Cipher Mining
A former bitcoin miner in Texas that now builds data centres and rents them, whole, to Amazon and other AI buyers on ten- to twenty-year leases.
The scores
Growth1/4 ↑
Filed revenue is all bitcoin mining and fell 43% on the year as rigs and joint ventures were sold; no lease revenue has printed yet.
Business quality3/4 →
Approved Texas interconnections won Amazon and a Google-backed tenant, but rivals sign the same tenants and owners build most capacity themselves; untested by a cycle.
Financial position2/4 ↓
The three campuses are funded, but the $359.3M overrun takes 43% of corporate cash and the pipeline needs money not yet raised.
Execution record2/4 ↓
One hall two months early, but Barber Lake moved a quarter with $359.3M of cost after twice being called on schedule.
Per share capture2/4 ↓
Shares up 9% in a year, 204.7M potential shares outstanding, convertibles near the price and a warrant top-up that grows as the price falls.
Potential3/4 →
A bounded upside that is committed: 700 MW signed, financed and dated. Not a 4 - the pipeline beyond waits on Texas regulators.
Risk profile2/4 ↓
Three tenants and one state hold everything, and the pipeline turns on a single external step: ERCOT's audited batch, report due December.
Valuation3/4 ↑
11.4 times contracted NOI prices the signed rent roughly as described and nothing for the pipeline. Not a 4 - one hall bills today.
Evidence quality3/4 →
Leases, debt, warrants and potential new shares are filed facts; the NOI and contracted totals are presentation figures graded, not verified.
If it works 2.0× · if it fails you keep about 2%.
The Verdict
We rate Cipher neutral. A former bitcoin miner, it now builds data centres in Texas and rents them whole to Amazon and a Google-backed tenant on long leases. The business is weak for now: the first campus slipped with a bill attached, and growth beyond the signed sites waits on new leases. The price pays roughly what the signed rent is worth after head-office costs, so it is fair, not cheap. As a bet it is fragile: if more sites lease on the same terms, the shares might not quite double; if leasing fails, the lenders come first and almost nothing is left. Little to win, much to lose.
The long caseAI needs new buildings with power already connected, and Cipher holds sites the grid has approved; filling them with long leases turns that power into steady rent.
The evidence under this - 8 claims
The 10-K's risk factors say every tenant guarantee takes effect only after rent commencement and is capped, and that a termination for construction delay would not trigger it; the $5.5bn is the company's own figure, not a filed schedule.
- The NOI is management's (C03) and only one hall pays rent today, so the denominator is a promise rather than a filed figure.
- Netting all $4.56bn of cash flatters the multiple only if the restricted cash finishes the campuses with nothing left over; the company says each is funded to completion, but Barber Lake already needed $359.3M more.
The same 477 MW was 'in discussions for HPC hosting lease' in the May presentation and no lease followed in the five months to October, while TeraWulf and Applied Digital signed more capacity with the same class of tenant (C38, C39).
The 8-K calls the changes 'tenant-modification change orders' and the tenant reimburses 50% of costs above $359.3M with a return, so part of the slip is the tenant's own redesign rather than a construction failure.
The 161.9M is the maximum the indentures could deliver; the 2031 notes carry capped calls bought to offset dilution, and Cipher may settle conversions in cash.
The company
Cipher Digital finds land with a grid connection, builds a data centre on it, and rents the whole building to one tenant for ten years or more. The tenant brings its own computers; Cipher supplies the power, the cooling and the shell.
Buyers come to it because powered space is scarce: under 1,500 megawatts was free across the main U.S. markets in mid-2026, about six months of demand. The alternative is to build it themselves, which the largest buyers do for almost 60% of their space.
The evidence under this - 3 claims
How it makes money
Today almost all revenue still comes from mining bitcoin at one Texas site, Odessa: $24.8 million in the second quarter of 2026, falling as the business is wound down.
The new money is rent. The three leases add up to about $887 million a year once every building is finished, of which about $793 million should be left after running costs. The first hall started paying rent in August 2026.
The evidence under this - 5 claims
Rents escalate, so a straight average overstates the first years and understates the last; the NOI path in C03 rises from $686M in 2027 to $894M in 2035.
- No filing carries a lessor payment schedule yet: the 10-Q shows no lease revenue, and the figure is a presentation number nobody outside the company has checked.
