Why the question comes up
Blockchains are sometimes described as wasteful, and sometimes as being unfairly blamed. Both views contain part of the truth because “blockchain” covers designs that differ enormously in how they work. A permissioned ledger shared among a few dozen companies might use no more electricity than a few ordinary servers. A large public network built on competitive computing can use a significant amount of power. The useful question is not whether blockchains use energy, but which design choices cause the difference.
The main driver: how the network agrees
The biggest factor is the agreement method, covered in the article on consensus mechanisms.
In proof of work, participants compete to find a valid answer to a hashing puzzle. Success is determined by how many guesses a participant can make per second, so more computing power means a better chance of winning the reward. Power is converted into guesses, and almost all of those guesses are discarded. The “waste” is deliberate: the cost of the effort is what makes rewriting history expensive, as it ties the security of the record to real-world resources.
In proof of stake, participants are chosen to add blocks based on assets they have locked up, not on computation. There is no race to burn electricity, so the network mainly needs ordinary servers. At least one very large network has switched from proof of work to proof of stake, and its developers reported a dramatic fall in energy consumption afterward. Even so, claims should be checked against independent estimates.
In permissioned voting among known members, nodes exchange messages and tally votes. The computing load is modest.
Why competition drives consumption upward
In proof of work, there is an economic feedback loop worth understanding. Rewards, usually newly issued tokens and transaction fees, attract participants. When the reward is worth more in the market, more participants join and buy more machines. As more computing power enters, the network’s difficulty setting adjusts so blocks keep arriving at a roughly steady rate. Each participant must therefore spend more to keep the same share, until the cost of operating approaches the value of the reward.
Two consequences follow. First, energy use tends to track the value of the reward rather than the number of transactions processed. Second, adding more users does not by itself force a proportional rise in energy, because the puzzle work is not performed per transaction. Per-transaction comparisons with payment systems therefore need to be read carefully, since dividing a fixed-ish total across a small number of transactions gives large figures, and the result depends on assumptions.
Hardware matters
The computers used also change the picture. Early participants used ordinary processors. Over time, specialized chips built for one hashing task became dominant because they are far more efficient per unit of electricity than general-purpose machines. Efficiency gains do not necessarily reduce total consumption, though, since if efficiency makes mining more profitable, more machines may be added. This pattern, where savings are offset by increased use, is a general feature of many technologies.
Machines also have a limited lifespan and become obsolete, so manufacturing and electronic waste are part of the environmental picture, not just electricity.
Electricity source versus quantity
How much power is used is one question; where it comes from is another. A large load running on mostly renewable energy has a different climate impact than the same load on fossil fuel generation. Some operators locate near cheap surplus power, such as remote hydropower or output that would otherwise be curtailed. Supporters argue this can use intermittent renewables that would go to waste, and that it can add flexible demand to a grid. Critics respond that it may still divert power from other uses, prolong the life of old plants, or raise local prices. The real balance varies by location and is hard to measure. The effect on smart grids and local supply is one reason regulators have taken an interest.
What else uses energy
Beyond mining or validation, other parts of the system consume power, though generally much less.
- Running nodes. Every full node stores and checks the ledger, which uses storage, bandwidth and some computing. Multiplied across thousands of copies, this redundancy is a cost of having no central keeper.
- Applications. Smart contracts execute on every node, so complex programs add work across the network.
- Data centers and devices. Some node operators rent space in data centers, and users’ own devices and wallets add small amounts.
Why estimates differ
Numbers quoted in news stories often vary widely and sometimes disagree by large factors. Reasons include:
- Energy use cannot be measured directly on a decentralized network, so analysts estimate from hardware assumptions and market data.
- Assumptions about which machines are in use, and how efficient they are, change the result.
- The energy mix is difficult to determine, and varies over time.
- Comparisons with banks, gold mining or card networks depend on what is included, and these systems are not strictly equivalent in what they provide.
For these reasons, this article avoids quoting specific totals. When you meet a single precise figure, ask who produced it and how.
Design choices that reduce energy use
Beyond switching agreement methods, designers have used several approaches: moving most activity to secondary layers that bundle many transactions into a few ledger entries, limiting participation to known members, and choosing storage and data formats that reduce the load on nodes. Each comes with trade-offs, often in openness or in how much trust is placed in particular parties.
Weighing the trade-off
Whether the energy used by a given network is justified depends on what you think its security and independence are worth, which is a value judgment, not a technical fact. A reasonable approach is to ask three questions: what does this design use energy for, is that use necessary to the goal it serves, and are there lower-energy ways to meet the same goal? The honest answers differ from design to design, so generalizations about blockchains as a whole usually mislead.