Loading...

ESP

AI-powered news analysis from 85+ sources

articles

5

Long

4

Short

0

m

meme10/23meme

Getting Espresso to 1 million TPS meant rebuilding consensus, networking, block building, and data availability as one system, instead of tuning four separate ones. Here’s a deep dive about how our engineers did it: https://medium.com/@espressosys/the-engineering-optimizations-behind-espressos-new-performance-benchmarks-b81f105e73e1

55/100·BLong

m

meme10/20meme

reply Espresso is built for what comes next: data-heavy, trustlessly verifiable apps that still finalize in real time. Private payments. Crosschain collateral management. RWAs that carry their own compliance data. Quantum-safe apps. Espresso is ready when they are. https://x.com/1491564545931984902/status/2112654489870352609

35/100·CLong

m

meme10/20meme

Espresso mainnet can now handle up to 1 million transactions per second. And finality stays under a second, even at that peak load. How we did it, and why it matters ↓

72/100·B+Long

n

news10/20news

Scale Without Sacrificing Speed: Espresso Now Handles Up to 1 Million TPS with Subsecond Finality One million transactions per second, each finalized in under a second. That’s Espresso mainnet’s newly benchmarked performance ceiling, following an upgrade the network’s validators completed last week. We benchmarked it on 100 nodes with emulated global network latency. Espresso was built to be a base layer for independent chains and financial applications, which means staying fast at peak volume while leaving room for the much larger transactions ahead as post-quantum signatures and privacy proofs go mainstream. At 1 million TPS with subsecond finality, Espresso can do both. Most blockchains trade throughput for speed, or speed for throughput. Delivering both meant rethinking how Espresso handles consensus, networking, block building, and data availability. Espresso by the numbers Our latest benchmarks, achieved on 100 nodes with emulated global network latency: 181 MB/s of data throughput, equivalent to 1 million stablecoin transfers per second assuming an average transfer size of 180 bytes 337ms average finality time under normal conditions 778ms average finality time under peak load of 1 million TPS More than 27 million blocks finalized since Espresso mainnet launched in November 2024 Running fully permissionless proof-of-stake consensus since early 2026 Integrated chains include ApeChain, with Celo, LitVM, and MegaETH in the process of integrating A base layer built for the job, not retrofitted for it Espresso leaves execution entirely to the chains and applications that integrate with it. That means Espresso has no VM, which makes sense when you step back and ask what a base layer for other execution environments really needs to do. The list is short: Reach BFT consensus on transaction ordering and data availability Handle massive throughput Do it all with fast, deterministic finality Execution isn’t on the list. That leaves Espresso-integrated chains free to pick their own VM, fees, sequencer, and privacy and compliance rules while still inheriting Espresso’s decentralized security, real-time finality, and throughput. Much as today’s internet infrastructure doesn’t care whether it’s serving a streaming movie platform, VR game, or a news blog, Espresso is agnostic to how the chains using it are designed, including whether they bridge to Ethereum, Solana, Litecoin, or nowhere at all. Leaving out execution is also what makes the performance possible, because Espresso’s 100 globally distributed nodes can devote themselves to one job: reaching consensus on ordering and data availability as fast and at as large a volume as possible. HotShot, Espresso’s purpose-built proof-of-stake BFT consensus protocol, was designed for exactly that. General-purpose chains weren’t. Ethereum was built as a single execution environment; when it adopted its rollup-centric roadmap, it quickly became clear that its slow finality while not an issue in its own execution environment (because of atomic execution) was causing massive fragmentation among the chains that settle to it. On Solana, posting large amounts of arbitrary data is costly enough that an ecosystem of chains settling to it never formed. Neither was built to be a base layer for other chains, and that’s okay. Espresso was. Thousands of independent chains We expect the world’s financial activity to run across thousands of independent chains, each with its own execution environment, rules, and users. Crypto calls them L2s, rollups, or appchains, yet once you strip away the labels, each one is simply an application built to its own specs, batching transactions and sending them to a base layer that uses BFT consensus to finalize ordering and data availability. What these applications need from that base