One internet user in 321. That is, at best, the share of the connected population touching web3 on any given day: 18.7 million daily active wallets recorded by DappRadar in the third quarter of 2025, set against the 6 billion internet users counted by the International Telecommunication Union for the same year, or 0.31% (1). The word itself is everywhere. Job ads, billboards, television panels, dinner-table arguments. You have probably nodded along without daring to ask what it actually covers. This guide answers head-on: what web3 is, where its name came from, who coined it, what genuinely runs today and what remains a promise. You will find exact dates, detailed calculations and the sources behind them. Nothing else.
Key takeaways
📋 Index
- Web3 runs apps on blockchains instead of company servers
- The word was coined by Gavin Wood in April 2014
- The same name first meant the semantic web, back in 2006
- A wallet replaces your username and your password
- Only 0.31% of internet users connect on a given day
Web3: a definition that fits into three verbs
Web3 describes a version of the internet where applications execute on blockchains rather than on a company’s servers, and where each person keeps their data and assets inside a wallet whose key only they hold. No platform arbitrates the exchange: the shared ledger settles it. So much for the theory.
But why does such an abstract idea generate so much noise? Because it strikes the sore spot of today’s web: dispossession.
Read, write, own
In the 1990s, you read pages. Then came social networks, blogs and comment threads: you write, you publish, you review — and the platform collects. Web3 grafts a third verb onto the first two: own. Your avatar, your loyalty points, your purchase history would become transferable digital objects, resellable and portable elsewhere.
The ecosystem sums up that progression with a sequence repeated like a mantra: read, read-write, read-write-own. Easy to remember. Far harder to build.
Web1, web2, web3: what actually changes
| Generation | Dominant period | Your role | Who hosts the data | Who can shut you out |
|---|---|---|---|---|
| Web1 | 1990s to early 2000s | Reader | Personal servers, hosting providers | Your host |
| Web2 | Mid-2000s onwards | Contributor | A handful of large platforms | The platform, unilaterally |
| Web3 | Proposed since 2014 | Contributor and owner | Thousands of independent nodes | Nobody, in theory |
Look at that last column. It holds the entire argument.
17 April 2014, the day the word was born
The term came neither from a marketing agency nor from a university lab. It appeared on a personal blog belonging to British computer scientist Gavin Wood, co-founder of Ethereum and author of its Yellow Paper. The post was titled « ĐApps: What Web 3.0 Looks Like ». The date: 17 April 2014.
A text written in Snowden’s shadow
Wood was not sketching a business opportunity. He was sketching a counterattack. His own phrase for it: the « post-Snowden » web. Ten months earlier, the disclosures from the former NSA contractor had exposed the scale of communications surveillance, and Wood drew a blunt conclusion from it: entrusting your information to any organisation rests on a broken model, since the odds of that organisation leaving your data alone come down to the effort required minus the gains expected.
His answer rested on four building blocks: static content publication, dynamic messaging, trustless transactions, one unified interface. The vocabulary has shifted since. The skeleton has not.
Here is the detail rival guides leave out: the network meant to embody that vision did not exist yet. Ethereum’s public Frontier release only went live on 30 July 2015, 469 days after the post (2). Wood named the child more than a year before its birth.
The name had already been taken for eight years
This is where the story turns delicious. « Web 3.0 » was already circulating, and it meant something else entirely. On 12 November 2006, the New York Times ran a piece by John Markoff about entrepreneurs trying to layer meaning on top of the billions of documents online — the semantic web theorised by Tim Berners-Lee and his co-authors in 2001. Markoff called it Web 3.0. Between that article and Wood’s post, 2,713 days went by (3).
Hence a confusion that persists to this day. Two unrelated technologies, one version number. The inventor of the World Wide Web voiced his irritation publicly at the Web Summit in Lisbon, in November 2022.
In fact, Web3 is not the web at all.
Tim Berners-Lee, Web Summit, Lisbon, 4 November 2022
He added that his own decentralisation project, Solid, uses no blockchain whatsoever: too slow, too expensive and too public, in his view, to hold personal data.
