Cipher Mining
Power-site arbitrage: Cipher's scarce asset is energized, grid-connected, interconnect-queued MW in Texas (Odessa, Barber Lake, Black Pearl), NOT its ASICs. It repurposes those sites from volatile hash-price-linked BTC mining (~$3-5M/MW/yr) into long-dated (10-15yr), contracted, credit-backstopped HPC leases (~$1.5-2M/MW/yr) at higher, more stable margins and richer multiples. Unlike a pure shell-lessor it retains full project ownership and self-develops the campuses, so it carries build/financing/execution risk to capture more of the value chain. A capital-intensive real-asset landlord, not a software business.
The thesis on this name
State of the AI Cloud
TeraWulf's structural twin — the entire 300MW Barber Lake TX site leased to Google-backstopped Fluidstack with full project ownership retained — but a more hybrid, still-mining model.
State of the AI Cloud
TeraWulf's structural twin — the whole Barber Lake site committed to Google-backstopped Fluidstack at ~207MW of contracted critical IT load, full project ownership retained — but a more hybrid, still-mining model.
State of the AI Cloud
Barber Lake committed to Google-backstopped Fluidstack at ~207MW of contracted critical IT load (~$3.8B base term), full ownership retained; smaller clip for single-site concentration, residual mining, and a counterparty the anchor customer has bypassed elsewhere.
Earnings, margins, COGS & capex
Cipher is mid-pivot: Q1 FY26 revenue fell to $34.8M (from ~$59.7M) as it decommissioned Black Pearl mining (Feb 2026) and wound Odessa toward exit (~end-2027), posting a -$114.3M net loss. The forward story is NOT the income statement today but ~$11.4B of contracted HPC revenue across three signed hyperscale leases (~700MW gross), with ~$787M average annualized NOI and first HPC revenue expected later in 2026 as Barber Lake energizes. It is funding a multi-hundred-MW build with a $2B 6.125% bond, a $200M revolver, and ~$4.2B cash — a leveraged bet that the sites deliver on a compressed 2026-2027 timeline and the neocloud/hyperscaler counterparties pay. The bull edge is fully-contracted, IG/Google-backstopped, long-dated cash flows on power it already controls; the bear edge is Fluidstack counterparty risk (only partially covered by Google's debt backstop), single-region concentration in Texas, execution on the compressed delivery, and residual hash-price exposure while mining runs off.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~0¢ is cost of goods and ~0¢ operating expense, leaving ~100¢ of operating profit (~114¢ net).
Revenue trend
Margins
negative - transition trough; mining revenue falling, HPC lease revenue not yet begun
company-disclosed forward metric - starts recognizing as campuses energize from H2 2026
declining by design - Black Pearl decommissioned Feb 2026; Odessa exit ~end-2027
up - interest expense increase cited as a driver of the Q1 net loss
COGS structure
Legacy COGS is BTC-mining power/electricity plus ASIC operations (Odessa ~207MW, ~11.6 EH/s, 17.2 J/TH fleet efficiency, ~$0.028/kWh power cost in Q1). As the model pivots, the economics shift to the HPC-landlord shape: the cost base becomes datacenter build (power delivery, liquid/air cooling to the racks, campus construction) plus interest on the $2B bond and project debt, against contracted lease revenue. The larger economic costs sit below any gross line as depreciation on the built campuses and interest — so contracted NOI, not near-term GAAP margin, is the relevant lens during the transition.
Capex
Directed at HPC build-out: delivering ~700MW of contracted gross HPC capacity (Barber Lake 300MW + Black Pearl 300MW for AWS + a third IG-hyperscaler campus) on a compressed 2026-2027 timeline, with a stated ~4.2GW total-portfolio ambition by 2030+. Mining capex is halted. Funded by a $2B 6.125% senior bond, a $200M revolver (Morgan Stanley-led, Goldman/JPM/Wells Fargo/Santander/SMBC), and ~$4.2B cash/restricted — not operating cash flow. Capex intensity is the defining feature and the source of both the re-rating option and the balance-sheet/execution risk.
