Formerly Cipher Mining. Bitcoin miner transforming into a 4.2 GW AI/HPC data center developer. Leopold Aschenbrenner holds $155M (new Q4'25 position). 66 employees, 10 sites across Texas + Ohio. | CIK: 0001819989 | AGI Score: 9/10 | Analysis date: 2026-03-12
Leopold Aschenbrenner initiated a new $155M position in Cipher Digital in Q4 2025 (10.47M shares of common stock). He also held $72M in CIFR convertible notes as of Q3 2025 -- so his total CIFR exposure is roughly $227M across equity + debt. This is Leopold's third-largest Bitcoin miner position behind CORZ ($419M, activist 13D) and IREN ($329M). Unlike CORZ where he's an activist, CIFR appears to be a passive position -- but the size is substantial. The question: Cipher has 4.2 GW of pipeline (3x CORZ's 1.4 GW) but is smaller and earlier-stage. Two hyperscaler leases signed (Google and Amazon, 600 MW). Is this a cheaper, earlier-stage version of the same CORZ thesis, and why did Leopold buy in Q4 2025 specifically?
2-year range: $2.10 (low, April 2024) to $24.71 (high, December 2024). Currently $13.71 -- down 45% from highs. Leopold bought Q4 2025 when price was $12-25 range.
Cipher Digital's investment thesis is simple: it owns or controls 4.2 GW of power capacity across 10 sites, and AI/HPC customers are desperate to lease energized data center capacity. The company has already signed two hyperscaler leases (Google and Amazon) for 600 MW and has 3.4 GW of additional pipeline sites targeting energization between 2026-2030.
This is the core asset. Power interconnection queues take 4-5+ years. Cipher has already secured interconnection for most of its pipeline. The question is whether they can convert these approvals into operating data centers with paying tenants.
| Site | Location | Capacity | Status | Tenant | Energization |
|---|---|---|---|---|---|
| Odessa | Odessa, TX (52 acres) | 207 MW | Operating | BTC mining (could retrofit to HPC) | Operating since Nov 2022 |
| Barber Lake | Colorado City, TX (250 acres) | 300 MW | Under Construction | Google/Fluidstack (Google backstop) | Phase I: Sep 2026, Phase II: Jan 2027 |
| Black Pearl | Wink, TX (75 acres) | 300 MW | Under Construction | Amazon Web Services (15-year lease) | Phase I: Q4 2026, Full ramp: Q1 2027 |
| Colchis | West Texas (620 acres) | 1,000 MW | Pipeline | TBD | 2028 |
| Ulysses | Ohio (195 acres) | 200 MW | Pipeline | TBD (first non-Texas site, PJM access) | Q4 2027 |
| McLennan | Central Texas | 500 MW | Pipeline | TBD | 2028 |
| Mikeska | West Texas | 500 MW | Pipeline | TBD | 2028 |
| Milsing | East Texas | 500 MW | Pipeline | TBD | 2028-2029 |
| Stingray | West Texas | 100 MW | Pipeline | TBD | H1 2026 |
| Reveille | Cotulla, TX | 70 MW | Pipeline | TBD | 2027 |
| TOTAL | 4,177 MW | = 4.2 GW | |||
| Operating (BTC mining) | 207 MW | 5% | Generating revenue now |
| Under construction (HPC contracted) | 600 MW | 14% | Google + Amazon, rent starting Q4 2026 |
| Pipeline (not contracted) | 2,870 MW | 69% | Interconnection secured, tenants TBD |
| Odessa (potential HPC retrofit) | 207 MW | 5% | Could convert from BTC, PPA through Jul 2027 |
| Small pipeline sites | 170 MW | 4% | Stingray + Reveille |
The honest picture: Only 14% of capacity is contracted with paying HPC tenants. 69% is "pipeline" with no signed leases. The thesis requires Cipher to sign 2,870+ MW of additional HPC leases over the next 3-4 years. Bullish argument: Google and Amazon as reference customers make signing additional leases much easier. Bearish: pipeline is aspirational until contracts are signed.
The central question for all Bitcoin miner-to-AI pivots: what's the right $/MW valuation?
| Company | EV | Operating MW | Total Pipeline | EV/Op. MW | EV/Total MW | HPC Status |
|---|---|---|---|---|---|---|
| CIFR (Cipher Digital) | $7.89B | 207 MW | 4,200 MW | $38.1M | $1.88M | 600 MW contracted (Google + Amazon) |
| CORZ (Core Scientific) | $6.07B | ~1,400 MW | 1,400 MW | $4.3M | $4.3M | 590 MW contracted (CoreWeave) |
| IREN (Iris Energy) | $14.5B | 810 MW | 2,810 MW | $17.9M | $5.2M | GPU cloud, own NVIDIA clusters |
EV/Operating MW is misleading for CIFR because only 207 MW is operating (all BTC mining). The right way to value Cipher is on its contracted + pipeline MW:
If you believe the full 4.2 GW pipeline will be developed and leased, CIFR is extremely cheap. If you believe only the 600 MW contracted will materialize, it's fairly valued. The truth is probably somewhere in between.
