cryptonist
A printed circuit board photographed flat, its traces and mounted components in rows

Explainers

Cloud mining and mobile mining apps: what the claims rest on

By Greg Lacan Reviewed 03/10/2026

Two products are sold as a way to earn bitcoin without owning mining equipment: a contract that rents a share of someone else’s hashrate, and a phone app on which a daily tap is said to mine. Both rest on one claim, that mining pays more than it costs and that the seller can hand part of the surplus to the buyer. Two listed miners’ annual reports show how much of the value mined is left after the cost of mining it, so the useful question is not whether an operator seems honest but what would have to be true for its terms to be met.

What mining pays per unit of hashrate

Mining is a lottery in which each hash is a ticket. A block joins the chain only if its header hashes to a value below a target, and the miner whose block is accepted collects the block reward: in the Bitcoin developer documentation’s words, a block subsidy plus any transaction fees paid by the transactions in the block. Hashrate is the number of attempts per second. Difficulty is how demanding the target is, and it is not fixed. Every 2,016 blocks the network compares how long the last 2,016 took with the ideal 1,209,600 seconds, two weeks, and rescales the difficulty so that, at an unchanged hashing rate, the next 2,016 should take two weeks. One adjustment can raise it “by as much as 300%” or lower it “by as much as 75%”, in the documentation’s wording. Bitcoin Core enforces the rule: CalculateNextWorkRequired clamps the measured timespan to between a quarter of the two-week target and four times it, and the mainnet parameters set that target at two weeks and the block spacing at ten minutes.

That has one consequence for everything below. The subsidy is fixed per block (the tokenomics piece traces that rule in the same software), and each adjustment pulls the block rate back towards one every ten minutes. What is paid per day stays tied to the block count and is shared among whatever hashrate is present, so revenue per unit of hashrate falls each time the total rises, and only hashrate leaving or a larger reward raises it again. The payer is the protocol’s issuance plus users’ fees, the question where crypto yield comes from puts to every yield: who is paying, and for what.

Figures read on 3 October 2026. mempool.space put the chain tip at block 969,748, and its mining statistics gave a difficulty of 132.7 trillion (132,716,002,350,731). The adjustment record shows that difficulty was set at block 969,696 that morning (07:16 UTC). The subsidy at block 969,748 is 3.125 BTC (computed from Core’s schedule: 50 coins, halved once for each 210,000 blocks, four times by block 840,000). Over the 144 blocks from 969,604 to 969,747, mempool.space’s reward statistics show 3.86 BTC of fees in 453.86 BTC of total rewards, 0.85% (computed).

From those inputs, computed: the expected number of hashes needed to find a block at difficulty D is D × 2⁴⁸ ÷ 0xffff. In Bitcoin Core, GetBitsProof gives a target’s expected hashes as 2²⁵⁶ ÷ (target + 1) and GetDifficulty reports difficulty as 0xffff × 2²⁰⁸ ÷ target, so the target cancels, the 1 being negligible against a target that large. The Bitcoin Wiki’s difficulty page simplifies it to about D × 2³², here 5.70 × 10²³. A miner running one terahash per second, or 10¹² attempts a second, makes 8.64 × 10¹⁶ attempts a day, so its expected output is 1.52 × 10⁻⁷ blocks a day. At a reward of 3.1518 BTC a block (the subsidy plus the 0.0268 BTC average fee over those 144 blocks), that is 47.8 satoshis a day per terahash per second, or 0.000000478 BTC, where a satoshi is a hundred-millionth of a bitcoin. It is a gross, expected figure: before any cost, before a pool’s fee, and not what any contract pays. One exahash per second (EH/s) is a million terahashes per second (TH/s). As a check on the order of magnitude, Riot Platforms’ annual report for 2025 gives 5,686 bitcoin mined at an average operating hashrate of 31.4 EH/s, which works out at 49.6 satoshis per terahash per day over the year (computed).

The cost a contract has to beat

Three costs sit under every terahash: electricity to run the machines, the machines themselves, and the site, staff and upkeep that keep them running, or, for a miner that rents space, a hosting fee in place of the site. Listed miners report these in annual reports, which makes the reports public evidence of what hashrate costs to produce.

Riot’s report for the year to 31 December 2025 calls power “overwhelmingly the largest marginal input cost”, and sets out its cost to mine one bitcoin in a table. Excluding depreciation of its miners, and after power curtailment credits, the cost was 49.0% of the production value of a bitcoin mined, the report’s measure being the revenue recognised from bitcoin mined divided by the number mined. Including depreciation, which Riot spreads over a useful life of three years, it was 90.2%. For 2024 the two shares were 48.5% and 96.9%. The cost covers more than power (the report lists compensation, insurance, repairs, ground rent and property taxes among the rest).

