Summary

Cipher Digital is a former Bitcoin miner betting its entire future on becoming a hyperscale data center developer. The company has 4.2 GW of grid-connected power across 10 sites, two signed HPC leases totaling 600MW (≈$9.3B contracted revenue with Amazon and Google per management's Feb 2026 presentation), $5.2B in debt, zero HPC revenue, 66 employees, and has never built a data center before.

The stock is up 295% in one year. Every analyst who covers it says Buy. No one says Hold or Sell. The 2031 convertible notes convert at $16.03 — three cents below the current price. The company reports a single operating segment: "Bitcoin Mining."

This memo is not a recommendation. It is a factual accounting of what the filings say, what the market implies, and where the two diverge.

I. What This Company Is

CIFR is a power-to-compute infrastructure play. The core asset is not mining equipment or data centers — it's connected electrical power in grid-constrained markets.

The business model:

1. Source large-scale power interconnections years ahead of demand
2. Mine BTC as interim monetization while building HPC (the bridge — now ending)
3. Build industrial-scale data centers on power-advantaged sites
4. Lease to hyperscalers under 15-year triple-net structures
5. Finance construction with non-recourse project debt backed by tenant credit

The value creation is in the spread between cost of power procurement and value of that power as data center capacity. Buy electricity at 2.8¢/kWh, sell it as racked compute capacity at $140-170/kW/month. Grid interconnection queues run 4-7 years — existing connections can't be replicated by throwing money at the problem.

Revenue streams today vs. tomorrow:

The entire thesis rests on this transition completing successfully.

II. The Contracts

Two signed HPC leases anchor the story. Everything else is pipeline.

Black Pearl — Amazon/AWS (300MW, Wink, Texas)

• 15-year lease, ≈$5.5B total contracted revenue
• Amazon.com parent guarantee on ALL rent and operating expenses for the full term
• Cost overrun protection: Amazon covers above $9.5M/MW cap
• $2.0B non-recourse project debt at 6.125% (issued Feb 2026)
• Construction: 5 sub-phases, first rack-ready Sept 30, 2026, fully ramped Feb 2027
• 3.0% annual rent escalator, triple-net structure, ≈100% NOI margin at project level

From the 8-K illustrative financials filed Feb 3, 2026:

These are illustrative, filed to market the bond offering, not audited projections. But they're the only project-level economics CIFR has disclosed. Total cash after debt service over 15 years: $2.7B. Simple average: ≈$181M/year, but the cash flow ramps significantly — from $29M in 2026 to $57M in 2027, rising to $127M+ by 2032 as amortization reduces the outstanding balance. Near-term distributable cash is well below the average. Lease support coverage ratio: 2.76-3.21x.

Barber Lake — Fluidstack/Google (300MW, Colorado City, Texas)

• 10-year lease, ≈$3B estimated total revenue
• Fluidstack USA II Inc. as direct tenant, Google LLC as backstop
• Google backstop activates AFTER rent commencement (critical limitation)
• $1.733B non-recourse project debt at 7.125%
• 50% excess cash flow sweep — traps half of free cash at the project level
• Phase I delivery: September 30, 2026 (244MW). Phase II: January 31, 2027 (56MW)
• Google received 24.2M warrants at $0.01/share ($525M liability) as payment for the backstop

The Barber Lake structure is materially worse for equity holders than Black Pearl. The 50% excess cash flow sweep, shorter lease term (10yr vs 15yr), and higher coupon (7.125% vs 6.125%) mean significantly less cash reaches corporate. Estimated annual distributable to CIFR: $30-50M vs $140M for Black Pearl.

Combined Contracted Cash Flow to CIFR Corporate

                        Black Pearl    Barber Lake    Combined
Annual NOI (stabilized)    ≈$367M        ≈$300M       ≈$667M
Annual debt service       -$227M        -$200M       -$427M
Annual excess CF sweep       $0          -$70M        -$70M
Cash to corporate          ≈$140M        ≈$30-50M     $170-190M

Management cites $669M average annualized NOI. That's the gross number. The net number — what actually reaches corporate after project-level debt and sweeps — is $170-190M. The ratio is 26%. Three-quarters of the NOI is consumed before it leaves the project entity.

