CoinDesk columnist Nic Carter is partner at Castle Island Ventures, a public blockchain-focused venture fund based in Cambridge, Mass. He is also the cofounder of Coin Metrics, a blockchain analytics startup.

What was once an idle supposition is now concrete. Public blockchains are destined to privilege the largest, most fee-tolerant transactions, at the expense of non-financial uses. 

When a robust block space market emerged on Bitcoin in 2017, some wrote it off as an aberration, believing the future was #FeeLess. Since then, Bitcoin’s bite-the-bullet approach to fees has taken hold: low-fee chains have suffered from bloat and irrelevance, and the second-most valuable blockchain, Ethereum, has effectively embraced the reality of meaningful fees. This heralds a shift in how major blockchains are perceived, moving away from generic computation layers and towards their destiny as financial infrastructure.

The logic for this shift is simple. Satoshi included the combination of fees and the blocksize cap in Bitcoin as both an anti-spam mechanism (to prevent the injection of arbitrary amounts of data that would make the chain impossible to validate) and as a method to compensate miners in the long term. Satoshi envisioned a future where fees alone would support miners, after the subsidy had run out. 

See also: Nic Carter - How Blockchains Become Great Big Garbage Patches for Data

Today, that doctrine is largely unchanged; Bitcoiners still expect fees to eventually grow to 100 percent of miner revenue. (Bitcoin miners currently make 9.7% of their income from fees, according to Coin Metrics.) Capped block space is critical to make this work. In a finite system, transactors are willing to pay up for inclusion in a block. In uncapped alternatives, fees are effectively zero – and one can imagine that these chains will be forced to rely on perpetual inflation to finance security, or fall back to permissioned validators.

Aside from paying for security and warding off perpetual inflation, fees have additional emergent impacts. Effectively, they force transactors to think hard about what they are using the blockchain for. This encourages higher-value transactions and discourages frivolous use cases. In fee-bearing blockchains like Bitcoin, marginal, spammy, or non-monetary usage simply gets priced out over time. 

As Bitcoin Core developer Greg Maxwell says, there is infinite demand for highly-available perpetual data storage. As a result, low-fee alternatives become great big garbage patches. If you imagine fees as attaching a weight to transactions, you can see how fee-bearing transactions would force out more marginal ones from the auction for block space, like a lead weight displacing water from a bucket.

One great example of this displacement is Veriblock. Veriblock is a protocol which bids a fixed amount of its token, VBK, for Bitcoin block space. At its peak Veriblock transactions accounted for more than 30% of Bitcoin transactions. But, as Bitcoin fees perked up in May 2020, and VBK fell in value (and hence so did the amount of BTC it was bidding for block space), Veriblock transactions were squeezed out. Ordinary Bitcoin transactors ultimately outbid the more fee-sensitive use case.

Consequently, many Bitcoiners believe that, in the long term, the base layer will come to resemble Fedwire or CHIPS, large-scale settlement networks with large average transaction sizes. This has been the working assumption among developers for a long time on Bitcoin, and it’s part of the reason why Bitcoiners speak derisively of SatoshiDice and coffee payments: They don’t expect these would be suitable for base-layer transactions at maturity. You don’t send a wire transfer to pay for a stick of gum; payment methods are a function of your convenience and settlement needs.

Mirroring Bitcoin’s high fee epoch in 2017, Ethereum has witnessed in 2020 the emergence of a healthy block space market. Driven by the popularity of stablecoins (most of which rely on Ethereum as the underlying infrastructure) and the rapid growth of DeFi, Ethereum fees have skyrocketed this year, peaking at 60% of miner revenue. Transactors on Ethereum paid $8.6m in total fees on Aug. 13, with per-transaction costs coming in at a median of $3.60.

Vitalik Buterin once stated in reference to Bitcoin that “the internet of money should not cost 5 cents a transaction.” It’s safe to say his attitude to fees, and that of Ethereans more broadly, has moderated with time. Delays in ETH 2.0, a growing understanding that unlimited block space has negative externalities, and a newfound appreciation for fees as the backstop of a potentially deflationary force have caused many Ethereans to embrace a higher-fee world.

The explosion of new liquidity mining opportunities on DeFi and the continued growth of crypto-dollars have priced out other use cases on the platform. The cost to deploy complex contracts has skyrocketed to hundreds of dollars in some cases. Today, a user sending a multi-million-dollar Tether transaction will most likely outbid someone deploying an Aragon DAO or minting an NFT. 

See also: What Is Yield Farming? The Rocket Fuel of DeFi, Explained

This is sobering for some Ethereans, as it has punctured some of the more expansive visions of what Ethereum could become – at least in its present form. With high fees, the most economically dense transactions come to occupy block space to the exclusion of all else.

Ever since the departure of the big blockers, Bitcoiners have made their peace with this, prioritizing a layered approach in which the base layer is reserved for larger settlements. Bitcoin scales not by increasing the available supply of block space, but by minimizing the quantity of data registered on chain. Lightning scaling amortizes potentially thousands of transfers into a handful of on chain transactions. 

Meanwhile, physical bearer instruments like Opendimes are funded once but can be passed around arbitrarily many times. Sidechains like Liquid hit mainnet for pegs in and out and thereafter allow asset issuance and transfer off chain. Even exchanges and custodians – many of which allow “on-us” transfers between users solely on their own database – can be understood as trusted sidechains. If exchanges move from a real time gross settlement model to a net settlement model, yet more block space will open up. 

The commonality here is the introduction of deferred settlement to save on fees (by bundling many off-chain transactions together) and winning efficiencies.

Keep in mind, the above methods rely on an array of trust models, and they are not all equivalent to base-layer bitcoin transactions. But this is how payments work: some high-assurance payments require immediate final settlement, whereas in other cases users are content with deferred settlement. In fact the latter is often preferred, because deferred settlement introduces efficiency and allows for recourse if something goes wrong. The important thing is that bitcoin users have the option of making a base level transfer should they need to.

It’s critical to understand that large-scale block space consumers have a strong incentive to minimize their footprint if wastefulness leads to higher fee liabilities. The rigidity of the blocksize and vibrant block space market means that Bitcoin punishes profligacy. This is why providers like Coinbase have latterly come to embrace space-saving measures like batching, after several years of demurral. Batching reduces the average size of a payment by incorporating many payments into a single transaction (which has a fixed cost in terms of bytes).

Ethereum’s relationship with fees is more complicated. Ethereum 2.0 is a specter that might produce a superabundant quantity of block space, although the execution of this vision remains in doubt. Ethereum itself is far more malleable in its key properties than Bitcoin, with the supply of available block space constantly changing. Increasing the gas limit effectively socializes transaction costs from users to node-operators who must shoulder a costlier validation burden.

Perversely, increasing block space means that heavy users have a reduced incentive to optimize their usage. The Ethereum technical community has devised multiple deferred settlement systems that could reduce the mainchain data impact of transactions, many of them falling under the ‘rollup’ designation. But for industrial consumers of block space, lobbying the developers or miners to increase the gas limit might prove cheaper than rearchitecting their backend to be more parsimonious. 

In a sense, the willingness of the protocol architects to rapidly iterate on the core protocol parameters makes investing heavily in efficiency-enhancing processes less attractive. And the looming prospect of massively abundant block space threatens to derail this newly pragmatic attitude. Ethereum must choose between two visions: the low-fee, endlessly creative and resource intensive world computer, or the more economically dense financial settlement network.