CleanSpark’s report for the year to 30 September 2025 gives a comparable measure separately for sites it owns and sites where third parties hosted its machines. At the owned sites its direct cost to mine counts only energy, other utilities, miner depreciation and financing, and leaves out the compensation, insurance, repairs, rent and taxes that Riot’s figure includes. The direct energy cost was 44.0% of the average revenue from each bitcoin mined and the whole direct cost, depreciation included, was 84.7%. At hosted sites, hosting fees and profit-sharing together were 71.5% of revenue in the year to 30 September 2024 and 96.7% in 2025, when those sites produced 147 bitcoin (CleanSpark no longer mined at hosted sites as of 31 March 2025); with depreciation the 2025 share stayed at 96.7% and the 2024 share was 115.3%, so in 2024 the direct cost of those machines exceeded the value they mined.

These are the filers’ own measures of direct cost, not the total cost of running the company. They come from two filers, a year or two each, and nothing here extends them to the industry. Within them, the value left after the direct cost of producing a bitcoin, depreciation included, was 9.8% of production value at Riot in 2025 (3.1% in 2024), 15.3% at CleanSpark’s owned sites in its 2025 year and 3.3% at its hosted sites in that year, each computed as 100% less the reported share. At CleanSpark’s hosted sites in 2024 it was below zero.

A contract’s output arrives later as coin, while a buyer who pays in currency pays on day one. Per unit of hashrate the network pays every miner the same, 47.8 satoshis per terahash per day on 3 October 2026 and falling with each rise in difficulty, and the filers above put the direct cost of producing it at 84.7% to 96.7% of the value mined in their latest years. The cost bases are not identical, and the top of that range, CleanSpark’s hosted sites, is the closest analogue to a rented contract. A retail price set at or above a seller’s cost, if that cost resembles theirs, therefore leaves the buyer at most what those filers keep, 3.3% to 15.3% of the value mined, less the seller’s margin and fees. Any return beyond that is a position on the coin’s price, and on difficulty not rising, with a mining contract wrapped around it. For a contract to pay more than that without help from the coin’s price, one of two other things has to hold, and neither can be established from a contract page. The seller may produce hashrate for far less than these filers do and pass the saving to a retail buyer instead of mining for itself, which is a claim about its electricity contracts, and the kind of claim that the SEC’s complaint against Geosyn Mining, below, alleges was false in that case. Or the payouts come from somewhere other than mining, which is the structure described next.

How cloud mining contracts fail

By a cloud mining contract this piece means a purchase of a stated amount of hashrate for a stated term, run by the seller, with the output credited to the buyer.

Hashrate that the buyer cannot check on the chain. In pool mining, as the developer documentation’s mining guide describes it, a miner sends the pool a share, which proves the miner did part of the work. The block reward and fees are paid to the pool, and the pool pays miners in proportion to the shares it has received. In that arrangement, the division among miners is the pool’s own calculation, and what the chain records is the reward paid to the pool. A buyer’s slice is also too small to be seen on the chain alone: at the inputs above, a hypothetical contract of 100 TH/s would expect to find a block on its own about once in 180 years (computed from the 1.52 × 10⁻⁷ blocks a day per terahash, which is about 66,000 days). Unless the buyer can see the pool’s own records, the evidence that hashrate is pointed at the buyer’s account is a dashboard, which is the seller’s rendering of its records and not a proof.

A fixed fee against falling revenue. Revenue per terahash moves with the inverse of difficulty, so a contract that charges a fixed daily fee per terahash, set in currency, has the fee take a larger share of revenue at every rise. Computed: a 30% rise in difficulty, with the coin’s price unchanged, cuts gross revenue per terahash to 76.9% of what it was (1 ÷ 1.3), so a fee above 76.9% of revenue before is above all of it after. The protocol allows bigger steps, up to the 300% quoted above. In the 26 adjustments the same adjustment record lists for the past year, the largest single rise was 14.7% (block 937,440, 19 February 2026) and the largest fall 11.2% (block 935,424, 7 February 2026), so a 30% step is larger than any in that record.

Terms that decide whether the balance leaves. Two kinds of clause do: a minimum below which nothing can be withdrawn, and a right for the operator to pause payouts or the contract. A minimum that the whole term’s output never reaches leaves a balance that cannot be collected, and a right to pause leaves the timing with the operator. The SEC’s Office of Investor Education and Advocacy lists “difficulty receiving payments” among the red flags it gives for Ponzi schemes in its July 2013 investor alert on Ponzi schemes using virtual currencies, and notes that organisers sometimes encourage participants to “roll over” promised payments by offering higher returns. The alert concerns Ponzi schemes in general and does not discuss mining contracts’ terms. A holder whose balance sits with an operator that stops paying is in the position described for exchanges: a claimant, whose standing is set by the customer agreement and the law where the entity sits. An operator holding balances can also have them stolen, and how crypto gets hacked sorts the recorded thefts by how they happened.