III. The Pipeline

Beyond the contracted 600MW, CIFR claims ≈3.4 GW across 8 additional sites:

Additionally, the February 2026 investor presentation discloses a 500MW Barber Lake expansion MOU — suggesting Google/Fluidstack may want significantly more capacity at the existing site. Existing tenant expansion is a strong validation signal.

That said, zero of these pipeline sites have contracted HPC tenants. Management says "multiple parties interested in all sites" and "we're spoilt for choice." The language is hedged: "targeting," "we believe," "we hope to find a partner." Committed language is reserved exclusively for the two signed leases.

Tyler Page on Colchis (1GW, the largest site): "predicting the budgetary cost at Colchis is a little bit challenging" and they need to "find a partner before too long." This is an honest admission that the biggest pipeline asset requires a counterparty commitment before serious capital deployment.

The Colchis JV structure adds complexity. CIFR holds 53% via a VIE; the 47% NCI holder has redeemable interests "outside the control of the Company." The $30.3M redeemable NCI sits outside permanent equity. If Colchis development goes well, the NCI holder benefits. If it goes poorly, they can potentially force redemption — a put option against CIFR on their most important growth asset.

Nine of ten sites are in Texas (ERCOT). Geographic concentration is real. Texas SB 6 (enacted 2025) imposes new interconnection requirements for loads ≥75MW, including security payments and potential batching delays.

IV. Financial Profile

The Four-Year Arc

*Plus $2.0B Black Pearl notes Feb 2026 = $5.2B total.

CIFR has never generated positive free cash flow. FY2025 operating cash burn was $208M ($17M/month), accelerating as HPC construction overhead hits the P&L before HPC revenue starts.

The $822M net loss is mostly non-cash: $450M embedded derivative reclassification on the 2031 converts, $199M D&A, $96M write-down of assets held for sale, $45M miner impairment, $53M SBC. Adjusted earnings (company definition): $22M — down from $107M in FY2024 because interest expense rose from $2M to $37M. That interest expense is $150M+ annualized on the full $5.2B debt stack. It hits the P&L in 2026 before HPC revenue offsets it.

Capital Structure

The 2031 convertible converts at $16.03. Stock is at $16.48. These notes are at the money. If the stock goes to the consensus target of $27, conversion is near-certain, adding 161.9M shares (40% dilution). The $82.7M capped call mitigates dilution between $16.03-$23.32, but above $23.32, dilution resumes in full.

Fully Diluted Share Count

Outstanding:                          405.1M
Convertible notes (2030 + 2031):     +161.9M (combined, per 10-K anti-dilution table)
Google warrants @ $0.01:              +24.2M
ATM remaining ($725M):               +44.0M (at $16.48)
RSU/PSU pipeline:                    +15.8M (per 10-K)
                                     ─────────
Fully diluted:                       ≈651M

Fully diluted market cap:            ≈$10.7B
Fully diluted EV:                    ≈$15.3B

The 161.9M conversion shares reflect both the 2030 converts ($172.5M, conversion price $4.45 — deep ITM) and the 2031 converts ($1.3B, conversion price $16.03 — at the money), including anti-dilution adjustments since issuance. The 2030 notes will almost certainly convert; the 2031 notes convert if the stock stays above $16.03. The $82.7M capped call mitigates dilution on the 2031 notes between $16.03-$23.32.

Analysts using undiluted shares overstate per-share values by ≈60%.

The Debt Maturity Wall

CIFR must be generating substantial HPC cash flow by 2028 or face refinancing risk on the converts. If noteholders exercise the 2029 put on the $1.3B zero-coupon converts, CIFR needs $1.3B in cash or must issue equity.

V. Competitive Position

The Landscape

Every BTC miner is making the same pivot simultaneously:

CIFR has the best tenants (Amazon direct parent guarantee + Google backstop) but the worst execution timing (0MW HPC operating vs peers already generating HPC revenue). The market is paying $6.7B for zero operating HPC megawatts on the strength of the Amazon and Google names.

