Crypto & Digital Assets
A bitcoin miner that has successfully optioned itself into an AI landlord — $11.4B of hyperscaler lease backlog backstopped by Google and Amazon is real, but the equity is priced for flawless execution of a buildout that hasn't started cashing rent, and the whole thesis is a bet on shovels-in-the-ground beating the sector's 25%-delivered track record.
Research
The verdict
A bitcoin miner that has successfully optioned itself into an AI landlord — $11.4B of hyperscaler lease backlog backstopped by Google and Amazon is real, but the equity is priced for flawless execution of a buildout that hasn't started cashing rent, and the whole thesis is a bet on shovels-in-the-ground beating the sector's 25%-delivered track record.
Cipher is a vertically integrated US data-center developer that is mid-pivot from pure-play bitcoin miner to AI/HPC landlord. The 10-K states it plainly: "Over the past several years, we have intentionally evolved from a pure-play bitcoin miner into a vertically integrated data center development and operations platform focused on energy-intensive compute infrastructure". The company renamed itself Cipher Digital Inc. in February 2026 (still trades CIFR on Nasdaq; HQ moved to 1 Vanderbilt Avenue, NYC) — a deliberate signal that "mining" is no longer the identity.
How it makes money today vs. tomorrow. Today, ~100% of revenue is bitcoin self-mining: FY2025 total revenue was $157.0M, of which bitcoin mining was $151.27M (FY2024 $151.27M; FY2023 $126.84M). Tomorrow, the model is HPC colocation: long-dated, fixed-rent leases to hyperscalers, with Cipher as the landlord owning the power, land, interconnect, and shell.
The three anchor contracts (the actual business case):
Key payment-term structure: these are NNN-style, fixed-rate, decade-plus leases with investment-grade backstops (Google, Amazon) — the polar opposite of bitcoin's variable, hashprice-driven cash flows. That swap from commodity exposure to contracted annuity is the entire equity story.
Customers / suppliers / competitors: Customers = Fluidstack (Google-backed AI cloud), AWS. Suppliers = miner OEMs (Bitmain-class rigs) for the legacy fleet, EPC/construction, and power counterparties (ENGIE, Luminant/Vistra PPA, American Electric Power Direct Connect). Competitors = the other miners racing the same pivot: IREN, Hut 8, Riot, Core Scientific, TeraWulf, CleanSpark.
Map: Land + interconnect → Power → Shell/fit-out → Compute tenant → End AI demand.
Chokepoints / single-source dependencies: (1) Grid interconnection — multi-year ERCOT queues make permitted, energized sites the scarce asset; this is Cipher's actual edge. (2) Construction execution — turning 0 MW of HPC into energized, rent-paying MW on schedule. (3) Tenant concentration — two customers (Fluidstack/Google and Amazon) "represent a significant portion of our expected future revenues," and the 10-K flags that the loss of either "could have a material adverse effect".
What's durable:
What's weak: Cipher is a landlord, not a technology company. There is no IP moat, no network effect, no switching cost beyond the 10–15-year lease term. The "product" is megawatts; the differentiation is execution and cost of capital. Once energized, a hyperscaler's switching cost is the lease — real, but finite. Bargaining power is asymmetric: Cipher needs the hyperscaler's credit and demand more than AWS/Google need any single 300 MW site. The backstop economics (Google taking 5.4% in warrants to guarantee Fluidstack) show who held the leverage at signing.
The compiled segments.csv is empty, so segment figures come from the 10-K segment footnote:
| Segment | FY2023 | FY2024 | FY2025 | Trend |
|---|---|---|---|---|
| Bitcoin Mining revenue | $126.84M | $151.27M | ~$151–157M total co. rev | Plateauing — halving + difficulty offset fleet growth |
| HPC / hosting revenue | $0 | $0 | ~$0 (pre-commencement) | The entire forward story; first rent ~Aug–Sep 2026 |
Reportable structure: management runs essentially one revenue-generating segment today (Bitcoin Mining) with HPC as a development-stage second segment that has not yet produced material revenue. Geographically, the asset base is concentrated in West Texas (ERCOT) plus land options in Ohio (Ulysses). The decisive segment fact: revenue is decelerating in the only segment that exists, and accelerating (from zero) in the one that doesn't yet. That gap — months of bitcoin attrition before HPC rent switches on — is the execution window the equity is trading through.
The latest filed print (10-Q for the quarter ended 2026-03-31):
Unusual vs. own history: the GAAP loss optics ($822M FY2025) are almost entirely non-cash derivative/warrant remeasurement — a feature of having raised $1.5B+ in convertibles with conversion features and issued Google warrants. Don't read it as operating distress; read the cash line (~$715M) and the capex run-rate.