Latest earnings
Revenue in a planned trough (mining wind-down); the market focus is contracted-backlog growth (third lease + $11.4B contracted revenue) and delivery timeline, not the current-quarter print [fact]
First HPC lease revenue expected later in 2026 (Barber Lake); mining expected immaterial by 2030 and likely exited by end-2027; ~4.2GW portfolio ambition by 2030+ [company-disclosed]
- Contracted HPC revenue / capacity
- ~$11.4B across three hyperscale leases; ~700MW gross contracted; ~$787M avg annualized NOI
- AWS lease (Black Pearl)
- 15-yr, 300MW, ~$5.5B; delivery from Jul 2026, rent from Aug 2026
- Fluidstack/Google lease (Barber Lake)
- 10-yr, 300MW, ~$3.8B initial (up to ~$9.0B w/ extensions); Google backstop ~$1.73B + ~5.4% warrants
- Liquidity / debt
- ~$4.2B cash & restricted + $76M BTC + undrawn $200M revolver; ~$5.2B debt (net debt ~$0.96B); $2B 6.125% bond
- Mining (running off)
- Odessa sole mine ~207MW, ~11.6 EH/s, ~346 BTC Q1; Black Pearl decommissioned Feb 2026; exit ~end-2027
Growth drivers
- ~$11.4B contracted HPC revenue across three signed hyperscale leases (~700MW gross), 10-15yr terms, ~$787M avg annualized NOI [company-disclosed]
- AWS 15-yr lease — 300MW at Black Pearl TX, ~$5.5B total contract value; delivery in two phases beginning July 2026, rent from Aug 2026 [fact]
- Fluidstack (Google-backstopped) 10-yr lease — entire 300MW Barber Lake TX; ~$3.8B initial contracted revenue, up to ~$9.0B with two 5-yr extensions; Barber Lake energizing ~Sept 2026 [fact]
- Third investment-grade hyperscale tenant: 15-yr lease signed Q1 FY26 (Mar 2026) for a third campus [fact]
- Google strategic backstop of ~$1.73B of Fluidstack's lease obligations + warrants for ~24M CIFR shares (~5.4% pro forma) validates the counterparty [fact]
- ~4.2GW total-portfolio ambition by 2030+ on Texas power Cipher already controls; full project ownership retained (captures more value than shell-leasing) [company-disclosed]
Reported financials — SEC EDGAR
Audited GAAP figures pulled from SEC filings · latest filing 2026-02-24. The audited primary-source spine — not financial advice.
Revenue — annual (GAAP)
Margins & balance sheet — FY’25
Bull & bear
Cipher is a fully-contracted miner-to-AI-DC pivot with ~$11.4B of long-dated HPC revenue across three signed hyperscale leases (AWS, Google-backstopped Fluidstack, and a third IG tenant), on Texas power it already controls and campuses it fully owns — a re-rating from a Bitcoin-miner multiple toward a datacenter-landlord multiple as revenue begins recognizing in H2 2026.
- Entire 300MW Barber Lake leased to Fluidstack (Google-backstopped, ~$3.8B initial / up to ~$9.0B w/ extensions) plus a 15-yr ~$5.5B AWS lease (300MW Black Pearl) and a third IG-hyperscaler lease = ~$11.4B contracted, ~$787M avg annualized NOI [fact]
- Google backstops ~$1.73B of Fluidstack's obligations and holds ~5.4% warrants — a blue-chip credit + equity endorsement of the anchor counterparty [fact]
- Owns the scarce asset (energized, interconnect-queued Texas power) AND retains full project ownership, capturing more value than shell-leasing peers [fact]
- Near-term 2026 delivery (Barber Lake ~Sept 2026, AWS phases from Jul 2026) converts backlog to cash sooner than greenfield builds [fact]
- ~4.2GW portfolio ambition by 2030+ gives a long runway to compound additional hyperscale leases if AI-compute demand and financing hold [company-disclosed]
Cipher is a low-conviction, high-execution-risk pivot: a leveraged bet that a compressed 2026-2027 Texas build delivers on time and that a neocloud counterparty (Fluidstack) — only partially backstopped by Google's DEBT guarantee, not a Google tenancy — pays across a 10-year lease, all while the income statement is deeply negative and mining runs off. Single-site/single-region concentration and heavy capital intensity compound the fragility.
