The security model that has protected Bitcoin for nearly eight years may have a critical flaw lurking in its future. A groundbreaking research paper published this week by a team of Princeton University computer scientists warns that Bitcoin’s blockchain could become fundamentally unstable once mining rewards shift from block subsidies to transaction fees — a transition that is baked into the protocol’s design but has received surprisingly little scrutiny from a security perspective.
TL;DR
- Princeton researchers published a paper finding Bitcoin’s consensus mechanism becomes unstable without block rewards
- The paper identifies new “undercutting” strategies where miners steal transaction fees from competitors
- Selfish mining attacks become more profitable and more damaging in a fee-dominated regime
- Currently 99.8% of miner revenue comes from block rewards, but this will change over decades
- The research was presented at ACM CCS 2016, one of the world’s top cybersecurity conferences
The Block Reward Problem No One Wanted to Talk About
Bitcoin miners are compensated through two mechanisms: block rewards and transaction fees. As of October 2016, block rewards stand at 12.5 BTC per block following the second halving in July, and they overwhelmingly dominate miner income. According to the Princeton team’s analysis of the previous 1,000 blocks, approximately 99.8% of all miner revenue came from the fixed block subsidy, with transaction fees contributing a negligible fraction.
But Bitcoin’s monetary policy dictates that block rewards halve approximately every four years, eventually approaching zero around the year 2140. At that point, miners will rely entirely on transaction fees to secure the network. The implicit assumption in the Bitcoin community has been that this transition would be seamless — that it makes no meaningful difference whether miners receive 12.5 BTC as a fixed reward or 12.5 BTC in expected transaction fees.
The Princeton paper, titled “On the Instability of Bitcoin Without the Block Reward,” challenges this assumption head-on.
Undercutting: A New Mining Strategy
The researchers identified a class of previously unknown mining strategies that become profitable only when block rewards are replaced by variable transaction fees. The most concerning is called “undercutting,” where a miner deliberately includes only a small portion of available transaction fees in their block, leaving the rest in the mempool as bait.
The logic is ruthless but sound: if the next miner sees a large pool of uncollected fees sitting in the mempool, they have a financial incentive to extend the current chain rather than attempt to fork it. The undercutting miner essentially pays a “tax” to secure their block’s position in the chain, sacrificing some immediate revenue for a higher probability that their block remains part of the longest chain.
In a fixed block reward environment, this strategy makes no sense — every block is worth exactly the same amount. But when block values fluctuate based on accumulated fees, the incentive landscape shifts dramatically.
Selfish Mining Gets Worse
The paper also demonstrates that selfish mining — a well-known attack where miners withhold discovered blocks to gain a head start — becomes significantly more profitable in a fee-dominated system. The intuition is straightforward: if you mine a block just seconds after the previous one, the accumulated transaction fees will be minimal. Publishing immediately yields almost nothing, but withholding the block for potential selfish mining carries very little downside.
This creates a cascading effect where the variance in transaction fees between blocks actively encourages deviant mining behavior that simply does not exist when block rewards are fixed.
Simulations Confirm the Theory
To validate their theoretical findings, the research team built a comprehensive Bitcoin mining simulator and released it as open-source software. The simulator models various mining strategies competing against each other, and the results closely matched the paper’s analytical predictions.
In one particularly striking result, the theory predicted an equilibrium involving the Lambert W function — a mathematical relationship that arises in complex exponential systems. When the researchers simulated the same scenario, the “Lambert miner” strategy indeed emerged as dominant, providing strong empirical confirmation of the theoretical framework.
The 1MB Band-Aid
The paper acknowledges one temporary mitigating factor: Bitcoin’s 1 megabyte block size limit, which has been the subject of intense community debate throughout 2016. Because blocks are consistently full, the competition for block space creates a more predictable fee distribution, reducing the variance that enables undercutting and other deviant strategies.
However, the researchers are clear that this is far from a complete solution. The block size limit is a policy choice, not a fundamental security mechanism, and relying on permanently full blocks as a security feature introduces its own set of problems for Bitcoin’s usability and adoption.
Implications for the Blockchain Ecosystem
The findings have implications far beyond Bitcoin. Any blockchain that uses a declining block reward schedule — which includes the vast majority of proof-of-work cryptocurrencies — faces the same long-term challenge. Ethereum, currently the second-largest cryptocurrency by market capitalization at $12.07, has its own set of growing pains, including ongoing denial-of-service attacks that have slowed its network throughout October 2016. These parallel challenges underscore that the entire blockchain ecosystem is still in its formative stages, with fundamental questions about long-term sustainability yet to be fully resolved.
