Cipher Digital Inc. (NASDAQ: CIFR) — A Miner Reborn as a Landlord, Already Priced for the Whole Pipeline

Independent fundamental research. Report date: 2026-06-13.

⚡ The Author’s Take

This block is the author’s own independent opinion and general information only — not investment advice. Everything below it (the analytical body) is deliberately position-free and carries no price target.

Verdict: AVOID at $24.50 / accumulate-on-weakness only / NOT a short. Fair-value zone ~$12–18 per share on the contracted book plus a modest pipeline credit; the stock already pays close to the bull case. Conviction: medium.

Cipher has done something genuinely impressive: in roughly eight months it converted a fading, fully-commoditized Bitcoin-mining shell into a credible, institutionally-financed hyperscale data-center developer with three signed long-term leases, ~907 MW of operating-and-contracted capacity, ~$5.2B of investor-demanded project bonds, and — uniquely in the converted-miner cohort — a direct, no-strings Amazon Web Services anchor that is arguably the single highest-quality tenant relationship any of its peers (WULF, IREN, APLD, CORZ) has secured. The operational execution at Barber Lake and Black Pearl is real, the West Texas/ERCOT power position is hard to replicate quickly, and the GAAP losses that scare retail (-$822M in FY2025) are ~80% non-cash derivative and impairment optics, not cash bleed. This is not a fraud and not a meme; it is a real business being built well.

But the price already discounts the bull case. At an EV of ~$14B the market is paying roughly a 5–6% implied cap rate on management’s own unaudited ~$787M average annual NOI projection — a figure that is not yet earning a dollar, sits behind 6–7% non-recourse project debt that eats the first ~$250M of it, and rests on leases that have not commenced. On a fully-diluted (~560M shares after ~155M of in-the-money converts and Google penny-warrants), net-debt-adjusted, contracted-only basis, the equity is worth roughly $12–18; getting to today’s $24.50 requires capitalizing a 3.3 GW pipeline that is unsigned, unfunded, and arriving exactly as a tidal wave of capital floods the same trade (the late-capital-cycle tell). The framing is “great execution, wrong price” — a momentum darling priced for a flawless multi-year pipeline conversion. I would not short it (15.7% of the float is already short, every covering analyst rates it Buy, the behind-the-meter and pipeline catalysts are live, and squeeze risk is acute), but I would not pay up here. Flips bullish: a meaningful drawdown into the low-to-mid teens, OR an investment-grade refinancing of the SPV bonds that proves the levered spread to the common. Flips bearish: a slipped Barber Lake/Black Pearl commencement date, an ERCOT batch-process disappointment, or any crack in the unnamed third tenant’s credit. Tag: “They built the house beautifully — then sold you the whole subdivision at retail.”

1. Executive Summary

Cipher Digital Inc. (NASDAQ: CIFR), renamed from Cipher Mining Inc. in February 2026, is a New York–headquartered developer and operator of industrial-scale data centers. It is roughly two-thirds of the way through a fast, deliberate metamorphosis: from a single-site Bitcoin miner (the Odessa, Texas facility) into a vertically-integrated landlord that builds and leases hyperscale AI/HPC data centers to investment-grade tenants under 10-to-15-year contracts. The company came public via a SPAC (Good Works Acquisition) in August 2021, sponsored by the Bitfury group; it runs lean (~66–85 employees) and reports on a December fiscal year.

The investment debate is unusually clean. What is real: three signed campus leases — Barber Lake (300 MW gross / 207 MW critical IT, leased to FluidStack with a Google backstop), Black Pearl (300 MW, leased directly to Amazon Web Services), and a third 100 MW lease at the Stingray site (an unnamed investment-grade hyperscaler) — totaling ~907 MW of operating-and-contracted capacity and, per management, ~$11.4B of contracted revenue and ~$787M of average annual net operating income (NOI) from October 2026 through September 2036. Behind them sits a ~3.3 GW interconnection pipeline (mostly West Texas/ERCOT, plus one Ohio/PJM site). The company has financed the build with ~$5.2B of mostly non-recourse, project-level senior secured bonds that trade above par — a powerful third-party validation — plus $1.47B of corporate convertibles and a $200M undrawn bank revolver.

What is unproven: as of the March 31, 2026 balance sheet, no HPC lease revenue has been recognized. All ~$224M of FY2025 revenue is legacy Bitcoin mining, now declining as the Odessa fleet winds toward a ~2027 shutoff. The ~$11.4B/$787M figures are unaudited investor-presentation metrics with no corresponding ASC 842 lessor schedule in the filings. The company is deeply free-cash-flow negative (~-$696M FY2025), faces a ~$249M+ annual cash-interest wall that the not-yet-commenced leases must cover, and has diluted shareholders ~64% since the SPAC with another ~155M shares (~38% of the count) of in-the-money converts and Google penny-warrants overhanging. Insiders have never bought a share on the open market, and the Bitfury sponsor has been a relentless seller (still ~14.6%).

Valuation is the crux. CIFR trades at the 95th percentile of its own ~10-year P/S and P/B history, at ~$23–28M EV per contracted MW (in line with the richest converted-miner comp, WULF) and a ~5–6% implied cap rate on management’s NOI target — at or below the cost of its own project debt. That leaves essentially all equity value at $24.50 resting on the pipeline converting and an eventual investment-grade refinancing. This memo takes no position; it lays out the embedded expectations and the falsification tests. The single, fenced-off opinion is in the author’s-opinion block above.

2. Business Overview

What the company does. Cipher Digital develops, finances, builds, and operates large-scale data centers in the United States. It is vertically integrated across the value chain that matters most in 2026: power origination (securing grid interconnection rights and long-term power), engineering and construction management, and operations. The strategic identity, restated at the February 2026 rebrand, is “Built for Hyperscale” — purpose-built to deliver power-dense, large-footprint facilities to the exacting specifications of hyperscale and AI tenants, fast.

The business today is a barbell of one declining legacy and three ramping growth assets:

• Legacy — Odessa Bitcoin mine (Permian Basin, TX). 207 MW operating, generating ~11.6 EH/s at ~17.2 J/TH fleet efficiency, mining ~346 BTC in Q1-2026. Its competitive edge is a fixed-price power purchase agreement with Luminant at ~$0.028/kWh, among the lowest in the industry, locked until roughly July 2027. Odessa is self-funding, throws off “several million dollars” of positive cash per month, and is being deliberately wound down — no further mining capex is planned. Management expects Bitcoin to be “immaterial” to financials well before a full shutdown by end-2027.

• Growth — three HPC campus leases:

  • Barber Lake (Coke County, TX): 300 MW gross / 207 MW critical IT load; leased to FluidStack (a GPU-cloud/“neocloud” operator) under a 10-year term, with a Recognition Agreement from Google LLC that backstops the lease (Google can assume the lease or pay a termination fee on a FluidStack default). Targeted revenue commencement: ~September/Q4 2026. Construction is the most advanced — the ~800,000 sq ft structure “topped out” in April 2026, design is 100% complete, ~99% of equipment procured.
  • Black Pearl (Andrews County, TX): ~300 MW gross; leased directly to Amazon Web Services under a 15-year term. The site is a retrofit/expansion of a former Bitcoin mine (decommissioned February 2026). Phase I ~93% procured, Phase II ~80%; targeted commencement Q4 2026.
  • Stingray (Andrews County, TX): 100 MW gross; a 15-year lease signed Q1-2026 with an unnamed investment-grade hyperscaler. Target energization Q4 2026; financed by the June-2026 $810M bond.

