An enterprise value of $7.5 billion resting on a lease whose first rent is due in 2028

01Snapshot

Key figures, each tagged by provenance. Where a figure is computed here, the first chip in the row is its own tag and the arithmetic follows in the basis.

02Business model

Riot Platforms owns and operates large-scale power infrastructure in Texas and Kentucky and monetizes it three ways. It mines bitcoin for its own account, taking the coin as revenue. It leases developed data center capacity to third parties under long-term contracts. And it sells electrical engineering, switchgear and power distribution equipment through two acquired businesses, ESS Metron and E4A Solutions. The company reports these as three segments: Bitcoin Mining, Data Center and Engineering.

The unifying asset is interconnected power with the approvals already in hand. Developed power capacity was 1,292 MW at June 30, 2026, across Rockdale at 700 MW, Corsicana at 400 MW and Kentucky at 192 MW. What has changed in 2026 is not the asset but the customer: capacity that used to be consumed internally by miners is now being let to tenants under contracts measured in decades rather than in blocks.

The payment mechanics differ sharply between the three. Mining revenue is received in bitcoin at the moment it is earned, so it is non-cash on receipt and its value moves before it can be realized. Data center revenue is contractual rent, billed and paid in dollars, and it does not begin until capacity is delivered. Engineering revenue is project work recognized over time, and it carries contract assets of $19.9 million at June 30, 2026, of which $12.5 million is engineering and $7.4 million tenant fit-out.

03Revenue streams

Each stream, with its share of the total and its durability. The three-month columns are the clearest view: the second quarter of 2026 is the first full quarter in which all three segments produced revenue.

Total revenue rose 13.9% while the largest stream fell 19%. That is the whole of the business story in one line. Bitcoin Mining lost $27.2 million of quarterly revenue year on year and the two other segments added $48.4 million between them. On a six-month basis the same pattern holds: mining fell from $283.7 million to $225.6 million, Data Center contributed $56.4 million from nothing, and Engineering more than doubled from $24.5 million to $59.5 million.

The mining decline is not a volume failure. Quantity of bitcoin mined rose to 1,587 coins in the quarter from 1,426, and deployed hash rate rose to 44.4 EH/s from 35.4 EH/s. What fell was the value of each coin produced: $71,667 against $98,800. The company mined more and was paid less for it, which is the network economics rather than an operating problem.

Data Center revenue at $23.2 million in the quarter comes entirely from the AMD lease, since the 191 MW tenant lease was signed after the quarter ended and delivers nothing before December 2027. The forward step from $23.2 million a quarter to a lease expected to yield approximately $9.1 billion over twenty years is therefore not a growth rate to extrapolate; it is a different contract that starts in 2028.

04Industry & market backdrop

Two markets bear on this company and they are moving in opposite directions.

Bitcoin mining economics compressed over the year to June 2026. The production value of one bitcoin mined fell 27.5%, from $98,800 to $71,667, while the cost of power for self-mining rose from $62.2 million to $73.5 million in the quarter. Riot's combined all-in power cost was 3.6 cents per kilowatt hour in the quarter against 3.5 a year earlier. A miner with a low power cost stays profitable across a wider band of bitcoin prices, and Riot's position on that curve is what has allowed it to keep expanding hash rate while the value of each coin produced fell.

Demand for large blocks of near-term interconnected power is the opposite story. The scarce input is not land or capital but energized capacity with grid approvals already granted, because a project still in an interconnection queue cannot commit to a delivery date. Riot has leased 241 MW of critical IT load at Rockdale inside roughly seven months, 50 MW to Advanced Micro Devices, Inc. with an option to expand to 200 MW, and 191 MW to an unnamed counterparty. That is the market repricing capacity that already exists.

Texas is the venue for both, and its rules are being rewritten. The company's own risk disclosure states that Senate Bill 6 requires the Public Utility Commission of Texas to amend wholesale transmission cost-allocation rules by December 31, 2026, and that the Commission is considering measures requiring large loads to bear a greater share of transmission upgrade costs, to pay minimum demand charges on contracted peak demand for twenty years, and to move away from the four coincident peak allocation method. It also states these actions could delay or prevent interconnection of projects still advancing through the ERCOT process.

05Competitive position

Named peers on comparable metrics. The metrics below are Riot's own filed figures; where a peer figure is not carried here, it is because no peer filing was read for this document and the gap is named in the methodology rather than filled from memory.

