Web3’s Centralization Problem

A Brief History of the Internet

The origins of the internet can be traced back to the early 1960’s when Cold War nuclear tensions were at their highest. The US realized it needed a communications system that could survive a Soviet nuclear attack and began developing a wide area computer network to serve this purpose. This development culminated in 1969 when the US Department of Defense launched the Advanced Research Projects Agency Network (ARPANET) as a way for government officials, researchers, and military personnel to share data. ARPANET proved to be a massive success and kicked off a wave of network building throughout the 1970’s and 1980’s to meet the rising demand for automated information-sharing between public entities. 

In 1989 the British computer scientist Tim Berners-Lee conceptualized an easier to use internet that didn’t need to exist for solely governmental purposes. Berners-Lee submitted his ideas as a proposal to the European Organization for Nuclear Research (CERN) and began developing his project that he had dubbed the ‘World Wide Web’. After privately growing the network throughout in the early 1990’s, the World Wide Web was opened up to the public domain to be used for communication and commercial purposes in 1993.

Web1 was the first stage of this commercial internet and took place during the 1990s and early 2000’s. It assumed that internet users would be both creating and using content and web infrastructure. Anyone was free to build on the internet and technically experienced content creators spun up their own webpages that catered to small, fragmented audiences. However running your own servers is not a convenient practice for most users, and because of this very few hosted infrastructure. Web1 was open source, decentralized, slow – and was populated by isolated, static websites that posted read-only content for their users. This first version of the public internet was limited to a smaller audience of technically proficient users.

Despite this, entrepreneurs who recognized the potential of the internet flocked to create companies that could utilize it, while investors lined up to provide them funding. This frenzy created what has now been dubbed the ‘.com bubble’, which began in 1995 and created explosive stock growth on the NASDAQ until its spectacular collapse in 2001. Although nearly all of the publicly traded internet companies created during the bubble failed, the market cycle drew a lot of attention to the internet and its potential. Usage of the World Wide Web was climbing at a dizzying pace.

Rapid growth, combined with lessons learned during the ‘.com bubble’, created the need for usable, interoperable web platforms that treated users as a commodity. This need drove the creation of Web2 throughout the 2000’s. Web2 did not fundamentally change any of the technical underpinnings of Web1, websites still needed to be hosted on servers and have their own specific web addresses, instead it transformed the functionalities that webpages possessed. Interaction and collaboration between individuals was enabled by social media dialogues on websites like YouTube and Facebook – users could create their own platform accounts, talk to one another, edit/reshare content, and embed content from other websites. This lowered the barriers to entry for content creation and created an influx of user generated material that could be monetized. Individuals could now easily access the web through internet service providers (ISPs) and did not feel the need to set up their own websites and servers. 

In the private sector, large tech companies were eager to own and influence user data, while ISPs further consolidated into a few large entities. Encouraged by aggressive VC investment, companies raced to capitalize on user's excitement to generate free content. Social media sites began to employ algorithmically curated content to silo users into their shared interests and increase their engagement, while concurrently censoring certain types of user generated content. This planned nudging combined with the rise of web based advertising technology turned users into products for platforms through the monetization of their personal data and attention. In the public sector, high profile whistleblowers like Edward Snowden illuminated the pervasiveness of government surveillance across the internet. This ease of use, combined with the increasingly platform based nature of the internet, eliminated the decentralized ethos of Web1. It became clear that the benefits of Web2 did not come without tradeoffs, a better system needed to be designed.

The Rise of Web3

In 2009, in the middle of Web2’s meteoric rise, the Bitcoin network was launched as a peer-to-peer payment system secured and run by a decentralized network of unrelated, user-operated computers connected to the internet. This innovation created a new financial architecture for conducting commerce that did not involve centralized intermediaries. Bitcoin proved that a decentralized ledger could successfully transfer value between two distant, untrusting parties. The Bitcoin network laid a foundation for the development of subsequent public blockchains throughout the 2010’s, many of which had new uses and features. The most significant of these was Ethereum, with its inception in 2014 and initial coin offering (ICO) in 2015 serving as the founding moments of Web3.