• Pipeline: ~3.3 GW of additional grid capacity across Reveille (Cotulla, TX, 70 MW, interconnect-approved), Ulysses (SE Ohio, 200 MW, the first PJM site), and larger 2028-energization sites (McLennan, Mikeska, Colchis — each progressing through the ERCOT “batch 0” interconnection process), plus a 500 MW upsize option at Barber Lake. Total portfolio ambition: up to 4.2 GW of grid power.

How it makes money — today vs. tomorrow. Today, revenue is 100% Bitcoin mining: the company earns block rewards and transaction fees, recognized under ASC 606 with Bitcoin as non-cash consideration. This revenue is volatile (Bitcoin price × network difficulty × uptime) and shrinking by design. Tomorrow, the model flips to contracted rent: long-term leases under which Cipher delivers powered, built-out data-center shells (and, in the “gross modified” structure, also supplies power, operations, maintenance, and security) to hyperscale tenants. That converts a commodity-price-taking cash flow into (in management’s framing) “durable, high-quality, long-term” contractual NOI. The recurring-revenue quality therefore improves dramatically on paper — but the recurring stream has not started, and the precise revenue-recognition classification (operating lease under ASC 842 vs. a service arrangement under ASC 606, with power/O&M unbundled) is not yet settled in the filings. Verdict: a real, well-articulated business model transition from price-taking commodity production to contracted infrastructure rent — but as of this report the new model is a set of signed contracts and construction sites, not yet a revenue line.

The contract architecture matters more than the headline MW. Each lease is housed in a bankruptcy-remote project subsidiary (Cipher Compute LLC for Barber Lake, Black Pearl Compute LLC, Stingray Compute LLC), which both isolates risk and is the unit against which the project bonds are issued — so the “company” is best understood as a holding entity sitting atop a small portfolio of single-asset, separately-financed projects, plus the Odessa mine and the pipeline option. The leases run 10 years (Barber Lake) to 15 years (Black Pearl, Stingray) of initial term, typically with extension options and contractual escalators (the per-asset escalator schedules are undisclosed). Because the structure is “gross modified” — Cipher provides not just the shell but power, operations, and maintenance — a portion of each lease payment is effectively a service fee tied to Cipher’s own cost to operate, which introduces a margin variable (power cost, staffing) that a pure triple-net REIT lease would not have. This is why the ~$787M figure is described as net operating income rather than rent: it is already net of the operating costs Cipher bears. The economics are therefore better than a triple-net comparison would suggest on the revenue line but carry operating leverage (and operating risk) a pure landlord would not.

Customer types and end markets. The end market is singular and concentrated: hyperscale and AI compute demand, intermediated through three counterparties — a Google-backstopped neocloud (FluidStack), Amazon Web Services directly, and an unnamed investment-grade hyperscaler. There is no diversification across industries, geographies (ex-the one Ohio site), or customer size; the entire forward revenue base is three tenants buying the same thing (powered AI data-center capacity) in the same window. That concentration is the inevitable shape of a hyperscale-development business — these are the only customers that take 100–300 MW blocks — but it is a genuine fragility relative to a diversified colocation REIT serving hundreds of enterprise tenants.

3. Industry Dynamics

The demand backdrop is as strong as any in the market. Hyperscaler AI capital expenditure is running at a ~$700B/year pace across the largest cloud and model builders, and the binding constraint has shifted decisively from chips to power and the data centers to house it. Interconnection queues, multi-year transformer and switchgear lead times, and local grid limits mean that whoever controls energized (or near-energized) large-block power at speed controls a scarce, monetizable asset. The “power is the bottleneck” thesis is, in this analysis, correct and durable for at least the next several years.

Supply is where the caution lives. The economics are visible enough ($120–135/kW/month colocation rents, high-80s% NOI margins on stabilized assets, asset-level IRRs frequently quoted in the high-teens-to-20s%) that capital is flooding in from every direction: the converted Bitcoin miners (WULF, IREN, APLD, CORZ, RIOT, MARA), purpose-built neoclouds (CoreWeave, Crusoe, Nebius), the established colocation REITs and privates (Digital Realty, Equinix, Switch, Vantage, QTS), private-equity infrastructure vehicles, and — most importantly — the hyperscalers self-building. The hyperscaler is simultaneously CIFR’s best customer and its largest potential competitor. In capital-cycle terms, the industry sits mid-to-late cycle: demonstrably high returns are pulling in a wave of capital that will, on a multi-year horizon, compress rents and returns. The current ~$130/kW/month and high-80s% NOI are best understood as cyclical-high economics, not a permanent plateau.

Regulatory and power structure. CIFR’s deliberate ERCOT (West Texas) concentration is a double-edged sword. Upside: ERCOT’s interconnection process is faster and cheaper than PJM’s notoriously clogged queue, Texas is politically data-center-friendly, and West Texas sits atop abundant cheap natural gas (the behind-the-meter opportunity). Downside: ERCOT is an energy-only, weather-volatile grid (Winter Storm Uri precedent), exposed to curtailment, basis risk, and an evolving large-load regulatory regime (the PUCT/ERCOT “batch process” and large-load interconnection rules are still being finalized — a live source of timeline uncertainty for the 2028 sites). The single Ohio/PJM site (Ulysses) is intentional geographic and grid diversification, valued by multi-market hyperscalers, but PJM carries higher deposit requirements and a tougher queue.

Barriers to entry are real but replicable: long-lead power origination, interconnection rights, speed-to-build execution, and access to project capital. None is a structural monopoly; each is a head-start that a well-funded competitor can close in 18–36 months.

The capital-cycle lens is the single most important analytical frame here, and it cuts against the equity. The capital-cycle discipline says: follow the capital, not the demand. When returns in an industry are visibly high, capital floods in on the supply side, capacity overshoots, and returns mean-revert — and the returns look best precisely at the top, when everyone can see them and the capital is still arriving. Mid-2026 is a textbook expression of that pattern in data-center development: dozens of converted miners, neoclouds, REITs, sovereign funds, and PE infrastructure vehicles are racing to lock up power and build, debt and equity capital is being raised in the tens of billions, and the marginal economics (rent/kW, NOI margin, asset IRR) are being quoted at levels that assume the scarcity persists. History (telecom fiber ~2000, shipping ~2007, shale ~2014, the prior crypto-mining boom ~2021) says it usually does not. The signal to watch is not demand — demand is real and may stay strong — but the spread between new-build cost and achievable rent, and the cap rate the market pays for contracted NOI. When that cap rate compresses below the cost of the debt funding the build (as it arguably has for CIFR’s whole-EV), the cycle is pricing perfection. None of this means CIFR’s projects are bad — asset-level IRRs on already-signed leases can be excellent — but it does mean the industry-wide return on the next wave of capital (the pipeline the market is capitalizing) is more likely to disappoint than to surprise.