- Barber Lake's first rent has already moved from October 2026 to a phased start from the fourth quarter (C04), so the 2026 and 2027 figures are older than the schedule they rest on.
The CEO said on the call that the rest of Black Pearl's schedule is unchanged and that the acceleration was paid for under a cap on expenses, so one early hall says little about the remaining phases.
Where the money comes from, stream by stream
| Stream | Share of sales (FY2025 and Q2 2026) | How it earns it |
|---|---|---|
| Bitcoin mining | 100% | Odessa's 207 MW mines bitcoin on cheap contracted power; revenue is bitcoin earned, valued when received, and is being wound down as sites convert. |
| Data centre leasing | 0% filed; first rent August 2026 | Ten- and fifteen-year leases of powered data halls to Amazon, Fluidstack and a hyperscaler; Black Pearl's rent began in August. |
| Power sales | not in revenue ($7.9M netted in costs) | Excess Odessa power sold back to the Texas grid through Luminant; recorded against cost of revenue rather than as revenue. |
Who runs it
Tyler Page founded Cipher in 2021 and still runs it; a former banker became finance chief in October 2025. Insiders own about 3%, most of it Page's, while the largest holder, Bitfury's group, has been selling. Bonuses ride on megawatts powered, built and signed; the share awards ride on the stock price. Nothing pays on rent actually collected or on value per share.
The evidence under this - 3 claims
V3 Holding has sold since: insider_fetch.py shows discretionary open-market sales, mostly V3's, in the year to October 2026 (C29), so the Bitfury stake is likely below 15.11% now.
Why it wins, and who could beat it
Why customers come back
What Cipher holds is a place in line. Texas's grid operator is tracking more than 438,000 megawatts of requests to connect, and Cipher's three leased sites already have their connections approved.
That is why Amazon, a Google-backed tenant and a large cloud company signed long leases with a company that had never finished a data hall, and why the Barber Lake tenant has now committed to twenty years. It buys a seat at the table, not pricing power.
The evidence under this - 4 claims
- The advantage is a queue position, not a technology: TeraWulf and Applied Digital signed the same tenants on the same terms (C38, C39), so it buys a seat at the table rather than pricing power.
- ERCOT's batch process and the Governor's letter can re-rank queue positions by rule, which is the one thing that could take the input away (C21, C22).
The new lease is a commitment to sign a lease that begins about 2036, so none of its rent falls inside any horizon this thesis uses, and the $5.2bn is the company's expectation.
The threat
The biggest buyers mostly build their own data centres, and rivals sell the same thing: TeraWulf rents space to Fluidstack, Cipher's own tenant, and Applied Digital signs 15-year leases with the same kind of buyer. Cipher's answer is speed and long terms, not lower rent.
The evidence under this - 4 claims
The same 477 MW was 'in discussions for HPC hosting lease' in the May presentation and no lease followed in the five months to October, while TeraWulf and Applied Digital signed more capacity with the same class of tenant (C38, C39).
The rivals: AMZN, GOOGL, MSFT, META, WULF, APLD, MARA, RIOT.
Who else does this, rival by rival
| Ticker | Rival | Where it competes | How far along | Priced against it |
|---|---|---|---|---|
| AMZN · GOOGL · MSFT · META | Amazon, Google, Microsoft and Meta building their own capacityPriced against it: Different product | HPC data center leasing | On sale | Different product |
| WULF | TeraWulf Inc.Priced against it: Same product | HPC data center leasing | On sale | Same product |
| APLD | Applied Digital CorporationPriced against it: Same product | HPC data center leasing | On sale | Same product |
| MARA · RIOT | MARA Holdings, Riot Platforms and the other listed miners Cipher used to name as its competitorsPriced against it: One segment only | Bitcoin mining | On sale | One segment only |
How big it could get
How big this could get
Demand is running hot. North American data centres absorbed 25 gigawatts in the first half of 2026, double a year earlier, and 95% of what is being built was spoken for in advance.
But in Texas the limit is now the grid, not the buyers: the grid operator's first group of applicants is being audited, with a report due in December, and no date yet for the study after it.
The evidence under this - 2 claims
Who has signed
Amazon signed for 300 megawatts at Black Pearl for 15 years, with Amazon's parent guaranteeing the rent. Fluidstack took 300 megawatts at Barber Lake for 10 years, with Google backing part of it, and an unnamed AI lab committed to 10 more after that. A third large cloud company signed for 100 megawatts at Stingray for 15 years.
