How to check what a miner promised against what it delivered

A promise in a filing has three parts: a number, a date and the words around them.

Public companies say what they expect to do. A miner says it will reach so many exahash by a date, energize so many megawatts, or report revenue in a range. Those sentences are forward-looking statements, and every filing that carries one also carries a paragraph saying it may not come true. The paragraph is not the interesting part. The interesting part is that the sentence can be checked later, against the same company's own filings, without any opinion entering.

A statement can be checked only if it has a number and a date. "We expect significant growth" has neither, and this site's specification says such statements are recorded but not scored. "We are targeting 75 exahash by the end of 2025" has both, and it scores. In between sits the directional statement, a direction with no number, like "we expect to continue increasing our computing power": it can be checked for direction but not for size. CleanSpark wrote that sentence in its annual report filed Nov 25, 2025, when its average hashrate for September 2025 was 45.6 exahash; its quarterly report filed Aug 6, 2026 gives 42.6 exahash for June 2026. The sentence promised no figure, so the ledger prints no percentage for it.

A statement can also have both a number and a date and still not be a promise: if the result was already in hand on the day it was written, it reports and does not promise. The ledger shows such a statement, marks it a reported result, and never scores it. A later section gives the rule and its examples.

Where the promises live.

Four places, in order of formality. The annual report on Form 10-K, in the business section and the discussion of results, where a company describes what it is building and what it expects the building to produce. The quarterly results release, filed as exhibit 99.1 to a Form 8-K, where a quarter's numbers come with expectations for the quarter after it. The monthly production update, another exhibit 99.1, or an exhibit to a Form 6-K for a company organized outside the United States, where hashrate targets are restated month by month. And the investor deck, usually exhibit 99.2, where the targets are drawn as bars and the dates are printed under them.

The outcome lives in the same places, one or more periods later. That is what makes the check possible: the company that said 75 exahash by December is the company that reports its December hashrate in February. The ledger this article draws on records each promise with the document it came from and the document that answered it. Each quote on the ledger was searched in the cited document as an exact string, and the ledger marks a row verified only when the quote was found there. Where a row stores the sentence that reports the outcome, that sentence was searched in the outcome document the same way; a row that stores none has an outcome the ledger did not string-check.

Pin the number and the date, then read the outcome in the company's own words.

On Jul 1, 2025, with its June 2025 production update, MARA stated that it was targeting 75 exahash by the end of 2025. That is a number, 75 EH/s, and a date, Dec 31, 2025. The filing that answers it is the fourth-quarter shareholder letter filed as an 8-K exhibit on Feb 26, 2026, which reports 66.4 EH/s energized at year end and says, in the company's words, "While below our 75 EH/s target, this outcome reflects deliberate capital discipline."

Three things to notice. The company said the number itself, so no reading between lines was needed. The company gave its reason, and the ledger records the reason next to the shortfall rather than replacing it. And the variance is a fraction with the promise as its base, which is what lets a promise of 75 exahash and a promise of 1.7 exahash sit on the same chart.

Six hashrate promises from five companies, scored the same way.

Hashrate is the most common promise a miner makes, and the easiest to check, because every miner reports it every month or every quarter. Here are six on file whose outcome is in a later filing: four had a date that has passed, and for two, CleanSpark's and Argo's, the company gave no date and the later filing shows the figure reached or beaten. Each is drawn the same way: the figure promised beside the figure delivered.

The same six, with the variance as a share of the promise. A plus sign is more than promised.

The same six, with the variance as a share of the promise. A plus sign is more than promised.

Company	Promised	Delivered	Variance
MARA	75 EH/s by Dec 31, 2025	66.4 EH/s	minus 11.5 percent
CleanSpark	50 EH/s	50 EH/s, June 2025	0.0 percent
Bitdeer	50 EH/s by Dec 31, 2025	55.2 EH/s	plus 10.4 percent
Bitdeer	40 EH/s by Oct 31, 2025	41.2 EH/s	plus 3.0 percent
HIVE	25 EH/s by Nov 27, 2025	25 EH/s	0.0 percent
Argo	1.7 EH/s	up to 2.4 EH/s	plus 41.2 percent

HIVE and Argo are the two rows here that were read from company dossiers. The ledger searched their quotes in the cited filings and found each word for word, so both are verified. The ledger stores no outcome sentence for either, so it did not string-check their outcomes; a separate search of the filings that answer them on Oct 2, 2026 found the outcome wording of each, word for word.

