Library›Chains and how they work›Mining Hardware ASICs and the Cost of Production

Mining Hardware ASICs and the Cost of Production

Mining hardware is where the abstract question of what a coin is worth meets a power bill, a purchase order and a shipping date. This page is part of the series under How Cryptocurrencies Gain Value, and it covers the arc from laptop CPUs to the ASIC industry, the Butterfly Labs era that opened it, why every attempt to build an ASIC-proof coin has eventually failed, and the one piece of real economics underneath all of it — cost of production.

The reason it belongs in a series about value: proof-of-work is the only part of crypto where somebody must spend a known amount of real money to produce a coin. That gives a price something nothing else in this field has — a floor made of electricity.

1. The four phases

The operator of this chain was reading Bitcointalk while this happened, and wrote it down at the time. The account below is his, from the 2020 Steemit post Steemit: Under New Management and from the project's 2016 CryptoNight analysis; the phrasing in quotes is his.

• CPU (2009–2010). Ordinary laptops. "When it started, you could mine Bitcoin on your laptop, CPU mining, the more Processors your computer had, the better, so you wanted a Quad Core or more. People got thousands of Bitcoin that way." This is the only period in Bitcoin's history where the distribution problem described in Holders, Rarity, and Distribution solved itself, because the means of production was a thing people already owned.
• GPU (2010–2012). "Then some people made Mining software for Gaming Graphics Cards, GPUs." Still consumer hardware, bought at a shop, but now you had to choose to buy it.
• ASIC (2013 onward). Application-Specific Integrated Circuits, defined plainly: "ASICs, which is a whole box of processors running 1 app." That is the clearest one-line definition of an ASIC on this wiki. "Butterfly Labs was first, then Antminer and others, then it just got crazy with mining farms with warehouses full."
• The industry. What began with Butterfly Labs' first ASICs became, as the Convergence paper puts it, an ecosystem of 100-plus mining-rig companies. The same observation that paper makes about VR and about cannabis extraction applies here exactly: the vision preceded the hardware, sometimes by decades, and the hardware is what converted the vision into an industry.

2. Butterfly Labs, from somebody who was there

The firsthand version is two sentences long and contains the whole transition:

> "Then there was Butterfly labs, and the mining farms. It no longer became feasible to mine with a laptop or number of laptops, and the Graphics cards that had been being used for mining were also being overtaken by the ASICs."

Note what it says: Butterfly Labs did not merely beat laptops, it beat the GPUs that had already beaten the laptops. Two generations of consumer hardware were retired by one product category, inside about a year.

The context it landed in

It matters what the community was like when that happened, because the ASIC did not just change the economics — it changed who the project was for. By his account, he arrived via a news story about the Silk Road, saw a ฿ symbol being used where a dollar sign should be, pasted it into a search engine and landed on Bitcointalk.org:

• The stated goal then was "that everyone would use Cryptocurrency" — a decentralised currency not issued as fiat by a government. Early advocacy was door-to-door in the literal sense: "We would create email lists of companies and everyone, to tell them about the price rise, and ask them to start accepting Bitcoin for their products."
• "When I first got there, Bitcoins were $5.00 each. Soon after they were $30.00, then $100.00. Some people said it was a Bubble and it would 'be affordable again.'" Everyone traded on MtGox — an acronym for "Magic the Gathering Online Exchange" — and "you had to use a Money order to buy them."
• People were already working out that whole coins would become too valuable to transact in, so the smaller pieces "were dubbed 'Satoshis'."

That is the world Butterfly Labs shipped into: a forum of people trying to get shops to accept a $5 currency bought by money order. The ASIC arrived and the question stopped being how do we get everyone to use this and became who can afford the hardware.

What the ASIC actually changed

Before ASICs, the cost of mining was the cost of a computer you already had. After, it was a capital purchase of single-purpose hardware — ordered in advance, manufactured on somebody else's schedule, delivered into a difficulty level that had risen while it was in transit, and worthless for anything else on arrival. That last property is the whole shift. A GPU that stops being profitable gets sold to a gamer. An ASIC that stops being profitable is scrap, and that asymmetry is what turned mining from a hobby into an industry with inventory risk, pre-orders, delivery disputes and balance sheets.