Riot competes on two fronts that have different competitor sets, and conflating them is the usual error. In bitcoin mining it competes against every other operator for the same fixed block subsidy, so a rival adding hash rate reduces Riot's share of a total that does not grow. In data center leasing it competes for tenants against operators of interconnected capacity, where a rival's new site does not take revenue Riot already has under contract.

The vertical integration is the differentiator most likely to be underweighted. An operator building 191 MW of critical IT load on a two-year schedule is exposed to switchgear and transformer lead times, and Riot owns two businesses that make that equipment. The interim facility exists specifically to fund long-lead equipment procurement, which is the same constraint seen from the financing side.

Against that, the balance sheet is a competitive variable and not a favorable one here. $843.0 million of debt, of which $254.6 million is current, and a treasury half pledged, is a weaker position from which to bid for the next tenant than an unlevered balance sheet would be.

06Financial performance

Multi-period, in thousands of dollars, with the trend stated. Years ended December 31.

Filed 10-K for the year ended December 31, 2025, filed March 2, 2026. There is no Data Center line in any of the three years: the segment first appears in 2026.

The 2026 half year

Filed 10-Q for the quarter ended June 30, 2026, filed August 10, 2026. Diluted per-share figures for Q2 2025 were $0.58 on 382,677,364 shares; the 2026 periods are loss-making so basic and diluted counts are equal.

The cash statements say something the income statement does not. In the six months to June 30, 2026 operating activities used $272.8 million, investing activities provided $514.8 million, and financing activities used $2.9 million. The investing inflow is principally $131.8 million from selling bitcoin and $14.7 million from marketable securities, against $93.2 million spent on property and equipment and $86.4 million on miners. Cash and restricted cash rose to $548.9 million from $309.8 million.

So the company funded a $272.8 million operating outflow and roughly $180 million of capital spending by selling assets, not by raising money. No cash was raised through the issuance of common stock or debt in the period. Whether that is prudence or necessity is not settled by the filings; what is settled is that the treasury is the funding source and it is finite.

07Capital structure & dilution

Shares, preferreds, capacity, warrants and convertibles, at June 30, 2026 unless stated.

The share count fell, and the reason is not what it looks like

Three counts appear in the same document and they are all correct: 378,022,964 issued and outstanding at June 30, 2026, 375,258,935 on the cover at August 7, 2026, and 347,695,985 as the weighted average for the quarter. The first two are point-in-time counts five weeks apart and the third is an average over three months, which is why it is the smallest.

The fall of 2,764,029 between the quarter end and the cover date is the ordinary mechanism rather than an event. The equity rollforward shows the same movement inside the quarter: 378,978,341 shares at April 1, 2026, less 955,377 for issuance of restricted stock net of forfeitures and delivery of shares underlying awards net of tax withholding, giving 378,022,964 at June 30. The count falls when forfeitures plus shares repurchased to satisfy employee tax withholding exceed new awards, and the cash flow statement records $2.5 million of such repurchases in the half. There is no buyback authority and no reverse split anywhere in the filings, and the specific post-quarter movement is not itemized because it falls outside the period the report covers.

The direction of travel over three years

Weighted average shares went 175,026,051 in 2023, 275,980,010 in 2024 and 340,719,298 in 2025, an increase of 95% in two years. Set against that, no cash at all was raised from stock or debt in the first half of 2026 and the $500.0 million program stands untouched. The recent restraint is real and it is six months old; the three-year record is the base rate.

08Valuation

Multiples shown with their basis and pricing date. This document derives no target price and issues no rating.

The last row of that table is the valuation question in one figure. The whole enterprise is priced at roughly four fifths of the undiscounted base rent from a single lease that has not started, before any value is attributed to 44.4 EH/s of mining, an engineering business running at $37.3 million a quarter, the AMD lease, or the 1,051 MW of developed capacity not committed to the tenant lease. Whether that is cheap depends entirely on two things the filings do not settle: whether the lease is delivered on schedule, and whether the counterparty pays.

Third-party coverage

Every row is market data: attributed, dated, tagged, never adopted. Price at publication is the close on the last trading day before the action, which governs every row in this table. That convention is used because it is what the author could see when the target was struck: a same-day close is set after the note is out and can move because of it.

The price at publication of $19.40 is the Nasdaq close of August 10, 2026, the last trading day before the note. The firm's own stock data box shows $19.37, which is neither the August 10 close nor the August 11 close of $20.24, so it is an intraday print rather than a session close and is not used here.