Verdict: a structurally attractive industry on the demand side and a structurally crowded, late-capital-cycle one on the supply side. It is a good industry for an incumbent that already controls energized power and signed tenants, and a more dangerous one for fresh capital — including the pipeline value embedded in CIFR’s equity — deployed at mid-2026 valuations.

4. Competitive Position

Name the moat — or its absence. Applying the Greenwald barriers-to-entry taxonomy, CIFR does not possess a durable competitive advantage. It possesses a head start plus the best tenant-credit mix in its cohort.

• Cost / supply advantage (partial, replicable): in-house power origination and a ~3.3 GW interconnection pipeline assembled over years of on-the-ground sourcing is a genuine access edge, and the Odessa $0.028/kWh PPA is a best-in-class legacy cost position. But interconnection rights and construction speed are acquirable — there is no proprietary technology, patent, or regulatory franchise locking competitors out.
• Customer captivity (contractual, post-signing only): once a 10–15-year lease is signed with a hyperscaler, switching costs are high and the cash flow is sticky — but this captivity exists only on the ~907 MW already contracted. The 3.3 GW pipeline that the market is capitalizing has zero captivity; every megawatt of it must still win a competitive tenant negotiation.
• Economies of scale / network effects: absent. At ~907 MW contracted, CIFR is sub-scale versus the REITs and has no network effect (data centers are not a two-sided network in this colocation model).

Where CIFR genuinely differentiates — and it matters — is tenant credit. The direct AWS lease at Black Pearl is arguably the single highest-quality anchor in the entire converted-miner group. It rests on Amazon’s investment-grade balance sheet, required no equity give-up (no warrants), and was financeable with a $2.0B bond on its own merits. This is materially cleaner than APLD’s reliance on CoreWeave (a ~BB-standalone, ~67% Microsoft-concentrated neocloud) or the fragile-neocloud exposure elsewhere in the cohort. By contrast, the Barber Lake/FluidStack anchor is lower quality and expensive: FluidStack is a neocloud whose lease had to be backstopped by Google, and Cipher paid for that backstop in equity — 24,178,576 penny-warrants (~5.6% of the company) with a $430M value floor that ratchets additional shares/cash to Google if the warrants are worth less at exercise. CIFR effectively rented Google’s balance sheet and handed over real dilution to do it. The third (Stingray) tenant is investment-grade per management but unnamed and unverifiable — a concentration and disclosure gap.

Head-to-head: versus WULF (the closest twin — a miner-turned-landlord also backstopped by Google via FluidStack), CIFR is comparable on model and arguably ahead on the AWS relationship; versus IREN and CoreWeave (neoclouds that own the GPUs and bear the depreciation/obsolescence risk), CIFR’s landlord model is structurally lower-risk — it does not own the rapidly-depreciating compute; versus APLD, CIFR’s anchor credit is superior.

The landlord-vs-neocloud distinction is worth dwelling on, because it is the most important risk fork in the whole cohort and CIFR is on the better side of it. A neocloud (CoreWeave, IREN, Crusoe) buys the GPUs — tens of thousands of accelerators that depreciate on a 3–6-year curve, face technological obsolescence with each NVIDIA generation, and must be kept utilized to service the debt that bought them. That is a capital-destruction machine if utilization or pricing slips. A landlord (CIFR, WULF, APLD) builds the shell and the power and lets the tenant bear the GPU risk; the landlord’s asset is the building, the power interconnection, and the lease — far longer-lived and not subject to Moore’s-law obsolescence. CIFR has been explicit and disciplined about staying on the landlord side (“when you can get very elevated lease rates, we favor the colocation business… you have a fully paid for asset at really attractive returns”), with only a small, credit-supported “test kitchen” experiment in compute ownership at the sub-scale Reveille site. That discipline is a genuine quality marker. The flip side: the landlord captures less upside if AI compute economics stay euphoric, and the landlord’s terminal value still depends on the building being re-leasable after the initial 10–15-year term — management’s “West Texas terminal value the market doesn’t appreciate” claim is an unproven assertion about residual values two-plus decades out.

Run the Greenwald tests explicitly. Market-share stability: there is no stable share to measure — this is a new, fast-growing, fragmenting market where positions are being established by who signs leases this year, the opposite of the stable-share signature of a real moat. ROIC test: not yet measurable (pre-revenue on HPC); the legacy mining ROIC was unremarkable and declining. Barriers-to-entry test: the barriers (power, interconnection, capital, speed) are access advantages a determined, well-capitalized entrant can replicate, not the demand-captivity or proprietary-cost barriers Greenwald requires for a true franchise. CIFR passes none of the three cleanly. Verdict: a credible head start with the best counterparty credit in the converted-miner cohort, and the lower-risk landlord posture — but a replicable land-grab, not a durable franchise. Three-tenant concentration, an unnamed third counterparty, and unproven terminal values are real fragilities the market is not currently pricing.

5. Growth History and Forward Opportunities

History. Revenue grew from ~$3M (2022) to $126.8M (2023), $151.3M (2024), and $223.9M (2025) — but this is legacy mining growth driven by deploying more megawatts of ASICs, and it is now reversing: Q1-2026 revenue fell -29% YoY to $34.8M as Black Pearl’s mining fleet was decommissioned (for the AWS retrofit) and Odessa wound down. So the reported top line is shrinking precisely as the strategic story is accelerating — a discontinuity that makes trailing financials nearly useless for valuation.

The forward opportunity is the entire thesis. Management’s contracted-NOI bridge is the central artifact: the three signed leases are projected to generate ~$787M of average annual NOI from October 2026 to September 2036, rising to ~$892M by 2035, against ~$11.4B of cumulative contracted revenue. Layered on top:

• Pipeline conversion (~3.3 GW): Reveille (70 MW) and Ulysses (200 MW) are the near-term, interconnect-approved opportunities in “active and advanced” tenant discussions (energization 2027). McLennan/Mikeska/Colchis (~2.5 GW combined) target 2028, contingent on a favorable ERCOT “batch 0” interconnection outcome expected mid-2026. Plus a 500 MW Barber Lake upsize.
• Behind-the-meter generation (the upside-convexity option): management calls on-site natural-gas generation “the most upside convexity potential in our stock” — West Texas sites sitting “above an entire ocean of natural gas” could in theory power gigawatts off-grid, bypassing interconnection queues entirely. It is also, by management’s own admission, an unsolved engineering/financing/permitting challenge (“nobody but Elon has done it”). Real option, early stage, zero in the numbers.
• Selective compute ownership (the “test kitchen”): at the smaller Reveille site, management is weighing participating in GPU ownership for credit-supported neoclouds — a higher-return, higher-risk pivot away from the pure-landlord model. A small experiment, but a signal that the disciplined “we prefer colocation” line is not absolute.
• Odessa conversion: the 207 MW Odessa site, with its cheap PPA as a bargaining chip, could itself be converted from mining to an HPC campus (as Black Pearl was) — optionality, not a plan.