The evidence under this - 4 claims
The 10-K's risk factors say every tenant guarantee takes effect only after rent commencement and is capped, and that a termination for construction delay would not trigger it; the $5.5bn is the company's own figure, not a filed schedule.
The new lease is a commitment to sign a lease that begins about 2036, so none of its rent falls inside any horizon this thesis uses, and the $5.2bn is the company's expectation.
The backstop was described as rising to about $1.73bn when the lease was upsized, which no Cipher filing read here states, and nothing says it covers the ten-year commitment that follows.
Where the growth is coming from, business by business
Where the growth is coming from
| Business | Share of revenue | Growth: last quarter | Last half-year | Last year |
|---|---|---|---|---|
| Bitcoin mining | 100% | -43.0% | - | - |
| Whole company | 100% | -43.0% | - | - |
The filed split through 2026-06-30. These rates are not carried forward: the reading below says why.
How we read it: Neither reading: the blend projects a mining line being wound down. The score believes the filed decline; the lease business has no revenue history to project.
The results
The numbers
Sales grew 20% over the twelve months to Jun 2026, slower than the 48% in the year to Dec 2025.
Operating margin is −278%, down from −16% in the year to Dec 2023.
Running costs take 334¢ of each sales dollar, down from 1308¢ in the year to Dec 2022.
Shareholders' money lost 171% a year, down from −11% in the year to Dec 2022.
It used $1.65B of cash over the last twelve months, so the $832M it holds lasts about 0.5 years at that pace.
It owes $4.80B more than the cash it holds, and its operations bring in no cash to repay it.
The share count rose 65% in 3.5 years, and pay in shares costs 47% of sales.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-04. Fiscal years end in December.
| Dec 2022 | Dec 2023 | Dec 2024 | Dec 2025 | 12 mo to Jun 2026 | |
|---|---|---|---|---|---|
| Size and growth | |||||
| Sales | $3.04M | $127M | $151M | $224M | $191M |
| Sales growth | - | +41× | +19% | +48% | +20% |
| Profit | |||||
| Gross margin | 75% | 60% | 59% | 64% | 56% |
| Operating margin | −12× | −16% | −29% | −188% | −278% |
| Net margin | −13× | −20% | −30% | −367% | −586% |
| What growth costs, per sales dollar | |||||
| General and admin | 23× | 22% | 22% | 16% | 24% |
| Pay in shares | 14× | 30% | 28% | 24% | 47% |
| Capital spending | 13× | 16% | 92% | 218% | 731% |
| Cash | |||||
| Cash from operations | −$20.9M | −$94.2M | −$87.5M | −$208M | −$256M |
| Free cash flow | −$60.1M | −$115M | −$227M | −$696M | −$1.65B |
| Balance sheet | |||||
| Cash and short-term investments | $11.9M | $74.2M | $5.58M | $628M | $832M |
| Debt | - | $0 | $32.3M | $2.79B | $5.64B |
| Net cash | - | $74.2M | −$26.7M | −$2.16B | −$4.80B |
| Returns and shares | |||||
| Shares (diluted) | 248M | 252M | 323M | 382M | 409M |
What growth costs
It is too early to say whether growth will make Cipher more profitable. Today's sales come from bitcoin mining, which is shrinking, while the rent has barely started, so costs look huge against sales. Office costs and pay in shares are rising; the bet is that they stay small next to the rent once all three campuses pay.
The evidence under this - 2 claims
Part of today's compensation is a development team building three campuses at once; some of it may be capitalised into projects or fall once construction ends, so the run rate may overstate the steady state.
The last report (4 Aug)
Quarter ended June 2026, filed figures.
The second-quarter 2026 report, out on 4 August, showed a business between two lives. Revenue was $24.8 million, all from mining and down from $43.6 million a year earlier, and the loss was $267.5 million, more than half of it an accounting charge on Google's warrants.
The news that mattered came with it: the first Black Pearl hall was handed over two months early and started paying rent. Then, on 24 September, Barber Lake's dates moved: halls now arrive from late 2026 into early 2027, and Cipher pays the first $359.3 million of extra cost.
What it did to the thesisIt proved the landlord model can work: the first Black Pearl hall was handed over early and now pays rent. The later Barber Lake delay and extra cost cut the other way, which is why the view stays Neutral, with the signed rent priced in and the unsigned pipeline not.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $-0.65 against $-0.25 expected; $-0.65 as filed, with no adjustments.