What's Defensible

Power interconnections. 4.2 GW of connected power in grid-constrained markets. ERCOT queue is 4-7 years. You can't buy your way past physics and permitting. Tyler Page: "Cipher is basically the only firm that has extremely high levels of credibility with grid wildcatters and also the ability to talk to hyperscalers." This is a bridging capability — sourcing sites from landowners/power developers AND having the technical credibility to win hyperscaler leases. The claim is self-serving but partially validated by landing both Amazon and Google.

Tenant credit quality. No peer has two top-3 hyperscaler relationships with parent-level guarantees. Amazon's parent guarantee covers ALL rent and operating expenses for 15 years. This is the gold standard in data center leasing.

Switching costs. Once a 15-year lease commences on custom-built infrastructure, the tenant is locked in. Power density specs, cooling design, and interconnection rights are tenant-specific.

What's Not Defensible

Construction capability. The 10-K explicitly states: "We have limited experience in developing an HPC hosting business and we have NOT PREVIOUSLY constructed and operated an HPC data center." Quanta Services builds the facilities. CIFR is a development-stage infrastructure company relying on a contractor for its most critical activity.

Speed to market. Every peer is 6-12 months ahead on HPC delivery. CORZ has 16MW operating. IREN has GPUs running. APLD has 100MW delivered. CIFR's first HPC megawatt doesn't come online until September 2026 at earliest.

The physical data center. Shell construction is becoming commoditized. Multiple firms can build the boxes. The scarce resources are power interconnections (CIFR has), tenant relationships (CIFR has), and speed (CIFR doesn't have).

VI. Management and Governance

CFO transition. Ed Farrell left Oct 2025 after two years, replaced by Greg Mumford (33, ex-KBW investment banker). No operating CFO experience. Now running finance for a $6.5B EV entity with $5.2B in debt, two major construction projects, and complex derivative accounting. High-stakes debut.

Insider ownership signal: zero open market purchases. No insider has bought a single share with their own money (Form 4 code P). Both the CEO and COO have active 10b5-1 selling programs. Tyler Page's plan authorizes selling up to 1.5 million shares (≈$25M) through December 2026. Patrick Kelly is selling ≈35K shares per month.

High SBC ($800K average per employee, 24% of revenue) means insiders already have substantial equity exposure, and selling to diversify is rational. But the absence of open market buying during what management calls "the most exciting opportunity in a generation" is information. When you're pivoting the company to a new future but hedging your personal exposure to it, that's a signal.

Capital allocation: sophisticated but dilutive. The good: non-recourse project financing is smart capital structure engineering; landing Amazon + Google validates the strategy; zero-coupon $1.3B convertible at 0% cost of capital is remarkable execution. The bad: raised $195.5M through ATM at $5.88 average (stock now $16.48 — shareholders diluted at 1/3 current price); WindHQ JV sold to Canaan for ≈$40M all-stock after $96M write-down; Black Pearl mined BTC for five months before demolishing the mining infrastructure for HPC conversion.

Proxy not yet filed. Board composition, detailed compensation, and governance details deferred to the 2026 proxy. Tyler Page is CEO and Director. Bitfury (original SPAC sponsor) terminated its board observer agreement in July 2025. Notably, the February 11, 2026 8-K added Thomas Duda (VP of Real Estate, Henry Crown and Company) to the board — bringing institutional real estate/infrastructure investment expertise aligned with the HPC pivot.

Material weakness (ITGC) was identified in FY2024 and management asserts remediation in FY2025. For a 66-person company managing $5.2B in debt, internal control quality matters.

VII. Factor Profile

What Drives Returns

Factor regression (250 trading days, with BTC factor):

CIFR idio = 66%

Momentum (MTUM)     β = +2.33    22% of variance
Bitcoin (IBIT)      β = +0.67    13% of variance
Tech (XLK)          β = +0.27     3% of variance
Market (SPY)        β = -0.51    -4% of variance
Idiosyncratic                    66% of variance

Idiosyncratic variance is 66% — below the 75% target. With the standard model (no BTC factor), idio is 72%. Both are below the threshold where factor noise materially degrades information ratio (≈19% IR degradation at 66% idio per the Paleologo framework).