Unit economics, as far as they can be reverse-engineered. Management disclosed two anchors: ~$11.4B of contracted revenue and ~$787M average annual NOI over the ~10-year window, against ~507–607 MW of critical-IT load across the three leases. That implies roughly ~$1.3–1.5M of annual NOI per critical-IT MW, or on the order of ~$110–130/kW/month — consistent with premium colocation economics observed across comparable assets (WULF, APLD), and management confirmed on the Q1 call that the incremental Stingray lease priced at a higher NOI/MW than the first two (“pricing continues to trend in a positive direction”). On the call, management also flagged that pricing power is a function of speed-to-availability — an energized or near-energized site commands a premium because hyperscalers are power-starved and time-constrained — which is precisely where CIFR’s in-house construction speed converts into rent. The bull reads this as proof the model compounds: each delivered site de-risks the next and supports higher pricing (the “flywheel”). The skeptic notes that premium pricing is itself a cyclical-high phenomenon and that the disclosed NOI is a portfolio average hiding per-asset spreads, escalators, and tenant-specific terms that are not in the filings.

Quality of growth. If delivered, this is high-quality growth: long-duration, contracted, investment-grade-counterparty cash flow replacing volatile commodity production. But the quality is entirely prospective. The growth is capital-intensive (FCF deeply negative through commencement), debt-and-equity-funded, and gated by construction execution, interconnection approvals outside the company’s control, and tenant negotiations not yet won. And it is worth stating plainly that the reported revenue line will likely keep falling through most of 2026 — as mining winds down faster than HPC rent ramps — before it inflects, so investors anchoring on near-term reported growth will see the opposite of the story until lease commencement flips the trend (a setup that can pressure the stock even if the thesis is intact). Verdict: potentially very high-quality growth, but unproven and front-loaded with execution and funding risk — the single largest gap between narrative and realized cash flow in this report.

6. Financial Quality

The headline loss is mostly accounting optics — read past it. FY2025 net loss was -$822.2M and Q4-2025 alone was ~-$734M, figures that on their face look catastrophic against $224M of revenue. The quality-of-earnings work shows ~80% of the loss is non-cash and non-recurring:

Roughly ~$650M of the -$822M is non-cash/one-time. Management’s adjusted earnings (excluding SBC, D&A, fair-value swings, impairments, disposals) were +$46.2M / +$106.7M / +$22.2M for FY2023/24/25 — i.e., the legacy mining business was modestly profitable but declining. The correct conclusion: do not capitalize the GAAP loss, but do not mistake the non-cash nature of the loss for financial strength either. The real financial profile is a pre-revenue infrastructure build burning cash.

Cash flow tells the true story. Operating cash flow was -$94M (2023), -$88M (2024), and -$208M (2025); capex exploded to $488M (2025), producing FCF of roughly -$696M for FY2025 and another ~-$462M in Q1-2026 alone. The Q1-2026 reported +$92M operating cash flow is a mirage — it reflects an +$83M jump in construction-related accounts payable (AP rose 5x to $198M) plus $59M of non-cash interest add-back; underlying operating cash is negative.

Balance sheet — the heart of the matter. Total assets ballooned from $855M (12/31/24) to $4.29B (12/31/25) to $6.39B (3/31/26). But the composition demands scrutiny:

~$3.5B of the cash is restricted construction escrow — ring-fenced at the project SPVs and dedicated to completing Barber Lake and Black Pearl; it is not freely available corporate liquidity. Net of restricted cash, the company has ~$715M of unrestricted cash and BTC against ~$4.7B+ of gross debt (and ~$5.2B+ including the post-quarter Stingray bond). Book equity of $714M against a ~$10B market cap means the stock trades at ~13–14x book — the 95th percentile of its own history.

The interest wall. The aggregate project + convert debt carries a cash-interest run-rate of roughly $249M/year pre-Stingray (and ~$298M+ including it) once notes fully accrue. FY2025 cash interest paid was only $1.5M (most GAAP interest is non-cash OID amortization, and construction-period interest is capitalized/funded from escrow) — but as the bonds season and the leases commence, that cash interest must be covered by HPC rent that is not yet flowing. This is the central financial risk: the timing gap between the interest obligation switching on and the lease revenue switching on.

Dilution and SBC. Shares outstanding grew from ~247.6M (YE2022) to ~409.0M (5/2026), ~64% dilution in under four years, ~$549.6M raised via ATM. Stock-based comp was $52.8M in FY2025 (23.6% of revenue) and $27.0M in Q1-2026 (an extraordinary 77.6% of the shrunken revenue base). On top of the count sit ~155M potential shares (~38%): the 2031 converts (~81M @ $16.03), 2030 converts (~39M @ $4.45, deep in-the-money), Google warrants (~24M @ $0.01), and ~11M RSUs/PSUs. The capped calls that once hedged convert dilution are now exhausted (cap $23.32 < $24.50 price), so incremental appreciation dilutes fully.

Bitcoin accounting is now mark-to-market (ASU 2023-08): ~1,116 BTC carried at $76.2M fair value at 3/31/26 — below cost (~$101M), i.e., an unrealized loss — and being sold down to fund opex. Management has supplemented BTC sales with covered-call premium (selling calls above spot in Q1-2026), a sign of disciplined inventory management rather than forced liquidation, but the treasury is now small enough (~$76M, <1% of assets) to be immaterial to the thesis.

The debt architecture deserves a closer read, because it is both the strength and the risk. The genius of the structure is that the ~$3.7B of project bonds (Cipher Compute $1.7B + Black Pearl $2.0B, plus the off-balance-sheet-at-3/31 Stingray $810M) are issued by bankruptcy-remote subsidiaries, secured by each project’s assets and lease, and structurally non-recourse to the parent — so a single project’s failure cannot drag down the whole company, and the construction risk is ring-fenced. During construction, interest is funded from the dedicated escrow (the ~$3.5B restricted cash) and a debt-service reserve account (DSRA), which is why FY2025 cash interest paid was only $1.5M despite $36.6M of GAAP interest expense. The corresponding risk: those escrows deplete as construction proceeds, and once each project is built the bond’s interest must be serviced from that project’s lease NOI. The bonds amortize against the lease cash flows, aligning debt service with revenue — but only after the lease commences. The window between escrow depletion / construction completion and stabilized lease revenue is the precise pinch-point; management asserts both projects are “sufficiently capitalized through construction based on current estimate to complete,” which is a statement that must hold for the thesis and which a cost overrun or schedule slip would test. The $1.3B 0%-coupon 2031 convertible is a quietly important piece: it costs zero cash interest (all expense is non-cash OID amortization) and does not convert until $16.03 — well below the current $24.50 — so it is now deep in-the-money and will dilute, but it also represents ~$1.3B of nearly-free financing that materially lowered the blended cash cost of the build.

Verdict: economics will improve dramatically with scale if the leases commence and stabilize, but the present-day financial profile is a deeply FCF-negative, heavily-levered, serially-dilutive build with a looming cash-interest obligation and no HPC revenue yet earned. There is no going-concern qualification, the bonds trade above par, and the project financing is sized to complete construction — but the equity quality is speculative until rent flows, and the balance sheet has zero resilience to a demand or capital-markets shock relative to the debt-free miner of two years ago.