Compare the incubation times of the two labels and a pattern jumps out. « Web 2.0 » was coined by Darcy DiNucci in 1999 and described an installed present by the time of O’Reilly’s conference in 2004: five years of lag. « Web3 » was forged in 2014 and still, twelve years later, describes a future that concerns 0.31% of internet users (5). Put differently, web3 has spent roughly twice as long announcing itself as web2 took to take over. A label that still needs explaining to the general public after twelve years says a great deal about the gap between narrative and usage.
Under the hood: how web3 actually runs
Three components cover about 90% of everything calling itself web3. A blockchain that records operations. Autonomous programs that execute on it. A wallet that identifies you.
Your wallet replaces your login
On a web3 application, there is no sign-up form. You connect a wallet — a cryptographic key pair — and that address becomes your account, your signature and your safe. The operations you trigger follow the rules of smart contracts, programs published on-chain that execute without human intervention as soon as their conditions are met.
The saying in this world runs to five words: not your keys, not your coins. Lose the recovery phrase and nobody emails it back to you. No customer support sits behind a smart contract.
Why almost nothing is written to the chain
Here is the calculation nobody performs, and it explains most of the disappointments. Writing one empty 32-byte slot into Ethereum’s storage costs 20,000 units of gas. One gigabyte amounts to 33,554,432 slots, therefore 671 billion units of gas. A block, meanwhile, has been capped at 60 million units since November 2025, and a block arrives every twelve seconds: storing a single gigabyte would monopolise the entire network for 37 hours (4).
Thirty-seven hours for one high-resolution holiday photo. So web3 stores almost nothing: it records proofs, balances and addresses, then points to distributed file systems such as IPFS or, very often, to an entirely ordinary server.
Most buyers believe an NFT contains the artwork. It contains only a pointer to it. The cause lies in the calculation above: the image would cost a fortune in network fees, so it lives elsewhere and the token merely references that location. The concrete consequence was documented by cryptographer Moxie Marlinspike in January 2022: once the OpenSea marketplace removed his token, it also vanished from his wallet display. Before any purchase, check where the file truly sits and who controls that place.
What is web3 used for in 2026?
A fair question, and the answer falls into three families of uses that genuinely run.
- Decentralised finance: lending, borrowing and swapping with no bank or broker, through public, auditable contracts.
- Non-fungible tokens: tamper-proof ticketing, digital title deeds, resellable in-game items, certificates of authenticity.
- Decentralised autonomous organisations: collectives whose voting rules and treasury live in code rather than in filed articles of association.
Finance, collectibles, organisations
Gaming now dominates measured activity, ahead of non-fungible tokens and decentralised finance. A sector born of speculation has therefore reorganised itself around entertainment. Nobody saw that coming in 2021.
Since 30 December 2024, a European legal framework
Web3 no longer operates in a legal vacuum. The European MiCA regulation has applied to crypto-asset service providers since 30 December 2024, and France closed on 1 July 2026 the transition period that let firms registered under its national regime keep trading. Since that date, only a European authorisation permits offering such services. Check that authorisation before handing over a single euro: the AMF publishes the list of approved providers.
One test separates a real web3 product from a repainted web2 one. Look in the settings for the option to export your private key or recovery phrase. If it exists, you can walk away tomorrow and recover your holdings elsewhere with a different wallet. If it does not, the company holds your keys: you are using a conventional service dressed in new vocabulary, carrying the same account-freeze risk as anywhere else.
Do the criticisms hold up?
They rarely come from novices, which gives them weight. Three keep resurfacing.
December 2021, first salvo. Jack Dorsey, co-founder of Twitter, wrote that users do not own web3 and that venture funds and their limited partners do. His point concerned the initial distribution of tokens, frequently concentrated among early investors.
