The 1% Fee Crisis: Why Bitcoin Miners Are Preparing for a Post-Subsidy Reality

Bitcoin miners collected less than 1% of their total revenue from transaction fees in the second quarter of 2025. The last time this metric dipped this low was in 2015, when the network was a fraction of its current size. The ledger remembers what the hype forgets: the Ordinals and BRC-20 frenzy of 2023 was a speculative spike, not a structural shift. With the 2024 halving cutting the block subsidy to 3.125 BTC, miners now earn roughly $200,000 per block from newly minted coins, but less than $2,000 from fees. That is a 99% subsidy dependency. I do not cover the story; I follow the code. The code shows a mempool that is chronically empty, with transaction counts hovering near 1,500 per block – well below the 4,000 equivalent weight capacity. Utility vanished before the mint even cooled. Bitcoin mining revenue is a two-part equation: subsidy (inflation) plus fees. Since the genesis block, subsidy has dominated. But the 2024 halving halved the subsidy from 6.25 to 3.125 BTC. At current prices, that slashed dollar-denominated miner income by roughly 50% overnight. The fee component, which briefly surged to 20% during the Ordinals mania, has since collapsed. According to on-chain data, the average fee per transaction has fallen to under $1.50, a level that barely covers the cost of inclusion. The network's security budget – the total dollar value of rewards paid to miners – now stands at around $15 billion annually, down from $30 billion pre-halving. This is not a transient dip; it is a structural reality that will persist until either fees rise or the next halving arrives in 2028. Let me dissect the mechanics. Bitcoin's block space is a scarce resource – 1 MB (or 4 million weight units) every 10 minutes. When demand for blockspace is low, users pay minimal fees. The current mempool often shows zero backlogs, meaning any transaction with a dust fee gets included within minutes. This is a classic supply glut. The market for block space is oversupplied because the network does not have enough use cases demanding settlement. Lightning Network absorbs some microtransactions, but it does not generate on-chain fees. The result: fee revenue is negligible. Compare this to Ethereum, where EIP-1559 introduced a base fee that is burned, creating a dynamic fee market. Ethereum validators currently earn 10-20% of their revenue from tips, and the base fee adjusts with demand. Solana, despite its low fees, has a local fee market that prevents congestion spillover. Bitcoin has no such mechanism. It relies on a first-price auction where the fee floor is zero when mempool is empty. This is not a design flaw; it is a deliberate trade-off for simplicity. But it creates a vulnerability: the network's security is entirely dependent on the subsidy. Now, the subsidy is programmed to halve every four years. In 2028, it will drop to 1.5625 BTC per block. At current BTC prices, that would slash annual miner revenue to $7.5 billion. To maintain the same security level, the BTC price would need to double every four years, or fees would need to rise dramatically. Neither is guaranteed. The price of Bitcoin is volatile; fees have been declining. I have tracked miner behavior since the 2024 halving. The response has been a wave of diversification. Publicly traded miners like Hut 8 and Core Scientific have signed multi-year contracts to host AI and HPC computing. They are effectively converting their infrastructure – power assets, cooling, real estate – into AI data centers. This is not a small pivot. In 2025, it is estimated that over 20% of former Bitcoin mining facilities are now partially or fully dedicated to AI workloads. This is a silent admission that the Bitcoin mining business model, as a standalone operation, is not sustainable. The concentration of hash power is another consequence. When subsidy dependency is high, only the most efficient miners survive. The top three mining pools – Foundry USA, Antpool, and ViaBTC – now control over 60% of global hash rate. This centralization contradicts the ethos of a decentralized network. And with fee income low, small miners have no cushion. They exit, and the pools grow. Silence in the code is the loudest confession: the network is becoming more centralized under the guise of market efficiency. Furthermore, the assumption that BTC price appreciation will compensate for subsidy decline is flawed. Historical data shows that after halvings, price does not always rise proportionally. The 2025 market is sideways, with ETF flows providing a floor but no catalyst. If price stagnates, miners face a cash flow crisis. They are forced to sell more Bitcoin to cover operational costs, which creates selling pressure, which suppresses price – a negative feedback loop. The fundamental issue is that Bitcoin's fee market is not designed to sustain security in the long term. The capped supply means subsidy will eventually approach zero. The network must rely on fees. But fees are currently only 1% of revenue. This is a 10-year low because the network has not solved the 'fee problem' – it has only deferred it. To be fair, the bulls have a point. Low fees are good for user adoption. The Bitcoin network is not designed to be a high-throughput settlement layer; it is a store of value with occasional transfers. The Lightning Network and other L2s handle the bulk of transactions, and they rely on the main chain only for channel openings and closures. This means on-chain fee demand is naturally low. Some argue that the security budget is still sufficient because the absolute dollar value of subsidies is high – $15 billion annually is not trivial. Also, the difficulty adjustment mechanism ensures that if miners exit, mining becomes easier for the remaining ones, keeping the network secure at a lower cost. Finally, the AI/HPC diversification could be seen as a positive – miners are becoming more resilient businesses, able to weather Bitcoin price cycles. But these arguments miss the core issue: the network's security is only as strong as the revenue miners can earn. If the subsidy is the only real carrot, and it is shrinking, then the network is on a path to reduced security unless fees increase dramatically. We traded value for visibility, and lost both. The next halving is three years away. The clock is ticking. If Bitcoin's fee revenue does not grow – either through increased on-chain activity or protocol changes – the security model will face a stress test. Miners are already voting with their feet, moving to AI. The question is not whether Bitcoin can survive without miners; it is whether the incentives align for enough miners to stay. The ledger remembers what the hype forgets: security is not free.