7. Capital Allocation

A sophisticated, all-in, debt-funded bet — executed well, but with the equity bearing real costs. Three threads:

(1) Financing strategy — genuinely impressive, and shareholder-aware in structure. The pivot from a serially ATM-dilutive equity story to per-asset, non-recourse, project-level bonds is the right architecture: it isolates construction risk inside bankruptcy-remote SPVs (Cipher Compute LLC, Black Pearl Compute LLC, Stingray Compute LLC), aligns debt service with each lease’s cash flows, and — critically — the bonds were oversubscribed and trade above par (yields ~5.7–6.2% as of the Q1 call), an unusually credible third-party endorsement of the contracts and the build plan. The $200M committed bank revolver (Morgan Stanley, Goldman, JPMorgan, Wells Fargo, Santander, SMBC) is the first corporate-level facility in the converted-miner peer group and a real maturation marker. Management has repeatedly chosen callable, flexible structures over cheaper non-callable paper to preserve the ability to refinance toward investment grade as assets stabilize — a thoughtful, optionality-preserving choice.

(2) The equity costs embedded in the deals. Two giveaways temper the praise. First, the Google penny-warrants (24.18M shares at $0.01, with a $430M floor true-up) handed to make the FluidStack lease bond-financeable — ~5.6% of the company, real off-headline dilution. Second, the relentless ATM issuance (123.5M shares, ~$549.6M, 2023–2025) that funded the early build at low prices. The AWS deal, by contrast, was landed clean (no warrants) — the template for how this should be done. No buybacks (nor would they be appropriate for a cash-burning developer); no value-destructive M&A (site acquisitions have been disciplined, often from under-capitalized speculators).

(3) Incentive design — better than the sector, but untested and lightly endorsed. The 2026 proxy reveals a genuinely strategy-aligned bonus scorecard — the first year (FY2025) it was used — built on (i) non-GAAP operating margin, (ii) megawatts energized, (iii) approved/pipeline power-portfolio MW, and (iv) HPC utilization (MW under LOI/lease to hyperscalers). These are exactly the operating metrics a developer should be paid on, not “Bitcoin mined.” Long-term equity is 50% relative-TSR PSUs / 50% RSUs. The caveats are real: FY2025 paid out at maximum across the board (cash bonus 200% of target; PSUs certified at 225%), with the first PSU tranche vested early “to facilitate tax planning” — a program that has never faced a down year; and the first-ever say-on-pay vote passed with only ~72% support, well below the ~90% norm, signaling shareholder unease. The board is 7-of-8 independent (Lead Director James Newsome, ex-CFTC Chair; Caitlin Long of Custodia), and the legacy Bitfury board-observer arrangement was terminated in July 2025.

The insider signal is unambiguously negative. Across the entire ~164-filing Form 4 corpus, there is exactly one open-market purchase — a director’s ~$36,000 buy in September 2022 at ~$1.43. No officer (CEO Page, CFO Mumford, the Co-Presidents) has ever bought a share on the open market, even as the stock ran ~7x from $3 to $25. Every officer monetizes via option/RSU exercise and sell-to-cover; directors’ sales during the run were 10b5-1-planned but uniformly one-directional. Dominating all of it, the Bitfury sponsor complex (V3 Holding / Bitfury, ultimate owner Valerijs Vavilovs) has sold ~$600M+ of stock and is on Schedule 13D/A Amendment No. 23, still holding ~14.6% and still selling — a persistent supply overhang and the opposite of an insider conviction signal. Verdict: a sophisticated, well-structured financing program and unusually rigorous incentive metrics — undercut by maximum first-year payouts, a weak say-on-pay, embedded equity giveaways, and a complete absence of insider buying against a steadily-liquidating sponsor. Capital allocation is competent on the debt side and shareholder-diluting on the equity side; the jury on per-share value creation is still out.

8. Changes and Headwinds — Last Two Years

The last ~18 months are the story — this is a company that re-engineered itself in real time:

• Strategic pivot & rebrand (Feb 2026): Cipher Mining → Cipher Digital, formalizing the shift from Bitcoin miner to “Built for Hyperscale” data-center developer.
• Three hyperscaler leases in ~8 months (Sept 2025 – Q1 2026): Barber Lake/FluidStack-Google (Sept 2025), Black Pearl/AWS (Oct 2025), Stingray/IG-tenant (Q1 2026) — the proof-of-concept → proof-of-repeatability → proof-of-platform sequence.
• ~$5.2B of project financing raised in <12 months: Cipher Compute $1.7B 7.125% (Nov 2025), Black Pearl $2.0B 6.125% (Feb 2026), Stingray $810M 6.0% (June 2026), plus the $1.3B 0% convert (2031) and $172.5M 1.75% convert (2030), plus the $200M revolver (Q1 2026).
• Leadership: CFO transition — Edward Farrell retired and Greg Mumford appointed CFO (October 2025).
• Legacy wind-down: Black Pearl mining decommissioned (Feb 2026); WindHQ mining JVs sold to Canaan (Feb 2026); Bitcoin treasury being liquidated; Odessa on a ~July 2027 outside date.
• Governance: Bitfury board-observer agreement terminated (July 2025); first say-on-pay vote (~72% support, 2025).

Headwinds / watch-items: the ERCOT large-load “batch process” finalization (expected mid-2026) gates the ~2.5 GW of 2028 sites; a Bitcoin-price decline would pressure the still-mining Odessa cash flow and BTC carrying value during the transition; rising rates or any wobble in the AI-infrastructure narrative would hit both the cost of the next bond and the equity multiple; and the unnamed Stingray tenant’s credit is an open question. Verdict: the changes overwhelmingly advance the thesis — they are the thesis — but each also raises the stakes and the leverage; the company is far less financially resilient to a demand or capital-markets shock than it was as a debt-free miner two years ago.

9. Risk Analysis

Net: the dominant risks are timing (lease commencement vs. interest wall), pipeline conversion, refinancing, and dilution — a cluster that all resolves over the next 12–24 months, making this an unusually datable thesis. A total loss is unlikely given the non-recourse ring-fencing and above-par bonds, but a severe equity impairment is plausible in a demand-or-capital-markets shock given the leverage.

10. Valuation Discussion (Embedded Expectations)

CIFR is expensive against its own history and priced near the bull case on its forward economics. Trailing multiples are meaningless (all revenue is dying legacy mining; zero HPC revenue recognized), so the analysis must work off management’s contracted-NOI projection and the asset base — while flagging loudly that the key inputs are unaudited investor-presentation metrics, not filed lessor schedules.

Own-history and peer screens. CIFR trades at the 95th percentile of its ~10-year P/S and P/B history (composite valuation percentile ~95). On EV per contracted critical-IT MW, EV ~$14B / ~507–607 MW ≈ ~$23–28M/MW — in line with the richest converted-miner comp (WULF, ~$25M/MW) and ~1.5–3x the $8–15M/MW typical of operating capacity. The market is capitalizing a meaningful slice of the uncontracted 3.3 GW pipeline as if it were already signed and built.