January 2022, second salvo, far more technical. Moxie Marlinspike, creator of the Signal messenger, built two applications to inspect the machinery himself. His finding: nearly every decentralised application queries the chain through two companies, Infura or Alchemy, and even a wallet like MetaMask routes through that relay. Two firms to read a ledger supposedly belonging to no one: decentralisation stops at the last mile.
| The 2014 promise | What is observed |
|---|---|
| Everyone hosts their share of the network | Applications lean heavily on a few node-access providers |
| Genuine ownership of content | The token points to a file hosted elsewhere |
| A mainstream internet | 0.31% of internet users active on a given day |
| No central authority | European authorisation mandatory for providers |
Should the subject be buried, then? No, and technical history urges caution: the HTTP protocol took fifteen years to become obvious. The building blocks of web3 work, they settle transactions worth tens of billions of dollars, and their limits stem from the physics of the system rather than from fraud. The gap between the talk and the usage nevertheless remains enormous.
Frequently asked questions about web3
Who invented web3?
Gavin Wood, co-founder and first chief technology officer of Ethereum, in a post published on 17 April 2014. He went on to found the Web3 Foundation in 2017 in Zug, Switzerland, then the Polkadot network.
What is the difference between web3 and Web 3.0?
Web 3.0 originally meant the semantic web: a network whose machines grasp the meaning of content. Web3 means the web built on blockchains. Same number, separate projects, permanent confusion.
Do you need cryptocurrency to use web3?
Yes, as soon as you write to the chain, since every operation consumes network fees paid in the native currency. Reading stays free. Signing costs.
Will web3 replace web2?
No usage data currently points that way. The scenario most consistent with the figures is coexistence, where a few functions — payments, identity, ticketing, title deeds — migrate to shared ledgers while the rest of the web keeps running on ordinary servers.
Sources and statistics
(1) The adoption ratio combines two independent publications: 6 billion internet users in 2025 according to the Facts and Figures 2025 report from the International Telecommunication Union, released on 17 November 2025, and an average of 18.7 million daily active wallets in the third quarter of 2025 according to DappRadar’s State of the Dapp Industry report. 6,000,000,000 ÷ 18,700,000 = 320.9, that is one active wallet per 321 internet users, or 0.31%. DappRadar states that one person may hold several wallets, so the real share of human users sits below this figure, which acts as an upper bound.
(2) The gap between the name and the thing comes from two public dates: Gavin Wood’s post « ĐApps: What Web 3.0 Looks Like », dated 17 April 2014 on gavwood.com, and the launch of Ethereum’s Frontier release on 30 July 2015. That makes 469 days, or fifteen and a half months, between the appearance of the term and the activation of the network meant to embody it.
(3) The vocabulary collision is measured between John Markoff’s article in the New York Times of 12 November 2006, which applies the expression Web 3.0 to the semantic web, and Gavin Wood’s post of 17 April 2014: 2,713 days, or seven years and five months during which the label pointed to an entirely different project.
(4) The storage cost is derived from the protocol’s public parameters. Writing to an empty 32-byte slot costs 20,000 units of gas (the SSTORE_SET parameter in Ethereum clients, from EIP-2200). One gigabyte equals 1,073,741,824 bytes, hence 33,554,432 slots, hence 671,088,640,000 units of gas. The per-block ceiling has stood at 60 million units since November 2025, a figure confirmed by ethereum.org documentation in 2026, and a block is produced every twelve seconds: 671,088,640,000 ÷ 60,000,000 = 11,185 blocks, that is 134,218 seconds or 37.3 hours of full network occupation. This calculation explains in physical terms why bulky content is hosted off-chain.
(5) The comparison of incubation times draws on two documented genealogies: the expression « Web 2.0 » appears with Darcy DiNucci in 1999 and becomes popular at the first Web 2.0 Conference run by O’Reilly Media and Dale Dougherty in 2004, a span of five years, whereas « Web3 » dates from April 2014 and therefore stands at twelve years in September 2026. Set against the usage rate calculated in point (1), that span offers a simple measure of the distance between the fame of a term and its actual adoption.