Gavin Wood, a co-founder of Ethereum, first coined the term Web3 in 2014 and described it as a system in which “...all interactions will be carried out pseudonymously, securely and for many services, trustlessly. Those that require third parties … will give the users and DApp-developers the ability to spread the trust among multiple different, possibly competing, entities, massively reducing the amount of trust one must place in the hands of any given single entity.” This type of collaboration was first showcased within the Ethereum network.

The creation of Ethereum represented a different way of thinking about cryptocurrencies. It is similar to Bitcoin in that it reaches decentralized consensus, but different in that it is designed to be programmable, whereas Bitcoin was mainly designed to serve as a payment system and store of value with limited programmability. Ethereum’s peer-to-peer network securely executes and verifies user created application code, called smart contracts, using its network effect of decentralized consensus as an ‘all encompassing social contract’. Developers can program nearly any kind of smart contract that they want within Ethereum’s permissionless, composable, and turing complete environment. Ethereum’s native token, Ether, is the currency used during transactions on the network and enables a decentralized marketplace for computing power and digital resources.

Ethereum was a success and sparked a wave of development to create open source, programmable networks, with much of this innovation coming from Ethereum’s founding team. Two of its founders, Gavin Wood and Charles Hoskinson, started the rival blockchains Polkadot and Cardano, respectively. Joseph Lubin went on to found Consensys, a company dedicated to harnessing the power of Ethereum and building the infrastructure for Web3. Of the project’s 8 founding members, only one, Vitalik Buterin, is still actively working on the blockchain platform. 

Outside of Ethereum’s core team, thousands of developers entered the Web3 space and began creating their own blockchains and DApps. Decentralized financial applications ushered in a form of user controlled finance (DeFi) that democratized forms of yield previously only available to accredited investors. Non fungible tokens (NFTs) trailblazed digital forms of ownership and established a lively creators economy. Subsequent layer one networks were developed and began competing against one another while simultaneously establishing their own unique ecosystems. 

Web3 aims to decentralize the internet back to its original Web1 foundations while maintaining the interoperability and throughput of Web2. It does this by distributing the hardware, software, and monetary systems currently housed within and supported by centralized entities, whether public or private, to decentralized networks of unrelated, user-operated computers. This structure removes intermediaries and allows users to own the content that they create and the data that they generate while online. Peers transact directly with each other using tokens, which essentially give users digital property rights to the blockchains they are engaging with, allowing them to own a piece of the internet. In order for this system to function as designed, it needs to remain decentralized and censorship resistant. Unfortunately, Web3 is already showing signs of consolidation.

Risks of Centralization

Public blockchains decentralize information throughout a broad network of servers that store and compare permanent ledgers to achieve network consensus. Individuals that want to access or build on top of the blockchain need to do so through an operational node, whether it is self hosted or hosted by a third party. Similar to the dynamics of Web1, running your own node, particularly in a production environment, is difficult, capital intensive, and not a feasible option for the majority of Web3 users. As a result, giant cloud providers like Amazon Web Services (AWS), Google Cloud, and Azure have centralized the node provisioning within Web3. The founding network of Web3, Ethereum, is an example of this. The Ethereum network has a total of 8,182 nodes, of which 3,029 (37%) are hosted by AWS. Obviously, this is antithetical to decentralization. Nearly all of the popular layer 1 networks created after Ethereum show similar overreliance on cloud providers.

Infrastructure centralization of this degree magnifies the risk of power outages. AWS experienced a large-scale outage in December of 2021 that halted several centralized exchanges like Coinbase and Binance. However this outage also took down dYdX, a supposedly decentralized exchange (DEX). DEXs are decentralized and allow peer-to-peer transactions, yet dYdX was completely reliant on a centralized cloud provider. AWS has been prone to outages in the past, and applications centralized on their servers will likely continue to face issues.

Blockchain-based apps rely heavily on application programming interfaces (APIs), which allow applications to communicate with one another. Decentralized applications (DApps) use APIs to connect to the blockchain directly instead of running their own node, a significantly more economical route for developers. ConsenSys, the parent company of both Infura and Metamask, has capitalized on this and expanded rapidly by bringing users onto their centrally managed APIs and wallets, trading the redundancy of a distributed system for the ease and convenience of a centrally hosted product. This consolidation is reminiscent of Web2, and once again brings forth a host of technical risks.