- Fluidstack counterparty risk: Google backstops project DEBT, not a Google tenancy — the lease's value still hinges on a neocloud's ability and willingness to pay for a decade [fact]
- Income-statement air pocket: -$114.3M net loss in Q1 FY26 on just $34.8M revenue as mining collapses before HPC revenue recognizes — negative cash flow through the build [fact]
- Single-region concentration in Texas (Odessa/Barber Lake/Black Pearl) exposes the whole book to ERCOT grid, weather, interconnect, and regulatory risk [estimate]
- Compressed 2026-2027 delivery across multiple hundred-MW campuses is real execution risk; any energization slip defers revenue and pressures debt service [estimate]
- Leverage + refinancing: a $2B 6.125% bond and project debt against tiny current revenue mean a risk-off market or an AI-capex air-pocket bites hard; residual hash-price exposure remains while Odessa runs to ~end-2027 [fact/estimate]
What it is worth
Sum-of-the-parts / contracted-NOI capitalization, framed as a low-conviction satellite long. The relevant value is the ~$11.4B of contracted HPC revenue (~$787M avg annualized NOI across three 10-15yr leases) capitalized at a datacenter-landlord yield, LESS ~$5.2B debt and the build capex still to spend, PLUS residual mining/BTC and the ~4.2GW portfolio option. At ~$9.9B equity cap / ~$10.9B EV, the market is already pricing successful energization and counterparty payment on the contracted book - so the equity is a leveraged call on flawless 2026-2027 delivery rather than a cheap contracted-cash-flow stub.
A delivery slip, a Fluidstack counterparty scare (only partially backstopped), a refinancing squeeze, or an AI-capex air-pocket compresses the multiple and forces the leverage + tiny current revenue into focus - the low-conviction downside [estimate]
Near current levels (~$25): the contracted book is real but back-end-loaded and execution-dependent; the stock trades on each energization/leasing milestone and AI-capex sentiment, range-bound until HPC revenue actually recognizes [estimate]
Re-rating case: on-time energization + reliable counterparty payment convert ~$11.4B contracted revenue / ~$787M NOI into a datacenter-landlord multiple, plus optionality on the ~4.2GW portfolio ambition and further hyperscale leases [estimate]
Current GAAP earnings are meaningless here (transition trough, -$114.3M Q1 FY26); value hinges on (1) on-time energization of Barber Lake/Black Pearl, (2) Fluidstack/AWS/third-tenant payment across the lease life, and (3) refinancing the build without a risk-off shock. The Google backstop de-risks the anchor lease but covers debt, not rent. Low conviction: fully contracted but execution-, counterparty-, and financing-dependent, on a single-region Texas footprint.