The evidence under this - 3 claims
The CEO said on the call that the rest of Black Pearl's schedule is unchanged and that the acceleration was paid for under a cap on expenses, so one early hall says little about the remaining phases.
The 8-K calls the changes 'tenant-modification change orders' and the tenant reimburses 50% of costs above $359.3M with a return, so part of the slip is the tenant's own redesign rather than a construction failure.
What management did with the money
Insiders in 2026Q3: 0 purchases and 4 sales, from their Form 4 filings.
The price, and where we disagree
In short. Fair. The price pays roughly what the signed rent is worth after overhead and the Barber Lake bill, and nothing for the sites still unsigned.
The shares have halved from June's high. They fell 15.6% on results day, when the first question on the call was about the Texas governor's order to audit every data centre waiting for a grid connection.
The next weeks test the two things that matter: whether Barber Lake starts paying rent this quarter as promised, and whether any of the 477 megawatts Cipher says are ready for 2027 gets leased. The third-quarter report is due on 2 November.
The evidence under this - 6 claims
The CEO argues an audit removes unserious applicants and so strengthens a developer that has posted its deposits and filed its water surveys (S05).
The 8-K calls the changes 'tenant-modification change orders' and the tenant reimburses 50% of costs above $359.3M with a return, so part of the slip is the tenant's own redesign rather than a construction failure.
The same 477 MW was 'in discussions for HPC hosting lease' in the May presentation and no lease followed in the five months to October, while TeraWulf and Applied Digital signed more capacity with the same class of tenant (C38, C39).
How we read the multiple
Count what a buyer of the whole company takes on: the shares, Google's warrants, the 2030 convertible notes and the debt, less the cash. About $9.0 billion.
Against about $793 million of yearly rent profit, that is 11.4 times, an 8.8% yearly return on rent from Amazon and a Google-backed tenant. Head office costs about $113 million a year in cash before shareholders see any of it.
Each extra dollar paid per dollar of that profit adds about $1.66 a share. Nothing is paid for sites beyond the 700 megawatts.
Against its own history. Its own past ratios describe a bitcoin miner, not a landlord, so this comparison carries little weight.
The evidence under this - 5 claims
- The NOI is management's (C03) and only one hall pays rent today, so the denominator is a promise rather than a filed figure.
- Netting all $4.56bn of cash flatters the multiple only if the restricted cash finishes the campuses with nothing left over; the company says each is funded to completion, but Barber Lake already needed $359.3M more.
Above $16.03 the 2031 notes convert into about 81.1M shares and $1.3bn of debt leaves, so the upside cases dilute and the fifteen-times figure falls to about $20.70.
- No filing carries a lessor payment schedule yet: the 10-Q shows no lease revenue, and the figure is a presentation number nobody outside the company has checked.
- Barber Lake's first rent has already moved from October 2026 to a phased start from the fourth quarter (C04), so the 2026 and 2027 figures are older than the schedule they rest on.
Part of today's compensation is a development team building three campuses at once; some of it may be capitalised into projects or fall once construction ends, so the run rate may overstate the steady state.
Applied Digital sells the same product to the same buyer on the same terms (C39); its perimeter differs mainly because its first campus already bills, which narrows the gap each quarter.
Priced against its rivals
There is no group to price it against. No listed pure-play prices this asset, so the price faced no peer comparison; the multiple is read against the contracted rent alone. The price is judged against its own history instead.
The bet: what it pays if it works
This is a bet that Cipher leases the sites it says are ready for 2027 on the same terms as its first three, without new shares. Even then, at a fair value for the rent, the shares would be worth less than twice today's price.
If leasing fails, a tenant walks away after a delay, or the pipeline never signs, little is left for shareholders: the lenders and the convertible notes are paid first, and the mining site's cheap power ends in 2027.
| HPC data center leasing revenue in two to three years | Leased gross megawatts needed | At the stated multiple |
|---|---|---|
| $507.00M | about 400 at 100% kept | 0.7× |
| $1.49B (the big win) | about 1.2K at 100% kept | 2.0× |
| $2.76B | about 2.2K at 100% kept | 3.7× |
To pay 2× (a Fair bet) it would need about $1.48B of HPC data center leasing revenue (about 1.2K leased gross megawatts); to pay 5× (a Cheap bet), about $3.71B of HPC data center leasing revenue (about 2.9K leased gross megawatts), at the same multiple.