No transcripts are on disk (transcripts/ empty), so this is ``. Across the 2025→Q1 2026 calls the tonal arc is unmistakable: management has gone from defending bitcoin economics to selling a hyperscale-infrastructure narrative. Q3 2025 was framed as a "decisive pivot from Bitcoin mining to hyperscale AI/HPC infrastructure"; Q1 2026 was headlined "$11.4B contracted revenue, HPC shift" and management said it plans to fully wind down bitcoin mining by July 2027 at the latest, potentially repurposing Odessa for HPC. Recurring phrases now: "contracted revenue," "world-class hyperscalers," "energized power," "critical IT load." What they stopped saying: hashrate-growth and fleet-efficiency boasts that dominated the 2024 calls. The sentiment shift is from commodity-operator to real-estate-developer — and the market has rewarded it (+536% over 12 months).
Bitcoin-miner-to-AI peers (multiples ``, June 2026; n/a where I could not source a clean figure). All figures are USD.
| Company | Ticker | Mkt cap | Rev (TTM) | P/S | Notes |
|---|---|---|---|---|---|
| IREN (Iris Energy) | IREN | ~$19.4B | ~$757M | ~25x | Sector leader; already earning AI-cloud rev, run-rate >$3.7B targeted by YE2026 |
| Hut 8 | HUT | ~$12.6B | n/a | n/a | "Highest-quality pivot," IG backstops + non-dilutive financing |
| Cipher Digital | CIFR | ~$10.6B | ~$157M | ~67x | Pre-HPC-revenue; $11.4B backlog not yet in the P&L |
| Riot Platforms | RIOT | ~$10.6B | n/a | n/a | More BTC-tied; slower lease conversion |
| CleanSpark | CLSK | n/a | ~$0.76B | ~1.56x | Still BTC-centric; "mining doesn't make sense at current hashprices" per mgmt |
The pattern is crystal clear — this name moves on hyperscaler deals and power/regulatory news, not earnings:
What the market reacts to: (1) a new hyperscaler lease or backstop, (2) power/interconnect wins and favorable ERCOT regulation, (3) capital raises (convertibles — a negative dilution reaction). It essentially does not react to mining results. Up 536% over 12 months.
CEO & founder — Tyler Page. Founder-CEO; before Cipher he was Head of Business Development for digital-asset infrastructure at Bitfury (he did not found Bitfury), and earlier sat on the management committee / ran client strategies at NYDIG, with prior roles at Stone Ridge, Guggenheim, Goldman Sachs and Lehman derivatives desks, and began as a Davis Polk lawyer (J.D. Michigan, B.A. UVA). Archetype: a finance/dealmaker, not an operator. That fits the company's reality exactly — Cipher's value creation has been deal creation (SoftBank, Fluidstack/Google, AWS, ENGIE, AEP), and the most dangerous gap is operational execution (building 600+ MW on time), which is not Page's background.
Skin in the game / ownership: Page beneficially owned ~7.8M shares + ~2.8M RSUs after Dec 2025 transactions, but he sold ~858K shares in mid-December 2025 at $14.74. The dominant holder is Bitfury Top HoldCo (~93.4M shares, ~23%), which is actively monetizing: a variable-prepaid-forward on up to 2M shares (June 2, 2026) and a 1.2M-share sale at ~$25.35–26.10 (June 4, 2026), part of ~$47.9M sold. Ownership is ~75% institutional. Read: founder/anchor are selling into strength — rational portfolio management after a 5x, but not the insider-accumulation signal a bull would want.
Capital-allocation history: The defining moves are (1) pivoting permitted power to HPC at >10x the BTC valuation multiple — value-accretive if delivered; (2) funding it with $1.5B+ of convertibles (largely 0%-coupon) plus a $70M capped call to blunt dilution; (3) accepting Google warrants (5.4%) as the price of a bankable backstop. This is a dealmaker optimizing for the lowest cost of capital to fund a landgrab — coherent, aggressive, dilutive. ROE/ROIC are negative on GAAP (the warrant/derivative marks), so a clean returns track record doesn't exist yet; the bet is that contracted rent turns the ROIC positive from 2027.
Red flags: related-party lineage to Bitfury (the SPAC sponsor / largest holder) is worth watching as it sells down; heavy reliance on equity-linked financing; promotional cadence of operational-update press releases. Nothing fraudulent surfaced — see Lens 10.
Forensic lens. Cipher is not an accounting fraud risk but it is an earnings-quality minefield by construction:
Regulatory findings (required sub-section):
regulatory/regulatory-findings.md (EDGAR EFTS, LR + AAER, search window 2021-06-20→2026-06-20) returns 0 SEC findings for Cipher Mining.This is an operating-company battery, but Cipher is in a one-time transition where trailing EPS is meaningless and forward EPS hinges on a binary (do the leases energize on schedule?). So the projection is framed around contracted-revenue ramp, not a smooth EPS curve. No forecast.ts create is run (watchlist mode).