SWOT
Strengths
- Fully-contracted flagship — entire 300MW Barber Lake leased to Fluidstack (Google-backstopped) and 300MW Black Pearl leased to AWS — ~$11.4B contracted HPC revenue across three signed hyperscale leases gives long-dated (10-15yr) revenue visibility [fact]
- Google validation — Google backstops ~$1.73B of Fluidstack's obligations and holds warrants for ~5.4% of CIFR — a blue-chip credit and equity signal behind the anchor counterparty [fact]
- Owns the scarce asset (energized, interconnect-queued Texas power) and retains full project ownership of the campuses, capturing more value than a pure shell-lessor [fact]
- Tenant diversification beyond a single neocloud — AWS (investment-grade) and a third IG hyperscaler reduce reliance on Fluidstack alone versus single-counterparty peers [fact]
- Near-term delivery — Texas sites energizing in 2026 (Barber Lake ~Sept 2026; Black Pearl AWS phases from Jul 2026) convert backlog to revenue faster than greenfield [fact]
Weaknesses
- Fluidstack counterparty/neocloud credit risk — Google's backstop covers project DEBT, not a Google tenancy — if Fluidstack fails to pay, the recourse is limited to the backstop, not a hyperscaler lease [fact]
- Deeply loss-making today (-$114.3M Q1 FY26) on collapsing mining revenue ($34.8M) with HPC lease revenue not yet recognizing — an income-statement air pocket during the transition [fact]
- Single-region concentration — all flagship capacity is in Texas (Odessa/Barber Lake/Black Pearl), exposing the book to ERCOT grid, weather, and interconnect risk [estimate]
- Compressed 2026-2027 delivery timeline across multiple hundred-MW campuses raises construction/energization execution risk [estimate]
- Residual hash-price exposure while Odessa mining runs off to ~end-2027, and heavy capital intensity ($2B bond + facilities) strains the balance sheet like peers [fact]
- More hybrid than the cleanest pivots (e.g. TeraWulf) — keeping meaningful mining longer delays the full re-rating to a datacenter multiple [estimate]
Opportunities
- ~4.2GW total-portfolio ambition by 2030+ on power Cipher already controls — a large runway to sign additional hyperscale leases if demand and financing hold [company-disclosed]
- Re-rating from a Bitcoin-miner multiple to a contracted-datacenter-landlord multiple as HPC revenue begins recognizing in H2 2026 and mining runs off [estimate]
- Retaining full project ownership lets Cipher capture development margin and optionality (sale-leaseback, refinancing, or REIT-style structures) that shell-lessors forgo [estimate]
- The AI-datacenter capacity super-cycle keeps energized, interconnect-queued MW scarce — Cipher's Texas power is a durable entry ticket if demand persists [estimate]
- Additional Google/hyperscaler participation (warrants, backstops, or direct tenancy) could further de-risk future campuses [estimate]
Threats
- AI-capex air-pocket — two consecutive hyperscaler capex guide-downs or an AI-datacenter digestion regime would reprice the entire miner-pivot cohort, Cipher included [estimate]
- Counterparty failure — Fluidstack (a neocloud) stumbling — only partially mitigated by Google's debt backstop — would impair the Barber Lake lease value [fact/estimate]
- Execution/energization slip on the compressed 2026-2027 build defers backlog-to-revenue conversion and pressures the debt-service math [estimate]
- Financing/refinancing risk — a risk-off market makes rolling the $2B bond + project debt expensive while capex outruns revenue [estimate]
- Bitcoin price collapse hits the still-running Odessa mine and the ~$76M BTC on the balance sheet during the wind-down [fact]
- ERCOT grid/regulatory scrutiny on large Texas datacenter load could gate energization or raise power costs [estimate]
Moats, dependencies & bottlenecks
Moats
grid-connected, permitted MW is genuinely scarce and slow to replicate, but the sites themselves are not unique and rivals hold their own power The real entry ticket to AI-DC hosting; Cipher's edge is having energized power ready, not proprietary technology [fact]