The EV-per-MW comparison is the cleanest cross-sectional anchor. Stripping out the meaningless trailing P/S figures, the relevant peer metric is enterprise value per MW of contracted (or operating) capacity. CIFR at ~$23–28M/MW sits essentially on top of WULF (~$25M/MW) — unsurprising given they are near-identical models with the same Google/FluidStack backstop architecture — and at a premium to APLD (~$12–18M/MW). On the surface that APLD discount looks like CIFR is expensive; on inspection it partly reflects APLD’s reliance on a weaker anchor (CoreWeave, a ~BB-standalone, Microsoft-concentrated neocloud) versus CIFR’s direct-AWS lease, so some of CIFR’s premium is deserved tenant-credit quality. But “deserved relative to APLD” is not the same as “cheap” — both names, and WULF, sit at ~1.5–3x the $8–15M/MW that operating, stabilized third-party data-center capacity has historically commanded, because all three EVs capitalize a large slug of not-yet-built, not-yet-leased pipeline. The peer set is internally consistent and collectively rich; CIFR is the best-anchored member of an expensively-priced group, not a cheap outlier within a cheap group.

The implied-cap-rate crux. Capitalizing management’s ~$787M average annual NOI against the ~$14B EV implies a ~5.6% cap rate (~6.4% on the $892M-by-2035 figure). That looks roughly fair to slightly cheap versus stabilized data-center REITs (~5–7%) — until you remember three things: (1) the NOI is not yet earning a dollar and is unaudited; (2) it sits behind 6–7.125% non-recourse project debt that consumes the first ~$249M+/year, so the levered spread to the common is thin-to-negative at the current EV unless the bonds refinance toward investment grade; and (3) the implied cap rate itself embeds pipeline value, since ~$14B of EV exceeds a clean capitalization of the contracted book alone. In short, the whole-EV implied cap (~5.6%) sits at or below the cost of CIFR’s own debt — which means essentially all of the equity value at $24.50 is a call option on (a) the 3.3 GW pipeline converting to signed leases and (b) an eventual IG refinancing. Neither is yet in evidence.

Scenario analysis (illustrative; fully-diluted ~560M shares = 405M + ~120M converts + ~24M Google warrants + ~11M RSU/PSU; net debt ~$5.95B):

The current $24.50 sits between the base and bull cases, closer to bull. A sum-of-the-parts cross-check (contracted NOI ~$787M @ 6–7% = ~$11.2–13.1B; pipeline option risk-adjusted ~$1–2B; Odessa/BTC ~$0.1–0.3B; less ~$5.95B net debt; ÷560M FD) yields ~$11.6–17/share on the contracted book plus a modest pipeline credit — and the gap from there to $24.50 is the aggressive value the market assigns to an unsigned, unfunded pipeline.

A simple reverse-DCF frames how much pipeline is in the price. Take the contracted book at face value: ~$787M of average annual NOI, capitalized at a 7% cap rate (a fair stabilized data-center multiple, and roughly the cost of the project debt), is worth ~$11.2B of enterprise value. Net the ~$5.95B of net debt and divide by ~560M fully-diluted shares and the contracted book alone supports roughly $9–10 of equity value. To bridge from there to the $24.50 price, the market must be assigning ~$8–9B of additional enterprise value — i.e., on the order of $14–15 per share, or ~60% of the current price, to the uncontracted 3.3 GW pipeline and the prospect of an IG refinancing. Put differently: at $24.50 an investor is paying roughly book-plus for the three signed leases and the entire balance of the price for a pipeline that has not signed a single additional tenant. That can be the right call — power-constrained AI demand is real and CIFR’s pipeline is among the best-positioned — but it is unambiguously an option premium, not a margin of safety. The bull is buying the option; the bear notes that the option is being sold at the 95th percentile of the stock’s own valuation history, mid-capital-cycle, by insiders who have never bought it.

Embedded expectations — what must be true to justify $24.50: the three signed leases must commence on schedule and at the projected NOI; the ~3.3 GW pipeline must convert to signed IG-tenant leases at premium rents over the next 2–4 years (the ERCOT batch-0 sites must clear); the SPV bonds must refinance toward investment grade to open a levered equity spread; and the ~155M-share dilution overhang must not materially exceed expectations. That is a demanding, multi-variable, multi-year set of conditions — achievable, but priced as if largely de-risked. This memo states no target and no recommendation; the embedded expectations are simply demanding.

11. Variant Perception

Consensus view. Sell-side is uniformly bullish — 11 Strong Buy / 5 Buy / 0 Hold / 0 Sell, ~$32 average target. The Street narrative: a uniquely-positioned, vertically-integrated hyperscale developer with the best tenant credit in the converted-miner cohort, ~$11.4B of contracted revenue, above-par debt, and a 3.3 GW pipeline — “the year of execution” delivering on schedule.

The strongest bull case. CIFR is the cheapest landlord on EV-per-contracted-MW with the highest-quality anchor (AWS, direct, no equity give-up) in a power-constrained, demand-saturated AI buildout. The non-recourse, above-par project bonds prove institutional conviction in the contracts; the West Texas/ERCOT power position and in-house construction speed are a multi-year head start; behind-the-meter natural-gas generation is a free option on gigawatts of off-grid capacity; and the GAAP losses are pure non-cash optics over a modestly-profitable legacy business. If the pipeline converts and the bonds refi to IG, the equity is worth multiples of today.

The strongest bear case. The equity already pays the bull price. The ~$787M NOI is unaudited guidance, not yet earning, behind 6–7% debt that erases the levered spread at the current EV; the stock is at the 95th percentile of its own history and ~$23–28M/MW; ~38% dilution overhangs (capped calls exhausted); insiders have never bought and the sponsor is relentlessly selling; the third tenant is unnamed; and the whole trade is arriving mid-to-late in a capital cycle that historically compresses exactly these returns. A single slipped commencement date, a failed IG refi, or a neocloud-credit wobble re-rates the equity hard.

The 3–5 assumptions that matter most (and what falsifies each):

1. The three leases commence on time at projected NOI. → Falsified by a slipped Barber Lake/Black Pearl date or a tenant renegotiation in 2H-2026.
2. The 3.3 GW pipeline converts to signed IG leases. → Falsified by an ERCOT batch-0 disappointment or 2027 passing with no new signed lease.
3. The SPV bonds refinance toward IG, opening a levered equity spread. → Falsified by the next bond pricing wider, or rates/credit backing up.
4. Tenant credit holds (especially FluidStack and the unnamed third). → Falsified by a neocloud-sector credit event or disclosure that the third tenant is sub-IG.
5. Dilution stays within the ~155M overhang. → Falsified by a large new ATM/equity raise to plug a funding gap.

Where we land relative to consensus: the bull business case is largely correct and the execution is real — but consensus is under-weighting the price and the capital-cycle timing. The variant perception here is not “the company is bad” (it isn’t); it is “the equity is priced for a flawless, fully-converted pipeline that is neither signed nor funded, at the 95th percentile of its own valuation history, mid-cycle.”

12. Fact vs. Interpretation Table

13. Open Questions

1. Lease accounting: Will the HPC leases be classified as operating or sales-type under ASC 842, and how will the power/O&M “gross modified” components be unbundled (ASC 606)? This determines reported revenue/NOI shape once commencement hits.
2. Who is the third (Stingray) tenant, and what is its standalone credit? “Investment-grade hyperscaler” per management, but unnamed and unverifiable.
3. What are the actual NOI economics per lease (rent/kW/month, escalators, term-end residual)? The $787M is a portfolio average; per-asset spreads are undisclosed.
4. Will the SPV bonds refinance toward investment grade, and on what timeline? The entire levered equity-spread thesis hinges on it.
5. ERCOT batch-0 outcome (expected mid-2026): do the ~2.5 GW of 2028 sites get their interconnects?
6. Will CIFR actually pursue compute ownership at Reveille (a model shift), or stay a pure landlord?
7. Behind-the-meter: is there a concrete project, counterparty, and financing path, or is it still purely conceptual?
8. Convertible note holders’ identities (Coatue/SoftBank/other?) and their hedging/selling behavior — relevant to the dilution and supply overhang.
9. How much further ATM dilution is contemplated to bridge the pre-commencement funding gap?

