless onboarding at scale

These benefits remain even as the hardware universe expands.

---

### **Streamlined Onboarding**

Onboarding historically required manual back-and-forth, cloud SKU detective work, and metadata debugging--all for very little ROI. This doesn’t scale to a future with:

* global home provers
* hundreds or thousands of operators
* permissionless submissions

With standardized hardware:

* provers report only `num_gpus`
* Ethproofs handles the rest (power, capacity, normalization)
* onboarding becomes instant and self-serve

---

### **Enabling the Long Tail of Home Provers**

Our long-term goal is **thousands** of independent provers—enthusiasts, researchers, students, solo stakers—running one or two GPUs from home submitting proofs to Ethproofs.

To get there:

* onboarding must be simple
* metrics must be uniform
* hardware assumptions must be globally accessible
* dashboards must be self-populating

Standardized profiles unlock this scale.

---

### **The Four Proving Modes**

With hardware standardized, the prover ecosystem resolves cleanly into four modes:

1. **Single GPU – 1:100 (altruistic)**
2. **Single GPU – 1:1 (real-time)**
3. **Multi GPU – 1:100 (batch)**
4. **Multi GPU – 1:1 (real-time)**

This allows clean leaderboards, automated onboarding, consistent visualizations, and cleaner data.

*(Note: we expect to raise the 1:100 threshold for single-GPU provers to 1:10 by mid-2026.)*

---

### **A More Aligned Metric: Electricity (Watts) per Proof**

Dollar-denominated cost is nice to see, but far from a northstar metric at this point in the journey. We’re shifting to **energy consumption per proof (kWh)**—a neutral, verifiable, decentralization-preserving metric.

**Why energy is better:**

* universal
* independent of geography
* verifiable via GPU TDP
* naturally improves as zkVMs get faster
* environmentally aligned with Ethereum post-Merge

**How it’s measured:**

> **Energy per proof (kWh) = num_gpus × TDP × proving_time_ms / 3,600,000**

Dashboards will move from cost-per-proof to energy-per-proof.

---

### **Centering Liveness: Reliability & Block Difficulty**

Ethproofs will stand out by tracking **who is reliably online, slot after slot**, under realistic conditions.

#### **Online Reliability**

Tracked per prover, per mode:

* single 1:100
* single 1:1
* multi 1:100
* multi 1:1

#### **Sub-8s P99 as the New Target**

We align with the zkEVM team’s tighter requirement:

* % of proofs under 8s
* P99 latency
* long-tail distributions

#### The High-Reliability “RTP Cohort”

On a monthly basis, we'll maintain a curated set of top-performing provers—our RTP Cohort—defined as provers that:

* achieve P99 < 8 seconds
* maintain ≥90% online reliability

The RTP Cohort represents the ecosystem’s most consistent real-time provers and serves as the reference set for zkLighthouse-style consensus experiments.

#### Surfacing “RTP Prover Killers” (Difficult Blocks)

When multiple RTP Cohort provers fail to produce a proof for the same block within 8 seconds, that block will be automatically classified as a “RTP prover killer.” These blocks populate a continuously updated research block list—a targeted dataset for cryptographers, zkVM teams, and client developers to study pathological edge cases, performance cliffs, and other anomalies.

---

## **Looking Ahead to 2026: Metrics, Benchmarks, and Mandatory Proving**

2026 will focus on aligning with zkEVM guidance and preparing for a world where **proving becomes mandatory** in consensus.

Key areas:

1. **Liveness Tracking**
Reliability becomes first-class, especially for 1:1 provers.

2. **Electricity per Proof**
An accurate, neutral measure of efficiency.

3. **Security Metrics**
Clear reporting on bits-of-security, formal verification, and robust assumptions. More on this soonTM.

4. **Difficult Blocks (aka RTP prover killers)**
Automatic surfacing of challenging blocks for focused research.

5. **Expanded Benchmarking Collaborations**
Building on CSP, we’ll support new EF efforts—particularly the **leanVM project** for PQ-safe signature aggregation.
Ethproofs will host **recursion-only benchmarks**, including:

* 1-to-1 recursive proving
* n-to-1 recursive aggregation

These metrics generalize well to future zkVMs.

All of this supports zkLighthouse-style client work as we progress from **phase 0 → phase 1 → phase 2** toward mandatory proving.

---

## **Wrapping Up: The Metrics That Will Shape the Next Era**

We’re evolving our metrics not because the old system was “bad,” but because the ecosystem has outgrown it. The future of zk proving is decentralized, energy-aware, reproducible, always-on, and—eventually—baked into consensus. Our metrics need to reflect that future.

And as we move into 2026, we’re doubling down on the community that makes all of this possible. We have a full slate of initiatives designed to maintain alignment, momentum, and shared direction across the proving ecosystem:

* **A new private “Ethproofs Buildooors” channel** for teams submitting proofs, along with aligned developers and researchers. *(DM me for details: @corcoranwill)*
* **More frequent community calls**, beginning with the upcoming call on **December 19th**.
* **An Ethproofs event titled “Beast Mode”** and a full **Ethproofs Hacker House in Cannes, France**, taking place just before EthCC.

* Beast Mode: [https://lu.ma/beast_mode](https://lu.ma/beast_mode)
* Fort Mode: [https://lu.ma/fort_mode](https://lu.ma/fort_mode)
* Villa Lean (Hacker House): [https://lu.ma/villa_lean](https://lu.ma/villa_lean)

These gatherings are designed to keep the ecosystem tightly connected as we push toward mandatory proving and real-time verification together.

Thank you for building with us, challenging us, and pushing this space forward. Ethproofs is a neutral steward—but the progress belongs to you.

**Onward to Mandatory Proving!**