In November of 2020, Infura caused a dramatic Ethereum network outage when they did not update their nodes to the latest version of Ethereum’s client software. Ethereum’s core developers had added new code to fix a consensus bug, and Infura didn’t notice in time. This sync issue caused Infura’s services to crash, which forced crypto exchanges and wallets dependent on Infura’s API, like Binance and MetaMask, to temporarily suspend Ethereum and ERC20 token withdrawals because they couldn’t get complete transactions or get accurate price data. This was a sobering experience that illustrates the problems with centralizing a system that is built to operate as a distributed system, no single bug should cause such a major failure.

Beyond technical risks, centralized infrastructure strips away Web3’s censorship resistance and allows governments and private entities to easily influence the sector. Globally, organizations like FATF are creating anti-money laundering (AML) and counter terrorism financing (CTF) regulatory frameworks that are increasing in their adoption. Within the US, government institutions like the SEC are adopting an increasingly hardline stance against cryptocurrency. If Web3 is to deliver on its promise of censorship resistance in the face of increased governmental scrutiny, then its infrastructure needs to be sufficiently decentralized and antifragile.

The Tornado Cash ban is an example of how the government can impact a centralized Web3. Tornado Cash is a smart contract operating on the Ethereum blockchain that acts as a mixer – users send funds to the smart contract which then mixes their funds with the funds of others in the contract before withdrawing them to the recipient. This greatly obfuscates the origins and destinations of cryptocurrency transactions. On August 8th, 2022 the US Department of Treasury sanctioned Tornado Cash, citing that the service had been used to launder billions of dollars from cyber crimes and was frequently used by Democratic People’s Republic of Korea (DPRK), and placed 44 USDC and Ether addresses connected to the mixer on its list of Specially Designated Nationals (SDNs). 

Immediately after this announcement Circle, the issuer of USDC, froze the USDC in the wallets of those 44 addresses. Also immediately after the announcement, Tornado Cash’s GitHub, email, and website were taken down. Tornado Cash’s GitHub simply contained the open source code for the development of the smart contract, and under the current legal interpretation set by Bernstein v. Department of Justice, open source code is considered free speech. Clearly, this established a worrying precedent. 

On August 9th, 2022, Alchemy and Infura disabled access to their Ethereum API on the Tornado Cash front-end user interface. On August 10th, a suspected developer of Tornado Cash was arrested in the Netherlands, with authorities highlighting that ‘multiple arrests’ were likely. Two days later on August 12th the Pocket Network, a decentralized infrastructure provider that claims to be the most censorship-resistant, reliable, and performant RPC infrastructure provider in the world, banned interactions with the sanctioned addresses. The Tornado Cash ban illuminates how centralized services with geographically registered entities are left completely open to nation state level attacks and regulation because they have to obey the legal directives of their respective sovereign. 

This type of geoblocking was again on display in March, 2021 when MetaMask banned all wallets associated with Iranian IP addresses in order to comply with the sanctions list supplied by the US Treasury Department’s Office of Foreign Asset Control (OFAC). In this same episode, MetaMask mistakenly banned users in Venezuela for several hours before correcting the error. As sanctions lists change users’ funds in certain areas can be rendered unusable, an existential scenario.

Private sector entities can also serve as vectors for censorship. In January of 2021, AWS stopped providing services to Parler, an openly right wing social media platform, because the content being shared on its message boards violated Amazon’s terms of service. It's possible that one day certain decentralized crypto projects and applications will violate these same terms of service for a more vanilla regulatory reason, resulting in their termination from AWS’s servers. Other large cloud providers like Google and Azure are equally liable to take similar courses of action. As regulations tighten around crypto globally, these providers will increasingly be pitted against their Web3 customers.

Final Thoughts

An overreliance on centralized providers runs completely counter to the intended ethos of Web3. If Web3 wants to deliver on its promise of creating a decentralized internet that is user controlled, censorship resistant, and equitable for all of its users, then it needs to avoid the same pitfalls that characterized the shift from Web1 to Web2. Efforts should be made to decentralize infrastructure provisioning and promote the development of alternative hosting solutions. More service providers should be open source and allow everyone to participate in an autonomous development and decision-making process. If Web3 continues to rely on legacy service providers and Web2 business models, then it will simply repackage the very problems it was created to end.

‍