Why Scale When You Can Centralize?

Only a few years ago the big issue in blockchain design was scaling. Everyone was trying to find blockchain schemes that could, over time, grow to handle higher and higher volumes of transactions. The reasons may have varied – some focused on reducing transaction costs while others wanted to build ever-more-complex products with higher on-chain compute needs or maybe they wanted to bring the entire world on-chain – but there was a shared goal to do more and do it faster.

Simultaneously there were teams working on all kinds of different approaches to the scaling problem. Ethereum, for years, focused on "sharding." Solana claimed to solve the scaling problem with a new consensus design. Myriad projects like Arbitrum, Optimism and StarkNet sprang up with new ideas. But, looking back on all of that a few years later, we do not see much success in scaling. Bitcoin launched in 2009 with about 7 transactions per second. And computers roughly double in performance every 18 months .

So 15 years later computer improvements alone should give a 1,000x speed-up. And, indeed, 7,000 transactions per second is considered good performance for a blockchain in 2025! There is no great engineering achievement in doubling your blocksize every 18 months in line with performance improvements granted you by the semiconductor industry. And, notably, nobody has demonstrated a truly "scalable" architecture that engineers around the hard bottlenecks.

Today the web3 economy experiences bottlenecks all the time. A few years ago the problem was congestion during NFT mints. Now it is more likely to be fees spiking when users compete to buy a new memecoin off a pump.fun bonding curve. Why is this still a problem? Why is there far less discussion about scaling today if, still, it remains unsolved? Why do we see clearly centralized products like Hyperliquid with a closed validator set and base with a monopoly operator pushing the limits on throughput (but still, notably, not outrunning Moore's Law)? Why do so many projects still have security councils with ultimate authority over supposedly self-custodied assets? Why does it look like centralization – the thing web3 exists to eliminate – is somehow sitting there underneath all the working high-volume popular products?

We are going to argue three things, in order, that explain what is happening. The first claim is that the ideal scaling people were talking about a few years is not possible – and those same people simply changed their goals after accepting the original was not possible. The second claim is that the industry as a whole pulled some sleight-of-hand and switched over to a simpler and more achievable version of scaling without clearly disclosing this in a way non-technical users could plausibly decipher in real time. Finally, the last claim is that we see centralization popping up in surprising places because centralization is the only solution the industry has ever found for the harder version of scaling. And, since admitting that centralization presents problems, seemingly-unnecessary technical complexity is used to obfuscate this reality.