Inputs (all labeled):
Base case (FY2027): HPC rent partially online (Black Pearl + Barber Lake phases ramping), bitcoin winding down. Revenue inflects from ~$150M (2025/26) toward a ~$0.7–1.0B run-rate exiting 2027 as the first two campuses energize. GAAP still messy from D&A step-up + warrant marks; EBITDA turns meaningfully positive as fixed rent lands on a built-out base. Bull case: all 907 MW energize on time, Colchis (1 GW, 2028) and the 3.2 GW pipeline get leased to additional hyperscalers, and Cipher re-rates on contracted EBITDA toward IREN-like scale. Bear case: construction slips (the sector has delivered only ~25% of leased AI capacity ), build costs blow past $10M/MW, a tenant (Fluidstack) wobbles before the Google backstop fully seasons, and the multiple de-rates hard from "backlog" to "delivered MW."
The forecast that actually matters (logged narratively, not via forecast.ts): Does Cipher energize and begin recognizing rent on the Black Pearl (AWS) and Barber Lake (Fluidstack) campuses on roughly the announced 2026 timeline? My estimate: ~65% probability the first rent is recognized in H2 2026 broadly on schedule (some phase slippage likely), ~80% within 6 months of guide. That binary, not an EPS line, is the scoreable claim.
Bull case. Cipher has done the hardest thing in the AI-infrastructure trade: it converted scarce, permitted ERCOT power into $11.4B of contracted, investment-grade-backstopped rent before most peers signed anything. Google ($1.4B) and Amazon (full-rent guarantee) are underwriting the cash flows — a project-finance lender's dream and a validation no slide deck can fake. The interconnect moat is real and widening (queues stretch years), the pipeline is 3.2 GW, and a favorable ERCOT large-load framework is a structural tailwind. If the first two campuses energize in H2 2026, the market stops valuing trailing mining losses and starts valuing a multi-hundred-million-dollar contracted-EBITDA annuity — and on gross-energized-power multiples that supports a materially higher cap. The contrarian bull point: the market still half-prices CIFR as a bitcoin miner (it sits below IREN/Hut 8 on contracted scale), so successful delivery is a re-rating and a category change.
Bear case (permanent-impairment risks). (1) Execution gap. The sector has delivered ~25% of what it leased; Cipher has built zero MW of HPC and must turn 600+ MW into energized, rent-paying capacity across 2026 with a finance CEO, not an operator. Miss the milestones and you eat a "structural de-rating." (2) Two-customer concentration. Fluidstack and Amazon are "a significant portion of expected future revenues"; the 10-K itself flags either's loss as materially adverse. Fluidstack is a young AI-cloud reseller — the Google backstop is exactly because its standalone credit isn't AWS-grade. (3) Dilution + cost of capital. $1.5B converts, Google warrants (5.4%), ongoing equity-linked financing, and Bitfury selling into strength — the share count and the anchor's exit are both headwinds. Pre-mortem (18 months out, thesis broke): Black Pearl phase-2 slipped to 2027, build cost ran $12M+/MW forcing a dilutive equity raise at a lower price, a wobble at Fluidstack spooked the market on counterparty quality, and CIFR de-rated from a "contracted-backlog" multiple to a "show-me-delivered-MW" multiple — a 40–50% drawdown without any fraud, purely timing + dilution.
Are multiples too high? On trailing numbers, absurdly (no HPC revenue). On forward contracted economics, they're defensible only if delivery is near-flawless — there is no margin of safety for slippage. This is a momentum/execution story priced for success.
Contrarian view (what the market refuses to see): the bull crowd treats "$11.4B contracted" as if it were booked revenue; the bear crowd treats it as vaporware. The truth the tape is mispricing in both directions is timing — the 12–18-month window where bitcoin revenue decays toward zero before HPC rent fully lands is a genuine air pocket, and the stock's 14%+ short interest says skeptics are positioned for exactly that gap to bite.
Dismantling the bull case. What breaks this?
A profitable, fast-growing digital bank mispriced as a crypto name and now mispriced again by fear — but the entire bull/bear hinges on one binary question the market cannot yet resolve: are the +$2.0B of fair-value loan marks real, or is Muddy Waters right?
A levered, sub-NAV Ethereum proxy run by the best operators in the trade — own the discount-closing buyback, not the "treasury premium" that already died; the bet is ETH plus a 0.83x→1.0x mNAV mean-reversion, fighting a 95% drawdown and a structurally bigger rival.
A real, cash-generative super-app now valued like a crypto-beta call option — the bull thesis (tokenization + a 27M-customer flywheel) is genuine, but ~37x forward earnings re-prices violently every time the crypto/options trading cycle exhales, and the prediction-markets engine that drove Q1 growth is one adverse court ruling from being switched off.