The centralisation that followed was not subtle, and his summary of it is the one to remember: "This simply served to make it harder to mine, as more people come in the people with the most power get the most, but it was easier when Laptops were just competing."

Three consequences outlived the hardware:

1. Mining became a business, so it became concentrated. Capital, cheap power and wholesale chip access are advantages that compound, and nothing in proof-of-work prevents them from compounding.
2. The miners became a chokepoint. Once production is concentrated, so is the power to decide which transactions enter a block — which is the toll-booth problem in §5 of How Cryptocurrencies Gain Value and §7 of Economics 101 and 102. The fee market and the ASIC market are the same story at two different layers.
3. Difficulty eats every efficiency gain. The protocol adjusts difficulty to hold the block interval, so faster hardware across the network does not produce more coins — it produces the same coins for more electricity. Individually rational, collectively a treadmill.

And the halving pushes the same way

Running underneath all of it is the emission schedule, which he states compactly: "Blockchain started with 50 Coins per Block that were released, and it halves every few years, so was 25, then 12.5 and may be or is coming up on 6.25. That makes them harder to get and the fees mean more as it gets lower, and they become more rare."

That is the link most explanations miss. The halving does two things at once: it makes the coin scarcer, and it shifts miner income from the block subsidy toward fees. A chain whose producers are paid less by issuance over time is a chain whose producers care more about the fee auction over time — so the toll booth in §5 of the value page is not a phase, it is the designed destination of a halving schedule.

3. Every ASIC-resistant coin has lost, so far

This is the part a project is tempted to skip, and our own corpus does not skip it.

• Litecoin was built for exactly this reason, and his notes state the motive without decoration: "So Litecoin was made, to allow Laptops again, then Feather Coin because Litecoin got ASICs." Litecoin and Feathercoin adopted scrypt on the promise that it resisted Bitcoin's ASICs. The corpus records the outcome plainly: scrypt ASICs were developed, and the same centralisation occurred.
• Dogecoin arrived late enough that, as the corpus puts it, farms already existed by launch. The pattern repeated with no transition period at all.
• CryptoNight — named after kryptonite, the algorithm being the ASIC's weakness — was the most serious attempt: deliberately memory-hard so that CPUs stayed competitive. Monero was the flagship. The corpus's own retrospective marks it as a partial failure: "Monero eventually faced ASIC pressure, had to fork."
• The community correction recorded alongside that post is the honest footnote: what matters for CryptoNight is CPU cache, at least 2MB per core, not RAM, and many mainstream CPUs did not have it. The conclusion in our own file: "more accessible than Bitcoin, but not truly everyman's coin."

The general lesson is uncomfortable and worth stating: if a hash function is valuable enough to specialise for, somebody will specialise for it. ASIC resistance buys years, not permanence.

4. Which is why these chains do not mine for consensus

The corpus draws the conclusion itself, in 2016, about STEEM: because it is Delegated Proof-of-Stake, there is no advantage to ASICs — block frequency is identical regardless of hardware. The hardware arms race is not won, it is declined. MELEK inherits exactly that, and it is why MELEK has no mining at all: see Proof of Work, Proof of Stake, and DPoS.

PRANA does mine, with Etchash, and the distinction is deliberate: PRANA's work is the chain's product rather than only its clock. See Proof of Work Mining for the algorithms and PRANA's TASK lane. The browser miner on the pool is the same argument as CryptoNight's in a new form — put the means of production back on hardware people already own — and it carries the same honest caveat: a browser tab cannot compete with a farm, and saying otherwise would be a lie.

5. Cost of production — the only floor in crypto

Here is the piece that earns this page its place in a series about value.

For a mined coin you can compute what it cost to make: hash rate, hardware efficiency, electricity price. That figure behaves like a soft floor, for a mechanical reason — when price falls below the cost of production, miners who cannot cover power switch off, hash rate falls, difficulty adjusts down, and supply issuance gets cheaper to produce. Sellers who were selling to pay bills stop existing.