The research team suggests that the Bitcoin community will likely need to implement protocol changes, potentially through a hard fork, to mitigate these issues before block rewards become insignificant. With the next halving not expected until 2020, there is time to develop solutions, but the complexity of coordinating changes across Bitcoin’s decentralized stakeholder base means that the conversation needs to start well in advance.
Why This Matters
This research represents one of the first rigorous, peer-reviewed examinations of Bitcoin’s long-term economic security assumptions. While the block reward won’t reach zero for over a century, the transition period — where fees become a meaningful fraction of miner revenue — could arrive much sooner. Understanding these dynamics now gives the Bitcoin community decades to prepare, rather than scrambling to fix a crisis after it manifests. The fact that this work comes from Princeton’s Center for Information Technology Policy, one of the most respected technology policy research centers in the world, gives it additional credibility and ensures it will receive serious attention from both developers and policymakers.
Disclaimer: This article is for informational purposes only and does not constitute financial advice. Cryptocurrency markets are highly volatile, and readers should conduct their own research before making any investment decisions.
princeton was ahead of its time here. the undercutting attack vector is still mostly unresolved years later
ACM CCS is a top tier venue. this was not some blog post, it was peer reviewed and the undercutting analysis held up
hash_entropy the ACM CCS peer review is what makes this credible. undercutting attacks where miners steal fees from competitors blocks is a genuine consensus vulnerability
undercutting attacks where miners steal tx fees from competitors blocks is a genuine consensus vulnerability that nobody talks about
Andrei the undercutting strategy sounds theoretical until you watch miners during fee spikes. 99.8 percent from subsidies looked safe but the math flips fast
nonce_rotator unresolved is the key word. were 10 years out from this being urgent but nobody in governance is even drafting proposals for fee smoothing or consensus tweaks
Magnus H. governance moves at glacial speed. by the time fees dominate nobody will agree on a fix and miners will just undercut until something breaks
Magnus H. 10 years sounds far but bitcoin governance makes DNS look fast. if they start debating fee smoothing now maybe they ship something by 2035
99.8% of revenue from block rewards back then. we are slowly moving toward the fee-dominated model and the concerns are becoming real
Marek Zielinski the 2024 halving cut block rewards to 3.125 BTC. fees covered 30-50% of miner revenue on busy days. princeton was right to flag this early
marek the fee-dominated model already showed cracks during the 2024 run fee spikes to $50+ per tx. imagine that being the only miner incentive
fee_spike_ the 2024 halving cut rewards to 3.125 BTC and fees covered up to 50% of miner revenue on busy days. princeton flagged this in 2016 and everyone shrugged
Mikko R. the scary part is undercutting becomes profitable before fees dominate. even at 20% fee revenue the game theory shifts
marek 99.8% from block rewards back then. fast forward and were already seeing fee spikes during congestion events. the transition is happening faster than people think
fee spikes during congestion events are a preview of the fee-dominated future. when block rewards are negligible miners will chase fee revenue aggressively
block_subsidy saw fees hit 2000 sats/vbyte during the inscriptions craze in early 2024. miners were making more from fees than the block reward itself for days
undercut_obs fees at 2000 sats/vbyte in 2024 proved princeton right. miners chasing fee revenue during inscriptions was the preview
undercut_obs the inscriptions craze proved the princeton model correct faster than anyone expected. miners literally chasing fee revenue and ignoring subsidised blocks
princeton called this in 2016 and the 2024 halving already proved them right. fees hit 2000 sats/vbyte during inscriptions and miners literally chased fee blocks
difficulty_anchor the 2024 inscriptions fee spike was a preview but miners chasing fee blocks is different from undercutting. undercutting means actively stealing from other miners mempools. the game theory gets ugly fast
fee market solutions like packaged txs or commitment schemes could smooth this out but core devs barely discuss it. everyone assumes Lightning solves everything
Gintaras R. Lightning assumes base layer liquidity which assumes miners include txs. if undercutting makes tx inclusion unpredictable the whole stack wobbles
Carlsten et al identified the undercutting attack vector in 2016 and nobody cared because block rewards were 12.5 BTC. fast forward to 2140 and fee-only security becomes an existential question.
Florian K. 99.8% of miner revenue from subsidies today. the transition to fee dominance is glacial but the instability risk compounds with each halving
ACM CCS 2016 was the right venue for this. the paper got peer reviewed and still got ignored by the entire industry because it was too far out.