Each row is HPC data center leasing alone, plus the rest of the company (Bitcoin mining) counted at about $99.30M.
The steps it has to take
- Done 700 MW leased to Amazon, a Google-backed Fluidstack and a hyperscaler
- Done Project notes raised for all three campuses
- Done First Black Pearl hall delivered and paying rent
- Slipped Every Barber Lake hall delivered and paying rent due 2027-03
- In progress Black Pearl rent fully ramped due 2027-03
- In progress Stingray delivered to its hyperscaler due 2027-06
- Not started A lease signed on some of the 477 MW available for 2027 due 2027-03
- In progress ERCOT's audited Batch Zero report naming Cipher's Texas sites due 2026-12
The detail behind the label
The evidence under this - 8 claims
The same 477 MW was 'in discussions for HPC hosting lease' in the May presentation and no lease followed in the five months to October, while TeraWulf and Applied Digital signed more capacity with the same class of tenant (C38, C39).
Where we disagree with the market
Every analyst rates it a buy, with targets near twice the price: they see a large AI landlord with a long pipeline.
The signed rent is worth about today's price; the pipeline is real but slower, costlier and paid for partly in new shares.
The difference is in the filings: the first campus slipped with a bill attached, Google's warrants carry a top-up if the price is low, and about 205 million possible new shares sit above today's count.
We would be wrong if Barber Lake is fully paying rent by March 2027 with no further cost, and the 477 megawatts are leased at better rent than the first three.
What analysts say, and the evidence
The evidence under this - 6 claims
- The NOI is management's (C03) and only one hall pays rent today, so the denominator is a promise rather than a filed figure.
- Netting all $4.56bn of cash flatters the multiple only if the restricted cash finishes the campuses with nothing left over; the company says each is funded to completion, but Barber Lake already needed $359.3M more.
The 8-K calls the changes 'tenant-modification change orders' and the tenant reimburses 50% of costs above $359.3M with a return, so part of the slip is the tenant's own redesign rather than a construction failure.
The top-up is measured on the day before exercise opens, which follows Barber Lake's commencement now due in the fourth quarter, so the price on that day, not today's, decides it; Cipher may also pay cash.
The 161.9M is the maximum the indentures could deliver; the 2031 notes carry capped calls bought to offset dilution, and Cipher may settle conversions in cash.
The same 477 MW was 'in discussions for HPC hosting lease' in the May presentation and no lease followed in the five months to October, while TeraWulf and Applied Digital signed more capacity with the same class of tenant (C38, C39).
What could go wrong, and how we would know
The risks
- A slip becomes a pattern. Barber Lake was called on schedule in February and May, then moved with $359.3 million of cost on Cipher.
- A late campus can be cancelled. Tenants may end a lease after a serious delay, and that would not trigger the guarantee behind it.
- Texas holds the pipeline. Every Texas site beyond the 700 megawatts waits on a grid queue now under audit, and the state has stopped data-centre permits.
- Your slice can shrink. Over 200 million possible new shares could join 414 million, and Google's warrant top-up grows as the price falls.
- Insiders sell, none buy. Insiders sold 37 times and bought never in a year; the founding holder sold about 7.9 million shares at its own choice.
The evidence under this - 8 claims
The 8-K calls the changes 'tenant-modification change orders' and the tenant reimburses 50% of costs above $359.3M with a return, so part of the slip is the tenant's own redesign rather than a construction failure.
Barber Lake was described the same way - 'fully funded through substantial completion' in February - and then needed $359.3M from Cipher (C04).
The 161.9M is the maximum the indentures could deliver; the 2031 notes carry capped calls bought to offset dilution, and Cipher may settle conversions in cash.
The top-up is measured on the day before exercise opens, which follows Barber Lake's commencement now due in the fourth quarter, so the price on that day, not today's, decides it; Cipher may also pay cash.
V3 Holding is the founding shareholder unwinding a stake built before the HPC business existed; its selling says more about its own portfolio than about the leases.
What we still do not know
- What rent each lease pays per unit of power; only the totals are public.
- How much of the Barber Lake lease Google still guarantees, and whether that covers the extra ten years.
- Fluidstack's own finances: it is private and publishes no accounts.
- Which Cipher sites Texas's grid operator has provisionally let into its first group; only totals have been published.