Four honest limits:

• It is a soft floor. Miners with sunk hardware, subsidised power or a strategic reason will mine at a loss for a long time, and price has traded below estimated production cost in real cycles.
• The floor moves, because difficulty and energy prices move.
• It says nothing about the ceiling. Cost of production never explains a bull market.
• It does not exist for staked chains at all. MELEK has no production cost in this sense — its issuance is near-free — so nothing underneath its price comes from physics. That is the trade for being feeless and fast, and it means everything in Utilities and Sinks has to do the work that electricity does for Bitcoin.

That last bullet is the most useful sentence on this page for anyone holding our coins, which is why it is stated rather than buried.

6. Where the energy comes from, and "pollution mining"

§5 treated cost of production as a single number. It is not — and the correction is the most useful thing on this page for understanding both the pollution argument and why distributed mining is not a fantasy.

The floor is a curve, not a number

Every miner has a different cost per coin, because every miner pays a different price for electricity. So there is no single production cost; there is a cost curve, and the floor is set by the marginal miner — the highest-cost one still running. When price falls, the expensive miners switch off first, hash rate drops, difficulty adjusts down, and the remaining miners get cheaper coins. The floor does not hold the price up so much as it sorts the miners.

That is why the energy question is an economic question and not only a moral one. Where your power comes from decides where you sit on that curve, and therefore whether a drawdown kills you or hands you your competitors' share. A miner whose panels are paid off has a marginal cost near zero and keeps mining through prices that shut a grid-powered farm down. See Solar Mining.

What the pollution objection actually says

"Mining wastes electricity and pollutes" is the most common criticism of proof-of-work, and it deserves a straight answer rather than a deflection.

The objection is really about fossil-fired grid power, and in that form it lands: miners chase the cheapest electricity on earth, and in some regions the cheapest electricity is coal. Where that is true, the critique is simply correct — the cost of production is partly paid by people downwind who never bought a coin. That is an externality, and calling it one is more honest than arguing about it.

The counter-argument is narrower than miners usually claim but it is real: a miner is an interruptible load that can go where the energy already is and would otherwise be thrown away — curtailed solar and wind at peak generation, stranded hydro, flared gas. Converting energy that was going to be wasted is a genuinely different act from burning new coal for it.

And the honest limit, which Solar Mining states and this page repeats: solar mining removes the fossil-waste objection; it does not claim a magic zero. Panels and batteries carry an embodied manufacturing footprint, and off-grid batteries are a real recurring materials cost. The defensible claim is the strong true one, not the bigger false one.

The town that outmines the farm — the hashrate version of §2

The operator's answer to ASIC centralisation was written in 2018, in Cryptocurrency Towns, Part 1, and it is worth quoting because it is the same argument as the distribution arithmetic in Holders, Rarity, and Distribution with hash rate substituted for money:

> "Once we have established our Town Based Currencies, we will begin setting up Solar Powered Mining Machines. Which will allow for people to generate Income for Free. If there is a town where every house uses the Sun and Alternators/Generators Powered by Solar Panels, then just as an example, 25,000 people living in a Town with Alternators and Generators and Solar Panels, could probably outmine every existing Bitcoin/Litecoin/Ethereum Farm."

Look at the shape of it. Twenty thousand households at one roof each beats a few warehouses — the identical structure to ten thousand people at ten dollars beating twenty rich people. And it answers the §2 centralisation problem from a completely different direction than every failed ASIC-resistant algorithm did: it does not try to make specialised hardware useless, it changes who owns the power bill. The farm's floor is a monthly invoice. The town's floor is sunlight.

That is a thesis, not a measurement, and the arithmetic in it is an explicit "just as an example" rather than a calculation anyone has verified. What makes it serious rather than decorative is that it identifies the correct variable: in proof-of-work, decentralisation follows energy ownership, and nothing else has ever reversed the trend that Butterfly Labs started.

Where this ecosystem actually stands on it

Status, labelled, because the charter on this project forbids the alternative:

• PRANA mines — Etchash, live, with consensus paying every miner by share each 20 blocks. The browser miner is the "hardware people already own" idea in current form, and carries the same honest caveat: a browser tab cannot compete with a farm.
• Proof-of-Solar is written and tested, and not deployed. `ProofOfSolarOracleMint` exists as a contract with tests and a generated ABI — a mint that issues against attested kWh, with no pre-mine and no founder allocation — but it does not appear in the PRANA mainnet deployment record, and nobody is earning tokens from verified sunlight today. It is a design, and should be read as one until it has an address.
• The solar-mining knowledge is the part that is actually delivered — Solar Mining, Electricity and Photovoltaics and the dye-cell pages are live and free to anyone, which is the half of "provide the means" that does not require a contract.