14. What Must Be True (Bull and Bear)

Bull case — what must be true:

• The three signed leases commence on schedule (≈Q4-2026) and at the projected ~$787M aggregate NOI, with tenants performing.
• The ~3.3 GW pipeline converts to signed investment-grade leases over 2027–2029 (ERCOT batch-0 sites clear; Reveille/Ulysses sign).
• The SPV bonds refinance toward investment grade, opening a positive levered spread to the common; the next bond prices at/inside current terms.
• Dilution stays within the ~155M overhang; no large emergency equity raise.
• Falsification test: If, by year-end 2027, either Barber Lake/Black Pearl have not reached stabilized contracted NOI OR not a single new pipeline lease beyond the current three has been signed, the bull thesis is broken — the market will have paid a pipeline price for a stalled pipeline.

Bear case — what must be true:

• Lease commencement slips and/or the cash-interest wall (~$249M+/yr) opens a funding gap that forces dilutive equity or distressed refinancing.
• The pipeline stalls (ERCOT disappointment, rent compression from the capital flood) and the ~$23–28M/MW the market paid proves unsupportable.
• A tenant-credit event (FluidStack/neocloud stress, or the unnamed third tenant revealed sub-IG) impairs the contracted book.
• Falsification test: If, through 2027, the leases commence on time at projected NOI, at least one major pipeline site signs an IG tenant, AND the SPV debt refinances at or inside current spreads, the bear thesis is broken — the equity will have de-risked into its valuation.

The two falsification tests are near-mirror images, which is what makes CIFR an unusually datable security: the next 12–24 months of commencement dates, pipeline signings, and refinancing spreads will resolve the debate decisively.

15. Source Appendix

See Appendix B below for the full citation list. Key primary sources: FY2023/2024/2025 10-Ks (CIK 1819989); Q1-2026 10-Q (filed 2026-05-05); 2026 DEF 14A (filed 2026-04-20); the Form 4 corpus; Schedule 13D/A (Bitfury); the Q1-2026 earnings-call transcript (2026-05-05); and 8-Ks for the Barber Lake/FluidStack-Google, Black Pearl/AWS, Stingray, and financing events, all available on SEC EDGAR.

This analytical body carries no investment recommendation and no price target; the only position-taking content is the clearly-labeled author’s-opinion block at the top, which is the author’s own independent view.

APPENDIX A — Standard Diligence Questionnaire — Cipher Digital Inc. (NASDAQ: CIFR)

Supplemental to the research memo. Fact / Interpretation / Assumption labels applied where it matters. Report date: 2026-06-13.

General

What thoughtful questions have other investors asked about this company? From the Q1-2026 call and the analyst community: (1) Is lease pricing still trending up, and is the new Stingray NOI/MW better than the first two leases? (Management: yes — premium pricing holds for near-term-available sites.) (2) Will CIFR own its own compute rather than just lease (the “neocloud credit” question)? (Management: testing it only at the smaller Reveille site.) (3) The Odessa PPA — convert to HPC or keep mining? (4) Behind-the-meter natural-gas generation potential. (5) Will CIFR exit Bitcoin entirely, and by when? (Management: immaterial well before a ~2027 wind-down.) (6) How far ahead of energization will it sign leases? The unasked-but-critical question this memo presses: what is the per-share value after the ~155M dilution overhang and ~$5.2B debt, and is the unsigned pipeline already in the price?

Cyclicality & Earnings Nature

Cyclical high or low? Both, in different segments. Legacy Bitcoin-mining earnings are near a structural decline (deliberate wind-down), not a cyclical low. The HPC-landlord economics the market is paying for are arguably cyclical-high (rents/NOI margins at the top of a capital-flooded cycle — Interpretation). External vs. internal drivers? A mix: the AI/power demand wave is external and favorable; the lease-signing, financing, and construction execution are internal and so far well-executed. Revenue stability? Today: volatile (Bitcoin price × difficulty). Tomorrow: highly stable if the 10–15-year contracted leases commence — but they have not yet. Market size/outlook? Very large and growing (hyperscaler AI capex ~$700B/yr; power-constrained) — the demand backdrop is among the strongest in the market; the risk is supply/capital flooding in (a late capital cycle).

Business Quality & Competitive Moat

Industry more or less competitive? More — converted miners, neoclouds, REITs, PE, and hyperscaler self-build are all pouring capital in. Profitability (ROIC/ROE)? Currently meaningless/negative (ROE ~-121% TTM on non-cash losses; pre-revenue on HPC). Asset-level project IRRs are quoted high-teens-to-20s% if delivered (Interpretation). Industry profitability/barriers? Stabilized colocation NOI margins ~high-80s%; barriers (power origination, interconnection, speed, capital) are real but replicable head-starts, not structural monopolies. Easily understood? Reasonably — it is “build powered data-center shells and lease them to hyperscalers,” but the financing (non-recourse SPVs, converts, warrants) is complex. Undermined by foreign low-cost labor? No — US-domestic, power-and-location-bound assets. Do brands matter? No consumer brand; counterparty brand/credit (AWS, Google) is what matters, and CIFR’s is the best in its cohort. Nature of competition? Speed-to-energized-power and tenant relationships. Switching costs? High post-lease-signing (10–15-year contracts); zero on the uncontracted pipeline.

Financial Condition & Balance Sheet

Assets not on the balance sheet? The ~$11.4B contracted revenue / ~$787M NOI backlog is an off-balance-sheet (unaudited) value driver not yet recognized; the 3.3 GW interconnection pipeline is an intangible option. Off-balance-sheet liabilities? The Google warrant $430M floor true-up is a contingent obligation (partly captured as a warrant liability); SPV debt is on-balance-sheet but structurally non-recourse to the parent. Accounting conservatism? Mixed — fair-value/derivative accounting drives huge non-cash GAAP swings (use cash flow and adjusted figures); Bitcoin now marked-to-market (ASU 2023-08); no going-concern flag; auditor Marcum LLP. CapEx-hungry? Extremely — FY2025 capex $488M, FCF ~-$696M; this is a capital-intensity story financed by debt + ATM until leases commence.

Capital Allocation & Management

FCF generation & use? Deeply FCF-negative; all capital (debt + equity) is being deployed into the build. Acquisitions? Disciplined site/power acquisitions, often from under-capitalized speculators; no large M&A; sold the WindHQ mining JVs to Canaan (Feb 2026). Buybacks? None (appropriate for a developer). Issuing shares to insiders? Heavy equity comp (SBC 24% of FY2025 revenue, 78% in Q1-2026) and ~64% total dilution since SPAC + ~155M overhang; Google penny-warrants (~5.6%). Director/management comp? CEO Page ~$15.0M FY2025; metrics keyed to MW energized/leased, operating margin, and relative TSR (rigorous for the sector) — but FY2025 paid at maximum (bonus 200%, PSUs 225%) and say-on-pay only ~72%. Management motivation? Aligned via large equity stakes and operating-metric bonuses — but zero open-market insider buying ever, and the Bitfury sponsor relentlessly selling (~14.6%, 13D/A Amd. 23) is a clear negative signal (Fact).