Long-term contracted HPC backlog (~$11.4B across three 10-15yr leases) contractual switching cost, not technological lock-in; value depends on counterparty ability/willingness to pay and on energization Backlog is only as good as Fluidstack/AWS/the third tenant paying — and much of it recognizes only from H2 2026+ [fact]
Google backstop + equity participation (via the Fluidstack deals) covers ~$1.73B of project debt and aligns Google via ~5.4% warrants, but is a debt backstop, not a Google tenancy A meaningful credit signal that de-risks the anchor lease, but it is granted, not owned, and covers debt not rent [fact]
captures more value-chain margin and optionality, but adds build/financing/technology-obsolescence risk Differentiates Cipher from pure landlords by keeping development economics — at the cost of carrying execution risk [fact]
a vulnerability, not a moat; heavy leverage vs cash-rich hyperscalers and even better-anchored peers (IREN) Funds the build with a $2B 6.125% bond + facilities; refinancing access is a dependency, not an advantage [fact]
Dependencies
A neocloud renting the entire 300MW Barber Lake site on a 10-yr lease; its credit and demand ARE the thesis — only partially mitigated by Google's ~$1.73B DEBT backstop (not a Google tenancy) [fact]
Backstops ~$1.73B of Fluidstack's obligations and holds warrants for ~5.4% of CIFR; a credit/validation crutch under the anchor lease, but covers debt, not rent [fact]
Investment-grade hyperscaler anchoring ~$5.5B of contracted revenue; diversifies away from Fluidstack but concentrates the book on a second single counterparty [fact]
$2B 6.125% bond + $200M revolver + project debt funding a multi-hundred-MW build against ~$35M/qtr revenue; a risk-off window or AI-capex scare throttles the build [fact/estimate]
All flagship capacity is Texas; energizing contracted MW is gated by grid interconnect, power delivery, and regulatory scrutiny [fact]
Odessa (~207MW) still mines to ~end-2027 and ~$76M BTC sits on the balance sheet; a BTC crash hits residual cash flow and the treasury during the wind-down [fact]
Advantages
- Fully-contracted flagship — entire 300MW Barber Lake leased (Fluidstack/Google) + 300MW Black Pearl leased (AWS) + a third IG-hyperscaler campus = ~$11.4B contracted HPC revenue [fact]
- Owns the scarce asset (energized, interconnect-queued Texas power) and retains full project ownership of the campuses [fact]
- Google backstop (~$1.73B) + ~5.4% warrants validate the anchor counterparty with a blue-chip credit and equity signal [fact]
- Tenant diversification — AWS (IG) and a third IG hyperscaler reduce reliance on any single neocloud counterparty [fact]
- Near-term (2026) delivery on Texas sites converts backlog to revenue faster than greenfield peers [fact]
Weaknesses
- Fluidstack neocloud counterparty risk - Google's backstop covers project debt, not a Google tenancy [fact]
- Deeply loss-making now (-$114.3M Q1 FY26) on collapsing mining revenue ($34.8M), with HPC lease revenue not yet recognizing [fact]
- Single-region concentration in Texas (ERCOT grid/weather/interconnect/regulatory exposure) [estimate]
- Compressed 2026-2027 delivery timeline across multiple hundred-MW campuses is meaningful execution risk [estimate]
- Heavy capital intensity + leverage ($2B 6.125% bond + facilities) against tiny current revenue; refinancing dependency [fact]
- More hybrid than the cleanest pivots (keeps mining longer), delaying the full datacenter-multiple re-rating; residual hash-price exposure [estimate]
Bottlenecks
- Power energization + grid interconnect — converting contracted MW at Barber Lake/Black Pearl to revenue is gated by ERCOT interconnect and power delivery on a compressed 2026-2027 timeline [fact]
- Datacenter construction execution — standing up multiple hundred-MW campuses (air + liquid cooling to the racks) on schedule is the swing factor for backlog-to-revenue conversion [estimate]
- Capital access — the build cannot outrun the bond + facilities + cash; a rate/risk-off shock throttles the pace [fact/estimate]
- Counterparty payment timing — contracted revenue only becomes cash when Fluidstack/AWS/the third tenant take capacity and pay on schedule [fact]
- Skilled datacenter operations talent to run high-density GPU campuses at contracted SLAs [estimate]