layer is decentralized security, fast finality, and trust-minimized interoperability, without being asked to give up control of their own architecture. As more of the world’s applications, and its capital markets, become independent systems instead of features bolted onto someone else’s chain, the question stops being which chain wins and starts being whether all of them can act on each other’s state at the speed and volume real markets require. That comes down to finality time. Financial systems built as L2s on Ethereum wait 16 minutes for finality, splintering markets into isolated islands that are stitched back together by third-party bridges, trusted RPCs, and messaging protocols that each add cost, latency, and additional points of failure. With fast enough finality, they interconnect from the start. That’s why this milestone matters beyond the number itself. Subsecond finality at 1 million TPS is the strongest evidence yet that independent chains inheriting security and finality from a base layer are the right architecture for building onchain, whether the chain belongs to a bank, an exchange, or a crypto-native team. Achieving 1 million TPS Real markets put three demands on the infrastructure underneath them. It has to: Keep up with their busiest days, not just their average ones. Stay ahead of where transactions are headed. Today’s transactions are relatively small. Post-quantum signatures and built-in privacy proofs will make them much larger, and only higher-throughput infrastructure will be able to carry them without slowing down. Stay fast under pressure. A base layer that finalizes one transaction quickly but slows to a crawl at peak volume isn’t infrastructure real-world applications can reliably build on. That’s the bar our benchmarks were built to test. Espresso’s latest benchmarks place its erasure-coded data availability layer at 181 MB/s of throughput. Assuming an average stablecoin transfer size of 180 bytes, that puts the network’s theoretical throughput ceiling at roughly 1 million stablecoin payments per second. For context, that’s about six times the peak data throughput recorded on Solana (29.68 MB/s)¹, and well beyond Tempo (3.74 MB/s)², BNB Chain (1.18 MB/s)³, Arc (0.48 MB/s)⁴, Avalanche (0.43 MB/s)⁵ and Ethereum (0.35 MB/s)⁶. These results come from a benchmark run on 100 nodes with emulated global network latency; we added artificial latency to all network calls that matches the real latency nodes would experience were they deployed globally, though the nodes themselves were run in the same region. That distinction matters as throughput benchmarks depend heavily on the test conditions under which they were recorded: Nodes passing messages across a server rack will always post better numbers than nodes passing messages across an ocean, and benchmarks run on small networks will always post better numbers than ones run on more nodes. We want to share numbers that are intellectually honest and reflect what Espresso is capable of when real-world applications integrate with the network in production. The Foundation: HotShot Espresso’s finality guarantees come from HotShot, our purpose-built BFT consensus protocol. Rather than use an off-the-shelf implementation like CometBFT, we built HotShot on the academic HotStuff protocol and adapted it to our own needs. Our team of decentralized computing and consensus design experts continue to push its performance to the bleeding edge. That work started from a very different premise than most consensus protocols used by other L1s. For example, Gasper, the consensus protocol underpinning Ethereum, was built so that blocks could continue being produced even when a large number of validators go offline. That flexibility is a big part of why Ethereum takes 16 minutes to finalize transactions. From day one, we designed HotShot for another job: finalizing data from external execution environments as fast as possible, with BFT safety guarantees that keep finalized blocks from being rolled back. The version of HotShot that got us to 1 million TPS and subsecond finality on mainnet is a substantial improvement of that protocol. Our technical deep dives cover the engineering optimizations and consensus changes that got us here. The journey from 6s to <1s When we launched Espresso mainnet in November 2024, it finalized blocks in roughly 6–10 seconds, the product of HotShot’s legacy chained protocol, which needed roughly three full consensus views plus data availability certification before any single block was final. Two major upgrades have since brought that down to under a second: Proof-of-stake upgrade (early 2026), 6 seconds → 3 seconds: Alongside the move to proof of stake, we upgraded HotShot’s core protocol from HotStuff to HotStuff-2, which removes a full round of voting. We also replaced the KZG-based VID