The evidence under this - 4 claims
The backstop was described as rising to about $1.73bn when the lease was upsized, which no Cipher filing read here states, and nothing says it covers the ten-year commitment that follows.
Fluidstack itself is a private AI cloud with no public accounts; outside the capped backstop its own credit is unknown (open question).
Rents escalate, so a straight average overstates the first years and understates the last; the NOI path in C03 rises from $686M in 2027 to $894M in 2035.
What would change our view
- Neutral→AttractiveBarber Lake delivers and the next lease is signed
It needs all of:
- ○Every Barber Lake hall is paying rent by 31 March 2027, with no extra cost to Cipher beyond the agreed $359.3 million.watched as it happens
- ○A lease signed on some of the 477 megawatts, at rent per megawatt as good as the first three, without new shares.watched as it happens
- Neutral→UnattractiveThe build or a tenant gives way
It needs any of:
- ○Barber Lake slips past June 2027 or needs more than $359.3 million, or another campus runs over budget.watched as it happens
- ○Any tenant ends its lease, or a guarantor refuses to pay, at any campus.watched as it happens
We would also be wrong if
- None of the 477 megawatts offered for 2027 is leased by the end of 2027.
What is coming
- 2 November 2026 Q3 2026 results: The first lease rent from Black Pearl, whether Barber Lake's first hall is dated inside the quarter, the overrun held at $359.3M, and any lease on the 477 MW.
Every dated event ahead
- 2 November 2026 Q3 2026 results: The first lease rent from Black Pearl, whether Barber Lake's first hall is dated inside the quarter, the overrun held at $359.3M, and any lease on the 477 MW.
- 10 December 2026 ERCOT Batch Zero eligibility audit report: Whether Cipher's Texas pipeline keeps its place in the batch study; the gigawatts beyond the signed 700 MW depend on it.
- March 2027 Barber Lake's last data halls deliver and rent begins: Whether the amended schedule holds; rent commencement also opens Google's warrant exercise and the top-up if the shares are below the floor.
- April 2027 ERCOT batch study date: How far the audit pushes the batch study, and so when any pipeline site beyond the 477 MW could be leased.
- July 2027 Luminant power agreement at Odessa reaches its minimum term: Whether Odessa keeps cheap power for mining or converts to a leased data centre, the mining stream's end date.
Since we wrote this
- 14 OctNewsCipher priced $850M of 0% convertibles and swaps 2030 notes for 40.7M shares. View heldPricing of $850M 0% convertible notes due 2032 and exchange of $165.9M of 2030 notes for 40.7M shares.
- 13 OctNewsCipher proposes $850M of 0% convertible notes and a 2030 notes exchange. View heldProposed $850M convertible notes offering due 2032 and concurrent exchange of 2030 notes.
- 8 OctBet moveBet re-struck: Priced for success to SpeculationRe-priced at the 7 October close; the rating moved from Expensive bet to Fair bet.
- 5 OctThesis reviewThe bet now names what it ridesThe owner asked that every bet name the macro trend behind it, after BWXT's bet was missing its chief executive's case for nuclear over gas.
- 4 OctThesis reviewOwner's checklist: revenue streams, cost of growth, management, catalystsThe owner's checklist found the method silent on who runs the company, what growth costs, where the money comes from and what is dated ahead; this pass fills those four blocks from the proxy and filings.
- 2 OctThesis writtenFirst thesis: fair for signed rent, pipeline unsignedNeutral with medium conviction: 700 MW is leased to Amazon, a Google-backed tenant and a hyperscaler, but Barber Lake slipped with a $359.3M bill and the pipeline waits on Texas.
Also checked 2 times with nothing new to act on.