Top signals & trends
Top signals
On-time energization and the first lease revenue landing later in 2026 is the key bull proof; any slip is the near-term bear trigger [fact]
Fluidstack paying reliably (or Google stepping further in) de-risks the anchor lease; neocloud stress is the core credit risk [fact/estimate]
A fourth IG-anchored lease would extend visibility and validate the ~4.2GW runway; the third lease already improved tenant diversification [company-disclosed]
Heavy leverage against ~$35M/qtr revenue; watch bond spreads, revolver draws, and any new project debt into a risk-off window [fact]
Cleaner/faster exit removes hash-price volatility and speeds the re-rating; a slower runoff keeps residual BTC beta [fact]
As a high-beta miner-pivot, CIFR reprices fast on any hyperscaler capex guide-down independent of its own execution [estimate]
Trends
Drives demand for energized, interconnect-queued MW — the whole miner-pivot re-rating rests on this persisting [fact]
The structural mispricing Cipher rides: volatile ~$3-5M/MW/yr hash-price revenue converts to contracted 10-15yr leases at higher, richer multiples [fact]
Google's Fluidstack backstop de-risks the anchor lease (positive), but the underlying tenant is still a neocloud whose credit IS the thesis (risk) [fact]
Scarcity supports lease economics for those who already hold energized power, but interconnect/regulatory scrutiny gates energization [fact]
Heavy capex ($2B bond + facilities) drives losses during the build (-$114.3M Q1 FY26); financing risk if AI-capex sentiment turns [fact]
Retaining project ownership captures development margin but adds technology-obsolescence exposure on the built infrastructure over the lease life [estimate]
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Energized, interconnect-queued Texas power (Odessa/Barber Lake/Black Pearl) is Cipher's core input and scarce asset
GPU supplier to the tenants (Fluidstack/AWS) whose demand fills Cipher's campuses; the accelerator whose economics drive the AI-DC lease market
Datacenter construction + cooling (liquid/air) vendors Power, cooling, and thermal-management infrastructure to deliver high-density GPU campuses on the compressed 2026-2027 timeline (e.g. Vertiv-class suppliers)
Debt capital providers (bond + revolver banks) $2B 6.125% bond investors + $200M revolver syndicate (Morgan Stanley-led; Goldman, JPMorgan, Wells Fargo, Santander, SMBC) fund the build
running off) Historical ASIC supply for the Odessa mine, now being wound down toward ~end-2027 exit
Anchor HPC tenant leasing the entire 300MW Barber Lake site (10-yr, ~$3.8B initial / up to ~$9.0B w/ extensions); a private neocloud - Google backstops ~$1.73B of its obligations
15-yr, 300MW lease at Black Pearl (~$5.5B); investment-grade hyperscaler diversifying the tenant book, delivery from Jul 2026
Undisclosed investment-grade hyperscaler (third campus) 15-yr lease signed Q1 FY26 (Mar 2026) for a third data-center campus - further tenant diversification
Not a direct tenant but backstops ~$1.73B of Fluidstack's lease obligations and holds warrants for ~5.4% of CIFR - a credit/equity underwriter of the anchor lease
Cipher's structural twin - the other Google-backstopped Fluidstack counterparty (~360MW Lake Mariner, Google ~14% equity); TeraWulf is a cleaner/fuller pivot (HPC revenue already overtook mining) whereas Cipher stays more hybrid [fact]
Best-anchored miner-pivot - direct Microsoft (~$9.7B) + Nvidia (~$3.4B) book and an owned-GPU AI Cloud model rather than shell-leasing; Australia-domiciled/US-listed (non-US flag). Competes for the same AI-DC capital and multiple [fact]
Largest contracted HPC backlog of the pure pivots (~590MW/>$10B to CoreWeave); stayed independent after shareholders rejected CoreWeave's buyout - the CoreWeave-counterparty cluster analog [fact]
Purpose-built AI-factory landlord (~400MW Ellendale ND to CoreWeave + a ~$5B IG-hyperscaler lease), pursuing REIT conversion; a diversifying-tenant peer competing for hyperscale leases [fact]
The neocloud tier Cipher's tenants (e.g. Fluidstack) resemble; they compete for the same hyperscaler/AI-lab demand and set the reference economics for GPU hosting [fact]