How Sui Is Building the Infrastructure the Internet Was Always Missing

How Sui Is Building the Infrastructure the Internet Was Always Missing

Elijah is a 26-year-old call center agent in Nigeria. Last Tuesday he downloaded a mobile game, signed in with Google, earned a rare in-game sword, sold it to a stranger in another country for $12, and withdrew the money to his bank account. From download to cash-out: forty minutes. He has no idea what Sui is. He used it anyway.

That invisibility blockchain so seamlessly embedded in experience that users never encounter it is the goal Sui was built for. It is also the goal every blockchain project has claimed since 2017 while delivering approximately nothing of the sort. What makes Sui different is not a better pitch. It is a fundamentally different technical foundation, built by a team that had already solved most of the hard subproblems before they wrote a single line of Sui code.

The Problem Was Always Architectural

Mainstream blockchain adoption has failed for four reasons that have nothing to do with regulation, education, or marketing. They are architectural failures baked into the design of first-generation networks.

Ethereum processes 15–30 transactions per second less than a 1990s database. During high demand, gas fees have exceeded $200 for a single swap. New users must generate and permanently secure a cryptographic seed phrase before they can interact with anything. And virtually every ‘decentralized’ application stores its actual data images, game assets, video on Amazon Web Services, because on-chain storage costs make it economically impossible to do otherwise. The NFT you ‘own’ is usually a receipt pointing to a file on a server that could disappear tomorrow.

These are not problems awaiting clever workarounds. They are consequences of building infrastructure for something other than what people actually need. Solving them required starting over.

The Foundation: Five Engineers From Meta’s Failed Blockchain

In January 2022, Meta shut down Diem, its blockchain project that had spent three years attempting to serve Facebook’s two billion users. Five of its core engineers Evan Cheng, Sam Blackshear, George Danezis, Kostas Chalkias, and Adeniyi Abiodun incorporated Mysten Labs within months and began building what Diem was supposed to be, without the corporate constraints.

They named it Sui, from the Japanese word for water: fluid, adaptive, taking the shape of its container. The name is the design philosophy. A blockchain that conforms to users rather than demanding users conform to it.

What followed was not iteration on existing blockchain design. It was a reconstruction from the data structure up.

The Technology: What Sui Actually Changed

Objects Over Accounts

Every major blockchain before Sui organises state as a global accounts ledger. Every transaction, regardless of whether it conflicts with any other, waits in a single sequential queue to update that ledger. Picture ten thousand library patrons waiting in one line for one checkout desk even though they want books from completely different sections. The queue is the bottleneck, not the books.

Sui replaces the global ledger with objects. Every asset every coin, NFT, game item, DeFi position is a distinct object with a unique identifier, explicit ownership metadata, and a version number. Because ownership is tracked at the object level, the network can mathematically determine which transactions have zero dependencies. Alice sending Bob an NFT she owns, while Carol simultaneously sends Dave a different NFT she owns, involves no shared state. Both transactions execute in parallel across separate validator threads with no coordination required.

This is not an optimization. It is a different computational paradigm. The result: 297,000+ transactions per second in testing, versus Ethereum’s 15–30. Not because Sui’s hardware is faster. Because its architecture enables work that account-based blockchains cannot parallelize.

297,000+ TPS peak throughput vs. Ethereum’s 15–30 on equivalent hardware

Two Speeds, One Network

The object model enables Sui’s most practically significant feature: a two-speed transaction processing system. When you transfer an NFT you exclusively own, that transaction requires no traditional consensus — no one else can send that asset simultaneously, so there is no conflict to resolve. Sui routes these transactions through Byzantine Consistent Broadcast: verify signature, confirm ownership, collect quorum acknowledgments. Finality in approximately 250 milliseconds.

When transactions touch shared state competing bids on an auction, concurrent trades in a liquidity pool they go through full Byzantine Fault Tolerant consensus via Mysticeti, a protocol published and peer-reviewed at the 2025 NDSS Symposium. Mysticeti eliminates block certification overhead by running consensus directly on uncertified DAG blocks, reaching the theoretical minimum of three communication rounds. When deployed to mainnet in August 2024, median consensus latency dropped from 1,900 milliseconds to 390ms an 80% reduction validated on a live network with real economic value.