Valuation & Market Data

ADR/MLP/K-1? No — a Delaware C-corp, common stock, no K-1. Dividend? None (and none appropriate). Profitability? GAAP-unprofitable (mostly non-cash); adjusted legacy mining modestly profitable but declining. Net income vs. cash from operations diverging? Yes, massively — GAAP NI -$822M FY2025 vs. operating cash flow -$208M, the gap being ~$650M of non-cash derivative/impairment items. Use cash flow and the debt schedule, not GAAP NI. Stock at 95th percentile of own ~10y P/S & P/B; ~$23–28M EV/contracted-MW; ~5.6% implied cap rate on management NOI (at/below its own debt cost).

Risks & Downside

What would cause the stock to decline? A slipped lease-commencement date; an ERCOT batch-process disappointment; a failed/wider IG refinancing; a neocloud/tenant credit event; a large dilutive raise; a Bitcoin or AI-narrative drawdown; rising rates. Catastrophic-loss risk? Plausible severe equity impairment in a demand/capital-markets shock given the leverage and the price already discounting the pipeline — though the non-recourse SPV structure ring-fences the parent from any single project default. Total-loss risk? Low — above-par bonds, real assets, IG anchor (AWS), and structural ring-fencing make a zero unlikely barring a systemic AI-infrastructure collapse plus failed refinancing (Interpretation).

Recent News & Events

Has the environment changed recently? Yes, dramatically and favorably for the narrative: three hyperscaler leases in ~8 months, ~$5.2B of project bonds raised (latest: $810M Stingray notes, June 2026), the Feb-2026 rebrand to Cipher Digital, the first corporate revolver, and a CFO transition (Mumford, Oct 2025). Accounting changes? Bitcoin fair-value adoption (ASU 2023-08); embedded-derivative reclassification to equity after the Oct-2025 charter amendment (the source of the $450M non-cash Q4 loss). New markets/facilities/management? First PJM site (Ulysses, Ohio); Black Pearl mining decommissioned; Odessa on a ~2027 outside date. The news tape is light and positive, dominated by the June-2026 Stingray financing — supportive but not thesis-changing.

APPENDIX B — Source Appendix — Cipher Digital Inc. (NASDAQ: CIFR)

Primary public sources first. Report date: 2026-06-13. Issuer CIK: 0001819989. All filings are publicly available on SEC EDGAR.

Primary — SEC filings (SEC EDGAR)

1. FY2025 10-K — filed 2026-02-24 (cifr-20251231.htm). Consolidated statements of operations, balance sheet, cash flows; Notes on convertible notes (16), embedded derivative, senior secured notes, warrants (18), PPA fair value, Bitcoin (ASU 2023-08), impairments/disposals, MD&A non-GAAP reconciliation. Auditor: Marcum LLP.
2. Q1-2026 10-Q — filed 2026-05-05 (cifr-20260331.htm). Q1-2026 vs Q1-2025 income statement; 3/31/26 balance sheet; debt schedule (2030/2031 converts; Cipher Compute $1.7B 7.125%; Black Pearl Compute $2.0B 6.125%); restricted cash; warrant liability; revolver covenants.
3. FY2024 10-K — filed 2025-02-25 (cifr-20241231.htm). Prior-year comparatives.
4. FY2023 10-K — filed 2024-03-05 (cifr-20231231.htm).
5. 2026 DEF 14A (proxy) — filed 2026-04-20 (cifr-20260420.htm). NEO compensation; FY2025 bonus scorecard (operating margin / MW energized / power portfolio / HPC utilization); PSU/RSU LTI design; board composition; say-on-pay (~72%); related-party (Bitfury observer agreement terminated 7/2025).
6. 2025 DEF 14A — filed 2025-04-21. Prior comp design.
7. Form 4 filings — the full insider-transaction corpus. Transaction codes (P/S/M/F/A/G); one open-market purchase (director Grossman, 9/8/2022, 25,000 sh @ $1.4287); officer exercise/sell-to-cover activity.
8. Schedule 13D/A — Bitfury complex (V3 Holding / Bitfury Top HoldCo; ultimate owner V. Vavilovs), Amendment No. 23 (~14.6% as of 6/2/2026, still selling). Schedule 13G — BlackRock (~5.4%), Jane Street (~5.2%).
9. 8-K filings — Barber Lake/FluidStack-Google lease & Recognition Agreement (Sept 2025); Black Pearl/AWS lease (Oct 2025); Stingray/3rd lease (Q1 2026); financing events (Cipher Compute Nov 2025; Black Pearl Feb 2026; revolver Q1 2026; Stingray $810M notes priced 6/8/2026); CFO transition (Oct 2025); rebrand to Cipher Digital (Feb 2026).

Primary — Earnings-call transcripts (company IR / public transcript services)

10. Q1-2026 earnings call — 2026-05-05 (Tyler Page CEO, Greg Mumford CFO). Source of: 907 MW operating+contracted; ~$11.4B contracted revenue; ~$787M avg annual NOI (Oct-2026–Sep-2036), ~$892M by 2035; 3.3 GW pipeline; capital-structure walkthrough (~$5.2B principal); Odessa $0.028/kWh PPA, 11.6 EH/s, 346 BTC; behind-the-meter and compute-ownership commentary; Q1-2026 results.
11. Q4-2025 earnings call — 2026-02-24. Rebrand context; Q4 loss drivers (embedded derivative + mining write-downs).
12. Earlier earnings calls and conference presentations (Q2-2022 → Q3-2025; Morgan Stanley TMT; Barclays) — strategic-pivot evolution.

Market & valuation data

13. Public market data (price ~$24.50, market cap ~$10B, EV ~$13–14B, short interest ~15.7% of float, beta ~3.2, analyst ratings/targets) and own-history valuation percentiles (P/B ~95th, P/S ~96th, composite ~95th vs. ~10-year history), as of mid-June 2026.
14. Peer comparison set for cross-checks: WULF (TeraWulf), IREN, APLD (Applied Digital), CORZ/CRWV (Core Scientific/CoreWeave), NBIS (Nebius), RIOT, MARA — publicly-listed Bitcoin-miner / HPC-datacenter / neocloud comparables.

Notes on source reliability

• GAAP net income is dominated by non-cash fair-value and impairment items; cash-flow statements, the debt schedule, and management’s adjusted reconciliation are the load-bearing sources, not headline NI.
• The ~$11.4B contracted-revenue and ~$787M NOI figures are unaudited investor-presentation metrics with no corresponding filed ASC 842 lessor schedule (no lease has commenced as of 3/31/26). Treated throughout as management hypothesis pending commencement.
• The June-2026 Stingray $810M 6.0% notes post-date the Q1-2026 10-Q and are sourced from the related 8-K.
• Analyst ratings/targets are third-party signals, never the basis for any valuation conclusion in this report.