CIFR is simultaneously three things:

1. A momentum stock (22% of variance, +295% 1Y)
2. A crypto proxy (13% of variance, IBIT β=0.67)
3. A company-specific story (66%)

The "company bet" narrative only holds if you strip out BTC and momentum. Without hedging those factors, ≈34% of variance is unintentional beta with no informational edge.

Peer Comparison

CIFR has the highest BTC sensitivity in the group (2x CORZ) despite both pursuing HPC pivots. This makes sense — CIFR's revenue was 100% BTC mining in FY2025 while CORZ already had HPC revenue. Until HPC rent starts, CIFR trades as a BTC proxy with an HPC call option bolted on.

All peers load heavily on momentum (β_MTUM > 2.0 across the board). This is the sector signature. Owning multiple names in this group is NOT diversification — it's concentration in the same factor cocktail.

Options Market Signal

ATM implied vol: 110%. Put/call OI ratio: 1.32 (bearish lean). Short interest: 20.4%. Heavy put protection at $12-14 strikes (11K OI). The options market prices CIFR for extreme volatility but leans bearish on direction — consistent with institutions owning for the HPC option but hedging the execution/BTC risk.

VIII. Forward Expectations Gap

What Current Price Requires

The $6.7B market cap decomposes into two pieces: contracted projects and pipeline optionality.

Contracted projects (Black Pearl + Barber Lake): NPV of distributable cash to corporate, discounted at 8-10% (Amazon-guaranteed cash flows warrant lower discount than generic project finance):

Black Pearl: ≈$140-181M/yr avg × 15yr PV annuity     = $1.07-1.38B
Barber Lake: ≈$30-50M/yr avg × 10yr PV annuity       = $185-340M
Gross NPV of contracted distributable cash:           = $1.3-1.7B

Less: Corporate overhead allocated to contracted
      (≈$115M/yr × 600MW/4,200MW pipeline share × PV) = $100-150M
Less: Remaining unfunded construction capex            = ≈$200M
                                                       ─────────
Net equity value of contracted projects:               = $0.9-1.4B

Note: The Google warrant ($525M liability) and corporate converts ($1.47B face) are not deducted here because both are captured in the fully diluted share count — deducting them from NPV AND adding their shares to the denominator would double-count. Equity value is expressed on a fully diluted per-share basis below.

On ≈651M fully diluted shares, $0.9-1.4B of contracted project equity = $1.40-$2.15 per diluted share.

Pipeline optionality: The residual value. Current price of $16.48 implies ≈$14-15 per share is attributed to the uncontracted ≈2.3 GW pipeline.

At $14-15/share × 651M diluted shares = $9-10B implied pipeline value, or roughly $4.0-4.3M per uncontracted megawatt. For reference, Black Pearl's contracted economics imply ≈$3.6M/MW. The market is pricing the uncontracted pipeline at or above the value of contracted, Amazon-guaranteed capacity.

This requires a high conversion rate of the pipeline to HPC, hyperscaler tenants secured for most sites, $15-20B in additional project financing, on-time construction, and manageable incremental dilution. All of these are possible — management says "multiple parties interested in all sites" and the demand environment is validated by counterparties — but none are contracted.

Street vs. Filing Reality

The most actionable disconnect is FY2026 revenue.

What's actually happening in 2026: Black Pearl BTC mining ceased February 2026. Odessa is the sole remaining mining site, operating at a loss (FY2025: $81.2M direct cost on 1,925 BTC mined = ≈$42K/BTC in power alone; all-in operating costs including $79M comp and $36M G&A push well above current ≈$67.5K BTC price). HPC rent from Black Pearl doesn't start until October 2026 at earliest, ramping through February 2027. Barber Lake Phase I delivery is targeted for September 2026.

Our estimate, quarter by quarter: Q1 ≈$30-40M (mining only, declining), Q2 ≈$25-35M (mining only, further reduced), Q3 ≈$40M (mining + first HPC rent), Q4 ≈$60M (HPC ramping). Full year: $120-175M at current BTC prices (≈$67.5K). This is a BTC-price-sensitive estimate — if BTC recovers to $100K+, mining revenue in H1 could be $40-50M/quarter, raising the floor to $170-225M. But even at $100K BTC, the consensus $406M requires HPC revenue ramp faster than the construction phasing in the 8-K supports. The low end of the street range ($248M) is closer to our estimate.