390ms median consensus latency after Mysticeti down from 1,900ms at launch

Invisible Wallets: zkLogin and Sponsored Transactions

Speed and throughput solve the back-end. The front-end required two separate innovations. zkLogin allows users to authenticate with their Google, Apple, or Twitch account and receive a Sui wallet without ever seeing a seed phrase or private key. A zero-knowledge proof cryptographically binds the ephemeral session key to the OAuth authentication without revealing the user’s identity to the blockchain or their blockchain activity to the OAuth provider. It is genuinely self-custodial Google cannot sign transactions on the user’s behalf and it is invisible to the user.

Sponsored transactions allow any third party a game developer, a protocol, a foundation to pay gas fees on a user’s behalf. Combined with zkLogin, the result is what Miguel experienced: sign in with Google, interact with a blockchain application, never see a gas fee, never encounter the word ‘wallet.’ This is not a UX layer on top of existing blockchain complexity. It is native protocol infrastructure that makes the complexity genuinely disappear.

Move, Walrus, and DeepBook

Move, the programming language Blackshear designed at Diem and refined for Sui, treats digital assets as linear types resources that cannot be copied or implicitly discarded. A Move program attempting to duplicate a token will not compile. Entire classes of vulnerabilities that have cost the DeFi ecosystem billions reentrancy attacks, integer overflows, unauthorized state modifications are caught at the compiler level rather than discovered by auditors or exploited by attackers.

Walrus, launched in March 2025 with $140 million in institutional backing, addresses the storage problem directly. Using erasure coding, Walrus splits files across distributed storage nodes so that each stores only a fraction but any sufficient subset can reconstruct the whole. Storage costs drop to economic viability. Crucially, Walrus data is programmable via Sui smart contracts not just stored, but conditionally accessible, auditable, and managed on-chain. Applications can store actual user data, not just hashes pointing to AWS. DeepBook, Sui’s native on-chain Central Limit Order Book, brings institutional trading infrastructure limit orders, price-time priority matching, shared liquidity to DeFi, crossing $1 billion in cumulative trading volume by late 2024.

The Ecosystem: Real Numbers, Real Tensions

By October 2025, Total Value Locked in Sui DeFi hit $2.6 billion, driven by Suilend, NAVI, Cetus, and Bluefin. Monthly stablecoin transfers exceeded $70 billion. Daily active users grew 81 times during 2024. Over 65 gaming studios committed to 2025 releases on Sui. Bitwise, Canary Capital, and 21Shares filed spot Sui ETF applications with the SEC in December 2025.

Those headline numbers require honest context. By February 2026, TVL had pulled back to approximately $561 million a 78% decline from the October peak. Part of this reflects market conditions. Part reflects genuine damage from the Cetus exploit of May 2025, in which approximately $223 million was drained in 15 minutes through a rounding bug in a third-party math library. Sui validators coordinated to freeze $162 million in stolen assets. The response was fast and effective. But the ability of 114 validators to collectively freeze specific wallets achievable via group message, not a hard fork is a legitimate decentralization question that the Cetus incident made concrete. A network with 800,000 validators cannot do that. Whether that capability is a feature or a vulnerability depends entirely on what you believe blockchains are for.

“P50 consensus latency on Sui is just under 400 milliseconds. For derivatives, every millisecond is potential slippage, potential alpha lost to faster traders.” Rabeel Jawaid, Co-founder, Bluefin

The Vision: Infrastructure, Not Speculation

What Mysten Labs is building is most clearly understood as a complete technology stack for a new layer of the internet. Execution and consensus via Sui mainnet. Decentralized storage via Walrus. Invisible identity via zkLogin and SEAL. Institutional finance via DeepBook. Each layer is valuable independently. Together they describe something that has never existed: an infrastructure where ownership, programmable money, and verifiable identity are native properties of every application, invisible to the users who benefit from them.

The Miguel scenario is not theoretical. The primitives exist today. Any developer can build that application on Sui right now. What remains is the grinding work of ecosystem development: enough studios building enough applications reaching enough users that blockchain becomes infrastructure rather than product category. That is what water does. It fills the shape of whatever contains it. It does not announce itself.