New deposits paying old payouts. The SEC’s press release of 1 December 2015 announced charges against two companies, GAW Miners and ZenMiner, and their founder, for conducting a Ponzi scheme. According to the complaint it describes, they sold $20 million of purported shares in a digital mining contract called a Hashlet between August and December 2014, to more than 10,000 investors, touting it as “always profitable and never obsolete”. It adds that GAW Miners “directed little or no computing power toward any mining activity”, and that because the defendants “sold far more computing power than they owned” they owed investors a daily return larger than anything their limited mining earned, so investors were paid back out of funds collected from other investors. In the complaint’s account most investors never recovered their full investment, and few made a profit. The release states allegations; the outcome is in two later SEC releases. Litigation Release No. 23852 (5 June 2017) reports a final default judgment against the two companies, entered on 2 June 2017, with permanent injunctions, $10,384,099 in disgorgement and prejudgment interest owed jointly and severally, and a $1,000,000 civil penalty on each. Litigation Release No. 23960 (4 October 2017) reports a final judgment against the founder, which it says resolves the litigation in its entirety.

Claims about the cost base and the machines. SEC Litigation Release No. 25983 (24 April 2024) announced charges against Geosyn Mining, LLC, a Texas-based mining and hosting company, and two co-founders. According to the release, the SEC’s complaint alleges that between November 2021 and December 2022 the company raised about $5.6 million from more than 60 investors, telling them it would buy, maintain and operate mining machines and distribute what they mined for a fee. It alleges that the defendants falsely claimed favourable electricity contracts that let the company run the machines profitably, and failed to disclose to new investors that machines had never been bought or brought online for some earlier ones. It also alleges that about $354,500 was paid to investors as purported profit distributions although the company appears never to have operated profitably. These are allegations in a complaint, and no later SEC release on the case was read for this piece.

Mobile mining apps: where the balance lives

A phone app that offers to mine asks for no purchase, only a daily action, so the question moves from what is bought to where the number is kept. Apple’s and Google’s app stores publish rules on this. Apple’s App Review Guidelines, last updated 8 June 2026 as read on 3 October 2026, say in 3.1.5 that apps may not mine cryptocurrencies unless the processing is performed off device, with cloud-based mining as the example. Google Play’s Blockchain-based Content policy says it does not allow apps that mine cryptocurrency on devices and permits apps that remotely manage the mining of cryptocurrency. These are rules about what may be published, not observations of any app. By the stores’ own terms, a store-listed app is not meant to be hashing on the handset, which leaves a question for each one: if the hashing is not there, where is it, and what does the balance count?

A number inside an app is recorded wherever the operator records it, and whether the same entry exists on a public ledger is the first thing to settle. The Pi Network piece does that for one app against the project’s own documents and ledger.

An app that credits balances without charging for them has to be paid for by something. The candidates include advertising shown in it, data collected through it, growth driven by referrals, and a token whose terms are set later, and which applies is a matter for the app’s terms and privacy policy. The same Apple guideline says cryptocurrency apps may not offer currency for completing tasks such as downloading other apps or encouraging other users to download, so how an app words its referral reward is worth reading with that rule beside it.

For the balance to become value, four things have to be true, each with a party that controls it:

  1. A ledger. The balance has to be recorded somewhere other than the operator’s own database, on a chain whose contents anyone can read. Whoever controls the step that moves balances onto it controls who gets there and when.
  2. A float. Some of the units have to be free to move, and the amount has to be countable. The tokenomics piece traces how CoinGecko’s circulating figure rests on lists of locked addresses that the project supplies.
  3. A market. Buyers and sellers have to be able to trade the units at a size that matters, which depends on depth, not on a listing, so on the venues and on whoever quotes prices there.
  4. A redemption path. A right against someone to turn the units into something else. For a fiat-backed stablecoin that right is why the price holds; for a token whose issuer states none there is nothing equivalent, and only the issuer could grant one.

Where the first condition fails, the balance is whatever the operator’s records say, and the other three have nothing to attach to.

Questions that settle it quickly

  1. Can the hashrate be verified on a pool? Which pool, and can the buyer view the account behind the contract read-only, rather than only through the seller’s dashboard?
  2. Does the promised return fit inside the network’s? Divide the promised daily payout, in coin, by the hashrate sold and compare it with 47.8 satoshis per terahash per day, gross, as of 3 October 2026. A promise above it is being paid from some other source, the seller’s own funds or other buyers’ money, which is the arithmetic of the scheme described above.
  3. What happens to the contract if difficulty rises 30%? The terms should say who bears it. With a fixed fee the buyer does, and a fee above about 77% of the revenue at purchase exceeds revenue after the rise.
  4. Who holds the keys to the payout, and can it leave? A balance in a contract or an app is a claim on its operator, not funds in a wallet whose key the holder controls, which makes how wallets fail a description of a different position. What applies here is in the terms: withdrawal minimums, pause rights, and whether customer assets are kept apart from the operator’s own.
  5. What pays for it? For a contract, the answer should be block rewards and fees at a cost the seller can show. For an app, advertising, data, referrals or a token; if its terms do not say, the question is still open.