7. Read next

• How Cryptocurrencies Gain Value — the spine of the series; §5 is the fee market this page's chokepoint produces.
• Solar Mining — the panels, the economics after payback, and the pollution question in full.
• Proof of Work Mining · Proof of Work, Proof of Stake, and DPoS · Mining · Electricity and Photovoltaics
• Holders, Rarity, and Distribution — why the CPU era's distribution mattered.
• Economics 101 and 102 — the toll booth, the fee bottleneck and the monopoly framing.
• Utilities and Sinks — what substitutes for a production floor on a feeless chain.

Not investment advice

This page is hardware history and cost mechanics. It is not a valuation of any coin, not a claim that any price has a floor, and not a suggestion to buy, sell or start mining. Mining is a capital purchase with electricity costs and can lose money.

Sources

• `knowledge/cryptocurrency/cryptonight_cpu_mining.json` — the project's own corpus entry drawn from the marsresident Steemit post, "CryptoNight Coins — Mining with Regular Computers" (2016, STEEM era): the four hardware phases, Butterfly Labs and Block Erupter as early ASIC players, the scrypt/Litecoin/Feathercoin and Dogecoin outcomes, CryptoNight's kryptonite naming and cache requirement, the Monero ASIC-fork retrospective, and the DPoS comparison.
• `knowledge/scripture/the_convergence.md` — Butterfly Labs' first ASICs to an ecosystem of 100+ mining-rig companies, and the vision-precedes-hardware argument.
• @marsresident, "Steemit: Under New Management" (Steemit, 27 September 2020) — the operator's firsthand account: the Silk Road and ฿ symbol route to Bitcointalk, the $5/$30/$100 arc and the "affordable again" bubble talk, MtGox and money orders, the coining of "Satoshis", and the Butterfly-Labs-and-the-mining-farms passage quoted in §2.
• The project's "Hashtags of Money" teaching notes — the halving schedule and its effect on fees and rarity, "a whole box of processors running 1 app", "Butterfly Labs was first, then Antminer and others", the Litecoin/Feathercoin/CryptoNight sequence, and Bitshares Assets versus Ethereum Tokens.
• This project's "Economics 101" Bitcointalk thread (immakingacoin, 2018) — the miner chokepoint and fee-bottleneck framing.
• @marsresident, "Cryptocurrency Towns, Part 1: Getting Started" (Steemit, 11 July 2018) — the solar-powered town thesis and the 25,000-household example quoted in §6.
• `.local/publish-live/src/prana-contracts/` — `ProofOfSolarOracleMint.sol`, its test and ABI; and `deployments.json` (prana_mainnet, chainId 712217), which contains no solar entry. Basis for the not-deployed status in §6.

Coverage

Covers the CPU/GPU/ASIC/industry phases of mining hardware, the Butterfly Labs transition and its three lasting consequences (concentration, the miner chokepoint, the difficulty treadmill), the repeated failure of ASIC resistance across scrypt and CryptoNight, why DPoS declines the arms race while PRANA mines deliberately, cost of production as a soft price floor that staked chains do not have, and the energy question — the cost curve set by the marginal miner, the fossil-power externality stated as one, curtailed/stranded energy as the real counter-argument, and the 2018 solar-town thesis as the hashrate form of the distribution argument. Proof-of-Solar is labelled written-and-tested-but-not-deployed against the mainnet deployment record. Sourced from this project's own corpus and the operator's firsthand account rather than from external histories. The Butterfly Labs section quotes his 2020 Steemit post Steemit: Under New Management and his teaching notes directly. Company-level chronology that he did not record — Butterfly Labs' pre-order and delivery disputes and the FTC action against it — is deliberately not asserted here, because this page's sourcing is the project's own writing.

Filed under  Chains and how they workTools