One row needs its footnote. Argo's 2.4 is what its fleet "can generate", a capacity, set against a go-forward expectation of 1.7; the two are close cousins, not twins, and the ledger says so.

One promise is left out of the table on purpose, because the ledger has no percentage for it. On Mar 27, 2025, Bitfarms (now Keel Infrastructure) named three targets in one sentence, all due by Jun 30, 2025: 18 EH/s operational, 21 EH/s installed and 19 w/TH installed. The document that stated them already reported 18.6 EH/s under Management, and the company's second-quarter filing says it reached the 18 EH/s operational and 19 w/TH installed targets in March 2025. The ledger row for that sentence covers only the 21 EH/s installed target. In August the company said that target was no longer feasible on its original timeline, so the row is marked not delivered. It carries no percentage: the 17.7 EH/s the company reported for Jun 30 is operating hashrate, taken after power to its Argentina site was halted, and its second-quarter documents give no installed figure for that day. The other two targets are not in that row.

A result already in hand is a report, not a promise.

Not every sentence with a number and a date looks ahead. On Apr 14, 2026, TeraWulf said its first-quarter revenue would be between $30 million and $35 million. The quarter had closed two weeks earlier, so the company was giving preliminary results for a period that was over, not predicting one. The final figure, $34.0 million, fell inside the range, but by then the quarter's revenue had been earned and only the closing of the books was left. The same holds for a target the company has already reached. IREN's update of Jul 7, 2025 says it had reached its 50 exahash target by the end of June, so that sentence reports a result, and it is not in the table above.

The ledger scores a statement only if its target had not yet been met and its period had not yet ended on the day it was made. A statement that fails that test is still on the ledger, marked a reported result, but it is never scored, because a result already in hand can be neither a promise kept nor a promise missed. Eight statements carry the mark: preliminary revenue for a period that had closed, two quarters from TeraWulf and a quarter, nine months and a full year from Big Digital Energy; IREN's 50 exahash; Hut 8's list of four new sites it had already announced; and DMG's 75 megawatt utility approval. The last two were never scored.

Most hashrate promises landed within 12 percent, and the megawatt promise fell far short.

The ledger holds other kinds of promise, and the distance between the promise and its date matters as much as the kind. A megawatt figure depends on a substation, a utility and a construction crew.

Two rows in this chart were read from company dossiers: Argo's go-forward hashrate and HIVE's 25 EH/s, both from the table above. The ledger found each quote in its filing word for word, and the separate search on Oct 2, 2026 found the outcome wording of each in the filing that answers it, word for word.

The site's specification weights a promise by how specific it was and how far ahead it looked, for this reason: hitting a figure announced two weeks early, as Bitdeer did with its 50 exahash, is not the same feat as hitting one announced five months early, as HIVE did with its 25 exahash, and a ledger that counted them the same would flatter whoever makes the shortest promises.

Three kinds of miss that are not the same thing.

A shortfall in quantity is the plainest: Bitdeer said 221 megawatts at Massillon by the first quarter of 2026 and reported 121 online at the end of it. A slip in timing keeps the quantity and moves the date: Cipher's 2025 annual report expected the Stingray site energized in the first half of 2026, and its first-quarter deck aligned the work with a fourth-quarter energization; MARA's Long Ridge plant was to reach 505 megawatts nameplate in the second half of 2026 and the second-quarter 10-Q says the first quarter of 2027. A withdrawal replaces the promise with a different plan: Bitfarms (now Keel Infrastructure) said in August 2025 that its 21 exahash installed target was no longer feasible on its original timeline, and that it had no plans to raise hashrate beyond the 17.7 exahash it was operating.

The ledger records each with the company's own sentence about it, because the sentence is the fact. What the reader does with a moved date is the reader's business; what this site does is keep the first date next to the second so the move is visible at all. A beat gets the same treatment: Bitdeer's 55.2 exahash against a promised 50 is recorded as plus 10.4 percent, not as praise.

Check any promise in five steps.

1. Copy the sentence from the filing, word for word, with the filing's date. A paraphrase cannot be verified later; a quote can.
2. Write down the number and the date it names. If either is missing, note the statement as directional or vague and do not score it. If the result was already in hand on the day it was written, because the period had ended or the target was reached, note it as a reported result and do not score it.
3. Find the first filing after the date that reports the same measure, in the same unit and on the same basis: energized against energized, installed against installed, self-mining against self-mining.
4. Subtract, and divide by the promise. Copy the company's own sentence about the outcome next to the result.
5. Compare with the ledger, which holds every row above with both documents, and with the deals article for the megawatt promises that arrive inside a lease.