Q1 2026 earnings (May 5, est. -$0.06 EPS) will be the first test. Revenue should show a material decline from Q4 2025 ($60M reported). If the street hasn't adjusted models for the BTC mining cessation, there's a negative surprise.

The $669M NOI vs. $170M Distributable Cash

Management's Feb 2026 press release headlines ≈$669M average annualized NOI from contracted leases. This is the gross number at the project level. After $427M in annual debt service and $70M in excess cash flow sweeps (Barber Lake), approximately $170-190M reaches corporate. Analysts citing $669M NOI without the debt waterfall overstate the equity story by nearly 4x.

The Convertible Dilution

The 2031 convertible conversion price of $16.03 is three cents below current price. The stock has been oscillating around this conversion price for weeks. At the consensus target of $27, these notes are deep in the money and 161.9M shares get added to the denominator. Per-share metrics at $27 on 405M basic shares look very different than on 590M+ diluted shares. The consensus target implies a ≈$16B fully diluted EV, requiring ≈$1B in EBITDA to justify at 15x — roughly double what the contracted projects produce.

IX. Primary Source Findings — What Consensus Misses

From deep reading of the 10-K, recent 8-Ks, and counterparty filings:

1. Tenant guarantees only activate AFTER rent commencement. If construction delays cause lease termination BEFORE rent starts, the Amazon parent guarantee and Google backstop do not trigger. The protective structure only works if CIFR delivers on time. Miss the deadline, lose the protection AND the tenant.

2. Google's $525M warrant liability has a floor. Per the November 20, 2025 8-K amendment, if warrant shares are worth less than $435M at the exercise period start, CIFR must issue additional shares OR pay cash to make up the difference. At current prices ($16.48 × 24.2M = $399M), CIFR would owe Google ≈$36M in cash or additional shares. This is a guaranteed minimum payment to Google that grows if the stock falls.

3. ONE operating segment: "Bitcoin Mining." Despite the HPC pivot narrative, CIFR reports one segment. The CODM doesn't receive asset-level information for resource allocation. HPC segment won't appear until leases commence operations. The company's internal reporting structure hasn't caught up to the market narrative.

4. Quanta Services flags labor as "the tightest market" in data center construction. CIFR's construction partner explicitly identifies the constraint CIFR doesn't acknowledge. CIFR says "firmly on track." Quanta says labor for data center projects is the most constrained segment in their entire business. No independent engineer's report or construction milestone certifications have been filed.

5. Colchis (1GW) — the largest pipeline site — has a NCI holder who can force redemption. The 47% non-controlling interest holder has redemption rights outside CIFR's control. This is a put option against CIFR on their most important growth asset.

6. The Luminant power agreement (Odessa) is classified as a derivative. Marked to market every quarter using Level 3 (unobservable) inputs. Creates volatile non-cash P&L swings: -$29M in FY2025, -$8M in FY2024, +$27M in FY2023. The underlying contract expires July 2027 with "no renewal terms specified." Odessa's future after the PPA expires is unaddressed.

7. Google's Fluidstack intermediation is a strategic pattern, not a one-off. Google deployed $18.2B across CIFR, HUT, and WULF via the Fluidstack structure. This validates the demand but also reveals that Google uses Fluidstack specifically to avoid direct long-term real estate commitments. The Fluidstack layer gives Google optionality that a direct lease would not. This structure is untested through an economic downturn.

X. Key Risks

Construction Execution (FIRST risk factor in the 10-K)

The 10-K's own language: "We cannot guarantee we will complete construction of all or any portion of the Barber Lake Facility, the Black Pearl Facility, or any future strategic growth initiatives on time or within our cost estimates, if at all." And: "We have NOT PREVIOUSLY constructed and operated an HPC data center."

CIFR provides completion guarantees to the project-level debt but "there can be no assurance that we will have sufficient funds to meet our obligations under such guarantees." If a project goes sideways, CIFR is on the hook at the parent level despite the non-recourse structure.