The shape of the coverage

One firm is established, and that figure is a floor rather than a count of the coverage set. Riot's investor relations coverage page was not read and no aggregator was queried, so how many houses follow this stock is not established. With one observation there is no spread to report and no consensus to attribute, and none is struck here: an average of one number is that number wearing a different label.

What the firm discloses about this issuer is not established. Its disclosure page carries a coverage-universe base rate, that 34 of its 82 Buy-rated names are investment banking clients, or 41%. That says nothing about whether this issuer is one of them, and the only use a reader could make of it is an inference this document declines to draw. The cell reads Not established rather than being filled from the base rate.

Two further facts about the row are worth stating because they change how it should be read. The $29.00 target is not new: the firm's own rating and target history shows it set on October 13, 2025 and carried since, so the August 11, 2026 note reiterates rather than revises it. And the target implies a return of roughly 50% against the price at publication, which is a claim by that firm and is reported here as market data and adopted in no part of this analysis.

09Growth drivers

Ranked by how much of the current enterprise value each would have to justify.

1. Delivery of the 191 MW tenant lease. Approximately $9.1 billion of base rent over twenty years from full deployment in June 2028, with two five-year renewals at the lessee's option. Phased delivery of 96 MW in December 2027 and the balance in June 2028. This is the driver: nothing else on this list is of comparable size, and nothing of it reaches revenue for almost two years.

2. Expansion of the AMD lease. The existing agreement covers 50 MW of critical IT load with the potential for expansion up to a total of 200 MW at Rockdale. The expansion is an option and not a commitment, and the filings state no contracted value for either the base lease or the expansion. What it has already demonstrated is that the capacity is lettable to a named, investment-grade counterparty, which is a fact the unnamed tenant lease does not supply.

3. The remaining uncommitted capacity. Developed power capacity is 1,292 MW; the two Rockdale leases account for 241 MW of critical IT load. Critical IT load and developed power are not the same measure, since the second includes the overhead the first consumes, so the residual cannot be computed from these two figures alone. What can be said is that a substantial majority of developed capacity is not under a data center lease and is currently used for mining or is undeployed.

4. The engineering business. Revenue rose from $10.6 million to $37.3 million in the quarter and from $24.5 million to $59.5 million in the half. It is the fastest-growing segment in percentage terms and the one least discussed, and it has a strategic function beyond its revenue: ESS Metron and E4A Solutions supply the switchgear and power distribution that the Rockdale build needs, at a moment when equipment lead times are the binding constraint on delivery.

5. Mining efficiency. Hash rate rose to 44.4 EH/s from 35.4 EH/s year on year and Kentucky's developed capacity nearly tripled to 192 MW from 65 MW. This grows the smallest multiple rather than the largest: mining is 65.3% of current revenue and the segment whose revenue per unit of output fell 27.5% over the year.

10Risks

Severity-ranked, most severe first.

11Bull / base / bear

Each case with its preconditions: what must be true, not what might be.

12Research summary

Riot has converted power into contracted revenue faster than any figure in its accounts yet shows. Inside roughly seven months it let 241 MW of critical IT load at Rockdale across two counterparties, and the larger of the two leases carries approximately $9.1 billion of base rent over twenty years against total 2025 revenue of $647.4 million. The Data Center segment did not exist a year ago and produced $23.2 million in the second quarter.

None of that is in the reported numbers, and will not be until 2028. The lease term begins on full deployment anticipated in June 2028, with the first 96 MW in December 2027. Until then the company reports a shrinking mining business, a fast-growing engineering business, one small lease, and a loss driven substantially by the bitcoin mark: $737.6 million for the first half, of which most is the fair-value mechanism rather than trading.

The gap between those two facts is a financing problem, and it is the whole of the near term. The only committed money is a $573.0 million interim facility at a project subsidiary which the company itself dates two ways, October 15 or December 31 of this year. Everything beyond it is absent from the record: no permanent facility, no backstop, no commitment. Meanwhile the treasury that has been funding operations fell 37% in six months and over half of what remains is pledged.

Two things a reader should not take from this document. The first is a view on the counterparty: the filings say a leading frontier AI lab and name nobody, so no assessment of that credit is possible from the public record and none is offered. The second is a conclusion about value: the enterprise is priced at roughly four fifths of the undiscounted base rent from one lease that has not started, and whether that is cheap turns on delivery and payment, neither of which any filed document settles.

The tests that would settle the reading are stated in the catalyst calendar for this ticker as falsification tests, each with a threshold that can be read off a filing.