Sources 26
Sources - 26
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-02-24Item 1's description of the Barber Lake (Fluidstack, Google backstop) and Black Pearl (Amazon Web Services, 15 years) leases with their delivery targets, the site pipeline, the Odessa power agreement, the Competition section naming ten rivals, and the risk factors on construction delay, tenant termination and capped guaranteesbacksC01C02C04C21C24C25C36C54C55C59
- S02primarycompany filing2026-08-04the balance sheet and cash flow statement, the debt table, the convertible conversion rates, the Google warrant terms and its $430.0M make-whole, the antidilutive share table, the at-the-market programme, construction commitments and the accelerated Black Pearl deliverybacksC06C07C08C09C11C12C13C14C18C27C44C50C54C55C59
- S03primarycompany release2026-08-04Q2 revenue of $25M and adjusted EBITDA of negative $30M, the Black Pearl acceleration with rent commenced, the Apollo option and the Stingray bondbacksC06C07C14C43
- S04primarycompany release2026-08-04700 MW contracted, about $11.4bn of contracted revenue, about $793M average annualised NOI and the year-by-year NOI path, the capital structure table, equipment secured by site, the Odessa snapshot and the 5.3 GW pipeline timeline subject to ERCOT's batch processbacksC01C03C18C20C21C25
- S05primarytranscript2026-08-04Cipher Digital second-quarter 2026 earnings call transcriptmanagement's reading of the Texas Governor's letter, the 477 MW available in 2027, the equity-funding answer, the cost cap behind the early delivery, and nine analyst turns from seven analysts, the first of them on the Governor's letter: the questions are a separate origin from the answers
Cipher has 477 megawatts potentially available in 2027 outside of this process. And we are pretty involved in discussions with multiple tenants
backsC19C20C22C42 - S06primarycompany filing2026-09-25the phased delivery now running from the fourth quarter of 2026 into the first quarter of 2027, the first $359.3M of excess cost borne by Cipher, the 50% reimbursement above it, and the binding ten-year commitment from a leading AI lab worth about $5.2bnbacksC04C05
- S07primarycompany release2026-09-25Barber Lake contracted revenue rising from $3.8bn to over $9bn, and management's framing of the amendmentbacksC05
- S08primarycompany release2026-02-24the Amazon lease at about $5.5bn over 15 years with Amazon.com, Inc. guaranteeing base rent and a stated ~100% NOI margin, Barber Lake at about $3.8bn over 10 years, the bond loan-to-cost ratios (~71% Barber Lake, ~85% Black Pearl), the rebrand and the sale of the joint-venture sites to CanaanbacksC02C26C44
- S09primarycompany release2026-05-05the third campus lease, 15 years with an investment-grade hyperscale tenant, both existing campuses described as on schedule, and the $200M revolverbacksC01
- S10primaryvendor data2026-10-02Research pack, frozen at content/CIFR/research-pack.jsonthe 1 October close, the 52-week range and returns, the price-move decomposition, the analyst rating series and target drift, FINRA consolidated short interest, the insider roster and base rate, and the quarterly revenue seriesbacksC23C28C29C30C31
- S11primarycompany filing2026-10-0237 open-market sales and no purchases, 27 of them discretionary and 10 under a Rule 10b5-1 plan, read off each Form 4's own checkboxbacksC29
- S12primarygovernment data2026-09-15the underlying rate of change, measured by a statistical agency rather than the industry; quote as frozen by the Applied Digital thesis's read of the same page
Servers alone accounted for an estimated 7% of commercial sector electricity consumption in 2025. Data center server electricity use grows to 22%-33% of commercial building electricity use by 2050 across our cases.
backsC35 - S13secondaryindustry data2026-09-15vacancy at 1.4%, 80.4% of capacity under construction pre-committed, under 1,500 MW available: the scarcity the moat rests on, measured outside the company
primary market vacancy fell to a record low of 1.4% in H1 … commitments made on 80.4% of all under-construction capacity … Total under-construction capacity across primary markets increased by 24.8% in H1 to a record high of 7,481.1 MW … Less than 1,500 MW of future capacity across all primary markets remains available, representing approximately six months of demand
backsC32 - S14secondaryindustry data2026-09-15Synergy Research Group - hyperscale operators to account for 67% of all data center capacity by 2031 Synergy Research Groupthe size of the substitute: how much of their own capacity the concentrated buyers build rather than lease
Almost 60% of that hyperscale capacity is now in own-built, owned data centers with the balance being in leased facilities … they now account for 48% of the worldwide capacity of all data centers … by 2031 hyperscale operators will account for 67% of all capacity
backsC33 - S15primaryindustry data2026-08-11a broker's North American absorption, capacity under construction and the owner-occupied part of it
North America data center demand reached an all-time high in the first half of 2026 with 25 GW of absorption, double the level from a year ago and five times that of two years prior. … North America now has 66 GW under construction, 95% of which is pre-committed. … Hyperscalers remain the dominant force, accounting for 59% of 2026 data center tenant demand while meeting capacity needs through both leasing and self-building, with 28 GW of owner-occupied capacity currently under construction.