Mitigant: At Black Pearl, Amazon covers construction cost overruns above the $9.5M/MW cap. This is significant — it transfers the most common execution risk (cost blowout) to the tenant on CIFR's largest project. This protection does NOT exist at Barber Lake.

Revenue Valley of Death (H1 2026)

BTC mining ceased at Black Pearl in February 2026. Odessa mines at a loss. HPC rent doesn't start until Q4 2026. The first half of 2026 will show declining revenue, continued cash burn ($17M/month), and no HPC revenue. Management calls this "a year of execution." The P&L will show deterioration.

Dilution Pipeline

Known dilution vectors: 161.9M shares from 2031 converts (at the money), 24.2M Google warrants, $725M ATM capacity (≈44M shares), plus ongoing SBC. Fully diluted share count: ≈651M vs 405M outstanding. If the stock rises to the consensus target, dilution mechanically increases. If it falls, the Google warrant floor creates additional dilution or cash obligation.

Convertible Put Dates

The 2030 converts ($172.5M) are puttable May 15, 2028. The 2031 converts ($1.3B) are puttable October 1, 2029. If holders exercise, CIFR needs $1.47B in cash or must issue equity at prevailing prices. This is manageable if HPC cash flows are ramping but creates a cliff risk if construction delays or revenue falls short.

AI Efficiency Risk

From the 10-K: "The introduction of, development and advancement in the efficiency of AI models could potentially adversely affect data center usage by significantly reducing the computational power needed such as those we are building at Barber Lake." Management has explicitly flagged the risk that more efficient AI reduces demand for the high-power-density infrastructure CIFR is building. The DeepSeek moment in January 2025 showed this isn't hypothetical.

Texas Concentration

Nine of ten sites in ERCOT. Weather events (Winter Storm Uri 2021), regulatory changes (SB 6), and grid stability risks are concentrated in a single jurisdiction.

The Herding Problem

Every BTC miner is making the same pivot: CORZ → CoreWeave, IREN → Microsoft, APLD → CoreWeave, HUT → Google, CIFR → Amazon/Google. If AI capex slows, all former miners compete for a shrinking pool of HPC tenants with purpose-built facilities they can't easily repurpose. The scarcity today is power. The scarcity tomorrow might be tenants.

XI. Counterparty Cross-Check

No direct contradictions. The deals are genuine. The Black Pearl $2.0B bond offering was 6.5x oversubscribed (≈$13B in book interest), the strongest institutional validation signal available — sophisticated credit investors underwriting the Amazon parent guarantee at scale. The Barber Lake $333M tack-on priced at a premium to par (100.25%), another positive demand signal.

The main discrepancy is one of emphasis: CIFR paints frictionless execution, counterparties describe a strained ecosystem where execution is difficult and customers have alternatives. But the bond market's overwhelming response to the project debt suggests institutional investors are comfortable with the credit risk of the contracted projects, even if they share concerns about the uncontracted pipeline.

XII. What to Watch

The single most important near-term data point is Q1 2026 revenue. If it prints at $30-40M (filing-supported) versus street expectations that may assume $80-100M quarterly pace, that's the first crack in the consensus.

The single most important medium-term data point is whether Black Pearl Data Hall 1 is rack-ready by September 30, 2026. On-time delivery validates everything. A delay threatens the lease, the guarantee activation, and the entire re-rating thesis.

Appendix: Sources

All analysis derived from primary sources:

• CIFR 10-K filed February 24, 2026 (SEC EDGAR)
• CIFR 8-K filings November 2025 — February 2026 (10 filings reviewed)
• CIFR Q3 2025 Earnings Transcript (November 3, 2025)
• AMZN Q4 2025 Earnings Transcript (February 5, 2026)
• PWR Q4 2025 Earnings Transcript (February 19, 2026)
• VST Q3 2025 Earnings Transcript (November 6, 2025)
• CIFR Form 4 filings (3 filings reviewed)
• Black Pearl illustrative financials (8-K Exhibit 99.1, February 3, 2026)
• Barber Lake indenture (8-K Item 1.01, November 13, 2025)
• yfinance market data (February 26, 2026)
• Factor regression: iev regress (250-day lookback, multiple specifications)