13Null categories

Requested categories with nothing to report. Status reads Covered above, None where a named source was checked and found empty, or Not established where the category was not examined. A category nobody looked at is a gap, not an absence.

14Methodology & sources

Pricing basis, the tagging scheme, and, importantly, what was not checked.

Pricing basis

Every price-dependent figure is struck on the August 14, 2026 close of $19.015, which was the last completed session: August 15 and 16 fall on a Saturday and a Sunday, so no later close exists. The trading calendar is taken from the exchange's own daily record rather than from a weekday rule, over the fifteen sessions from July 27 to August 14, 2026, with no weekday absent from the series.

The half cent is load-bearing and is stated rather than rounded. On 375,258,935 shares the capitalization is $7,135.5 million at $19.015, $7,137.4 million at $19.02 and $7,133.7 million at $19.01, a spread of $3.75 million. A reader given a rounded price could not reproduce the figure. The close was corroborated against a second, unrelated provider: fifteen of fifteen overlapping sessions agree, and August 14 agrees to six decimal places.

The share count is the cover figure of the quarterly report, 375,258,935 as of August 7, 2026, which is seven days before the price. That is the closest count available and the gap is stated rather than smoothed.

Provenance tags

Sources

What was not checked

An unexamined area is a gap, not a clean bill. The following were not reached.

• Any filing made on August 17, 2026. The submissions record behind this document was retrieved on August 16, 2026, so the day this document is stamped is outside the sweep and nothing filed on it is covered.
• The credit agreement filed as an exhibit to the current report of August 14, 2026. It is the document that would settle whether the interim facility matures on October 15 or December 31, 2026, and both dates are reported here because it was not read.
• The tenant lease and the AMD lease. Their terms are as described in the quarterly report, and neither agreement was read. No contracted value for the AMD lease is stated in any filing read.
• The identity of the Rockdale tenant and any credit support behind its obligations. Nothing in the filings names or rates the counterparty.
• Peer filings. No competitor's report was read, so section 5 carries no peer figures.
• Riot's investor relations coverage page and any market data aggregator, so the size of the analyst coverage set is not established and the single row in section 8 is a floor.
• The 133 beneficial ownership filings on Schedule 13D and 13G and their amendments, the most recent dated August 6, 2026, and every Form 4. No holder and no insider transaction is identified.
• Three current reports carrying Item 5.02 in 2026, filed January 2, January 5 and June 15, so no management change is established. The June 15 report also carries the 2026 annual meeting voting results.
• What this registrant did between 2002 and 2017, when it was a diagnostics business under earlier names. None of those filings was read, and nothing here describes that period.
• Court dockets, ERCOT interconnection queue records, Public Utility Commission of Texas docket filings, and county and municipal permitting records for Rockdale, Corsicana and the Kentucky site. Every megawatt figure here is the company's own statement in a filing.
• Exchange short interest reports, credit rating agency actions and index provider announcements, each recorded in the null table above.

Filing status

Riot Platforms has filed fifteen late-filing notifications across the 1,592 filings on its index: eight on Form NT 10-K and seven on Form NT 10-Q, the earliest on November 15, 2002 and the most recent on March 2, 2023. Nine fall after October 19, 2017 and six belong to the earlier company. None has been filed in the three years and five months since. The whole index was searched on August 16, 2026 across both of its pages, the recent page holding 1,005 filings back to January 25, 2013 and an overflow page holding 587 more back to February 7, 2002.

That pagination is why the count is stated with the index size behind it. A search of the recent page alone would have covered 63% of the record, missed six of the fifteen notifications, and produced a figure that looked complete.

An earlier identity, and the date it ended

The Central Index Key long predates the business. Filings begin on February 7, 2002, more than a decade before any bitcoin operation, because the registrant was incorporated in Colorado on July 24, 2000 under the name AspenBio, Inc. The company states the transition in terms: "Effective October 19, 2017, we adopted the corporate name Riot Blockchain, Inc., and changed our state of incorporation to Nevada." It adopted its current name on December 30, 2022 and remains a Nevada corporation.

Two things follow that a reader would otherwise get wrong. An eighteen-filing history of annual reports is not eighteen years of this company: it is roughly nine years of this one and fifteen of a diagnostics business that happened to keep the same registration. And the state of incorporation changed with the identity rather than separately from it, so every filing before October 19, 2017 was made by a Colorado corporation and every one after it by a Nevada one.

15Document log

Newest first. The original build entry is never removed or rewritten.