Exponential Software Requires Exponential Hardware

Hardware is constantly evolving. Bitcoin (BTC) mining only needed Raspberry Pis in the very early days. Ethereum (ETH) could be mined with an average consumer laptop GPU. Now, validators on newer blockchains are required to have high-speed servers, gigabit internet connectivity and custom-built high performance hardware. The growth and demand of the hardware industry has never been greater.

This post is part of Mining Week. Sam Cassatt is the CEO and Founder of Aligned, a Web 3 infrastructure firm.

As blockchains continue to evolve and new innovations are created, so too will the demand for optimized computing – and further, the need for custom hardware. Just as Web 2 created the modern design of data centers, new Web 3 primitives will require the creation of a new structure of hardware, custom-built for the rapidly growing needs of the software.

Web 1 was almost exclusively reading and writing simple http content. Fairly simple needs limited by our fairly simple hardware stack. The computers were bulky and slow, the internet was at broadband (or even dial-up) speeds, servers were bigger and more expensive to host, etc.

Then came Web 2 – the era of social media and streaming. As the technology grew, so did the demands of consumers. The new shifts in workload motivated the design of the Web 2 data center. Facebook (FB) needed to process massive real-time video streams that informed the architecture of their data centers. Apple (AAPL) created the M1 chip for users to maximize the Web 2-centric experience. AWS’s entire business is based on catering to the needs of Web 2 – closed servers for gated networks.

While Satoshi began mining bitcoin on commodity CPUs, the most efficient hashing hardware eventually became GPUs. Fast forward a few years, and bitcoin is mined at an institutional level on large networks of ASICs, custom computer chips designed specifically for the purpose of mining bitcoin and nothing else. This was the progression of the first Web 3 workload, born in the era of social media.

Mining is the most simplistic version of this hardware growth – the first chapter. As more and more networks continue to use proof-of-stake consensus models and as total value locked (TVL) continues to grow in decentralized finance (DeFi), the stakes get bigger, the demand for computation increases, creating the demand for better hardware and infrastructure.

Currently, hashing is the main math that is deployed in blockchain architecture. Eventually, over time, our demands for blockspace will exceed the computational limits of networks based on hashing.

Zero-knowledge proofs are emerging as the next progression in the mathematical underpinnings of blockchain scaling. Zero-knowledge (ZK) rollups such as ZKSync, Polygon Hermez and many more are starting to allow scalability above layer 1, or base, chains. This type of advancement in Web 3 architecture will be necessary to handle the amount of on-chain activity to simultaneously hold massive industries such as DeFi, gaming, decentralized autonomous organization (DAO) governance, non-fungible tokens (NFT) and more.

The problem is that our traditional hardware is not built for zero-knowledge proofs. The general purpose processors of today are not designed for this type of computation, although they can do it however inefficiently. Over time, we will likely see the emergence of specialized hardware for this purpose, just as we eventually saw ASICs emerge for bitcoin mining once the computation was understood and solidified.

Similarly, miner extractable value (MEV) will ultimately be an optimization problem. Vitalik Buterin’s "Endgame" explores some of the future directions we are going, especially as block proposers and builders become separate. As this happens the amount of MEV per block will be optimized, likely creating a separate fee market for well-constructed blocks. Again, this will likely lead to hardware optimized solutions, seeking to optimize as far as we can go.

Everyone has slightly different imaginations of what Web 3 will ultimately look like. Regardless of your specific vision, if you believe in the Web 3 thesis of decentralization, open-source, permissionless access to the internet, you believe that eventually hardware workloads will get optimized. Right now, it is definitely not optimized, but as our demands continue to grow, technology continues to mature and so will the hardware that the next generation of the internet will be built upon.