backsC34 - S16primarycompany filing2026-07-29a named competitor selling the same product to the same buyer: approximately 15-year leases to investment-grade hyperscalers on a standardised designbacksC39
- S17primarycompany filing2026-08-05a named competitor leasing to the same tenant: Fluidstack HPC leases at Lake Mariner, 102 MW of critical IT energised in July 2026, and the sale of its Abernathy joint venture stake to FluidstackbacksC38
- S18primaryregulator2026-06-18the size of the Texas large-load queue and the batch process every Texas pipeline site waits on; read through press_fetch.py
ERCOT is tracking more than 438,000 MW of large load requests, nearly 89 percent from data centers alone. "Batch Zero" is the name of the first group of large-user applicants to go through the new process.
backsC48 - S19primaryregulator2026-09-11conditional classification of 204 projects (66.4 GW) as base load and 158 (127.9 GW) as studied load, the audit report due 10 December 2026, and an undetermined study timeline against the original 9 April 2027; the PDF was downloaded with curl and read with pdftotext by the research agent, a different client from press_fetch
ERCOT has provided conditional classification for all Batch Zero Large Loads: 204 projects (66.4 GW) met the qualifications to be conditionally included as base load; 158 projects (127.9 GW) met the qualifications to be conditionally included as studied load. … ERCOT has begun a verification and audit of the eligibility of all Large Loads provisionally included in Batch Zero. ERCOT will file the Batch Zero Eligibility Verification Report with the Commission by December 10, 2026.
backsC47 - S20primaryregulator2026-08-03Office of the Texas Governor, 3 August 2026: Governor Abbott Directs Comprehensive Data Center Audit Office of the Texas Governorthe directive behind the paused Batch Zero classification, and the 474 GW of requests, about 90% data centres; read through press_fetch.py
directed the Public Utility Commission of Texas (PUCT) and the Electric Reliability Council of Texas (ERCOT) to conduct a comprehensive verification and audit of all data centers advancing through ERCOT's interconnection process. The PUCT and ERCOT must complete this audit before any data center project moves forward.
backsC46C48 - S21secondarynews2026-09-21Texas Tribune, 21 September 2026: Abbott orders a halt to data center permits pending audits Texas Tribunethe second directive: no state environmental permits for data centre projects until the grid and water audits are complete; read through press_fetch.py
Data centers must pay their own way, protect our grid and water and complete the ERCOT and TWDB audits. Until they do, TCEQ will issue no permits sought by data center projects.
backsC49 - S22secondarynews2026-09-15San Angelo LIVE!, 15 September 2026: Colorado City given $10 million by AI data center companies San Angelo LIVE!a local outlet placing Anthropic inside Barber Lake beside Fluidstack, and stating a Cipher project near San Angelo is in Batch Zero; circumstantial on who the 'leading AI lab' is, and not used to identify it
Fluidstack and Anthropic are tenants inside it. Anthropic is the company behind Claude; those models run on the compute in the data center. Google backstops Fluidstack's lease.
backsC52 - S23secondarynews2025-09-25Coinspeaker via Yahoo Finance, 25 September 2025: Google gains 5.4% of Cipher in Fluidstack backstop Coinspeakerthe size of the original Google backstop, and that TeraWulf's Fluidstack lease carried the same Google structure
Under the arrangement, Google will backstop $1.4 billion of Fluidstack's lease obligations in exchange for approximately 24 million shares of Cipher common stock, giving the Mountain View company a 5.4% stake.
backsC50 - S24primarycompany release2025-11-12the end buyer behind Fluidstack's Texas leasing, in the buyer's own words: a $50bn commitment to build with Fluidstack
building data centers with Fluidstack in Texas and New York, with more sites to come. These facilities are custom built for Anthropic with a focus on maximizing efficiency for our workloads
backsC51 - S25secondarynews2026-09-11Houston Public Media, 11 September 2026: ERCOT to survey more than 400 data centers as part of Abbott audit Houston Public Mediahow much of the Texas queue was excluded from Batch Zero, measured by the grid operator and reported independently of any developer
According to the agency, 373 projects were excluded from Batch Zero. The projects were requesting more than 300 gigawatts of electricity from the grid.
backsC47 - S26primarycompany filing2026-04-20the executive officers and their start dates, the beneficial ownership table at 8 April 2026, the 2025 bonus goals and payouts, the 2025 PSU design and payout, and the CEO's 2021 market-capitalisation PSUsbacksC56C57C58