scheme used to distribute erasure-coded block data to validators with a faster Merkle-based one. Subsecond finality upgrade (October 2026) 3 seconds → 337 milliseconds: A redesigned version of HotShot that finalizes each block within a single view, plus networking, data availability, and block-building changes. The short version, covered in full in our engineering deep dive, is that we rebuilt consensus, networking, block building, and data availability together instead of treating them as separate systems to optimize one at a time. What subsecond finality and 1M TPS unlock Pairing 1 million TPS of capacity with real-time finality means any chain, centralized exchange, or real-world system can trust the finalized ordering of an Espresso-integrated chain, along with the state that follows from it, the moment a block lands. Crosschain activity starts to feel like it’s happening inside a single execution environment. An order placed on one chain can match and settle against an order on another, and a payment on one chain can trigger an action on a second, with no centralized sequencer to trust and no multi-minute wait on a slow base layer. The payoff arrives first wherever waiting has always cost the most. Institutional finance is moving toward a web of sovereign, purpose-built systems that interoperate on demand, much as banks and exchanges do today, only faster and more programmable. Real-time finality lets each of those systems act on another’s state as soon as a transaction becomes irreversible, and none of them has to surrender governance over its own environment to do it. Built for what’s next The next generation of onchain applications will demand more from a base layer than speed alone, carrying more data in every transaction and proving more about themselves while still finalizing the moment a user acts. Private payments will attach a zero-knowledge proof to every transfer. Crosschain collateral management will let a lender on one chain accept collateral held on another and respond the instant its value shifts. Real-world assets will carry their own compliance data, from investor eligibility to transfer restrictions, everywhere they move. Quantum-safe applications will sign transactions with post-quantum signatures, some more than 30 times the size of the ones in use today. A base layer short on throughput would buckle under that weight, and one with slow finality would undercut the reason to build these applications onchain in the first place. Espresso was designed to supply both at the scale they will require. The upgrade behind these benchmarks is already running on mainnet. If you’re building a chain or evaluating base layers for what comes next, get in touch. Footnotes Highest TPS observed on Solana mainnet Tempo’s TPS number extrapolated from gas limit + average block time of 555ms BNB’s TPS number assumes 70m gas limit per block on mainnet & 21k gas per tx Arc’s TPS number taken from benchmarks run on 20-node testnet Highest TPS observed on Avalanche c-chain Ethereum’s TPS number is historical peak found across L1 + L2 ecosystem on 7 January 2026 About Espresso Systems Espresso Systems builds digital asset infrastructure for institutions and enterprises that demand speed and customizability without sacrificing architectural control. It is lead developer of Espresso, a high-performance blockchain that provides the subsecond finality and massive data throughput making real-time settlement possible at institutional scale. Unlike legacy blockchains that impose their own rules on applications that build on them, Espresso provides a foundation for custom applications that need decentralized security and real-time finality without forcing them to sacrifice speed, connectivity, or architectural control. Companies and blockchains like Celo, ApeChain, Gate, Litecoin Foundation, and others have already chosen Espresso for the architectural control and native interoperability legacy infrastructure cannot match. Founded in 2020 by a team out of Stanford’s cryptography program, Espresso Systems has pushed the state of the art in blockchain, consensus protocol, and distributed network design. It is backed by over $50 million of capital from Andreessen Horowitz, Sequoia Capital, Greylock Partners, Coinbase Ventures, Electric Capital, and more. Website | Docs | GitHub | X | HackMD | LinkedIn | Espresso Foundation(Historical earnings: For 2026Q2 (period ended 2026-06-30), reported basic EPS of 0.42, diluted EPS of 0.39, net income of USD 0.103 billion, and revenue of USD 1.395 billion. Consensus expectations: Consensus for 2026Q3 expects EPS of 0.4 and revenue of USD 0.797 billion.)

10/100·CNeutral+16

n

news10/20news

The Consensus Changes Behind Espresso’s New Performance Benchmarks.
Espresso now handles up to 1 million TPS with subsecond finality

72/100·B+Long