| Company | Shares | Value (Q4'25) | % of Leopold AUM | Position Type | First Bought |
|---|---|---|---|---|---|
| CORZ (Core Scientific) | 28.76M | $419M | ~8% | Activist 13D | Q1 2025 |
| IREN (Iris Energy) | 8.70M | $329M | ~6% | Passive | Q1 2025 |
| CIFR (Cipher Digital) | 10.47M | $155M | ~3% | Passive | Q4 2025 (NEW) |
| BITF (Bitfarms) | -- | exited | -- | Sold | Q4 2024 |
| CLSK (CleanSpark) | -- | exited | -- | Sold | Q4 2024 |
| BTDR (Bitdeer) | -- | held | -- | Passive | Q2 2025 |
| RIOT (Riot Platforms) | -- | held | -- | Passive | Q4 2025 |
| HUT (Hut 8) | -- | held | -- | Passive | Q4 2025 |
Leopold's total Bitcoin miner exposure is roughly $1B+, or ~20% of his $5.2B AUM. This is not a token position. He is making a massive, concentrated bet on Bitcoin miners as AI infrastructure plays. CIFR is his newest addition.
Several catalysts aligned in Q4 2025:
Bottom line: Leopold already had the Bitcoin-miner-as-AI-infrastructure thesis from CORZ and IREN. CIFR was a way to add more MW exposure at a cheaper per-MW price, diversified across different hyperscaler tenants.
Cipher's balance sheet transformed in FY2025. Total assets went from $855M to $4.29B -- a 5x increase in one year. This was funded primarily by $3.15B of new long-term debt.
| Item | FY2025 | FY2024 | FY2023 | Change (YoY) |
|---|---|---|---|---|
| Total Assets | $4,292M | $855M | $566M | +$3,437M (+402%) |
| Cash & Equivalents | $628M | $6M | $86M | +$622M |
| Restricted Cash | $1,761M | $0 | -- | New (project financing) |
| Bitcoin Inventory | $125M | $93M | $33M | +$32M |
| PP&E (Net) | $645M | $532M | $282M | +$113M |
| Other Non-Current Assets | $837M | $18M | -- | Deposits, prepaid construction |
| Liabilities | ||||
| Total Debt | $2,767M | $56M | $22M | +$2,711M |
| Long-Term Debt | $2,712M | $0 | $0 | NEW |
| Current Liabilities | $699M | $132M | $34M | +$567M |
| Total Liabilities | $3,456M | $173M | $75M | +$3,283M |
| Equity | ||||
| Stockholders' Equity | $806M | $682M | $491M | +$124M |
| Retained Earnings | -$1,004M | -$181M | -$137M | -$823M |
| Shares Outstanding | 405M | 351M | 291M | +54M (+15%) |
| Book Value / Share | $1.99 | $1.94 | $1.69 | +3% |
Cipher went from virtually no debt to $2.77B of debt in one year. This is the financing for the HPC data center buildout. Key details:
The debt is presumably project-level, non-recourse (the 10-K emphasizes this approach), which limits downside to the project assets. But $37M+ of annual interest expense is significant for a company generating $224M in BTC mining revenue.
Despite the massive debt raise, Cipher maintains positive equity of $806M. This is meaningfully better than CORZ (-$963M). Key sources:
Tangible book value is $728M, or $1.80/share. At $13.71/share, the stock trades at 7.6x tangible book.
| Item | FY2025 | FY2024 | FY2023 |
|---|---|---|---|
| Revenue (BTC mining) | $224M | $151M | $127M |
| Cost of Revenue | ($81M) | ($62M) | ($50M) |
| Gross Profit | $64M | $28M | $19M |
| Gross Margin | 28.4% | 18.5% | 15.1% |
| Compensation & Benefits | ($79M) | ($61M) | ($57M) |
| G&A | ($36M) | ($33M) | ($28M) |
| Depreciation | ($199M) | ($102M) | ($59M) |
| Operating Loss | ($422M) | ($44M) | ($20M) |
| Other Expense (net) | ($404M) | ($1M) | ($2M) |
| Net Loss | ($822M) | ($45M) | ($26M) |
| EPS (Basic) | ($2.15) | ($0.14) | ($0.10) |
The FY2025 loss looks catastrophic but needs context. Major non-cash and one-time items:
| Depreciation & Amortization | $199M | Non-cash (miners depreciating over 3yr life) |
| Other operating losses | $174M | Impairments (mining equipment being retired for HPC) |
| Other expense | $406M | Likely fair value adjustments, derivatives |
| Unrealized BTC losses | $42M | Bitcoin price decline mark-to-market |
| Change in PPA fair value | $29M | Power purchase agreement revaluation |
| SBC | $53M | Stock-based compensation (non-cash) |
The "real" cash operating loss is much smaller. Operating cash flow was -$208M, and capex was $494M. Total cash burn of $702M, funded by $3.15B in debt issuance and $245M in equity raises. The company has $628M of unrestricted cash left.
| Metric | CIFR | CORZ | IREN |
|---|---|---|---|
| Market Cap | $5.55B | $5.12B | $13.72B |
| Enterprise Value | $7.89B | $6.07B | $14.5B |
| Total Pipeline (GW) | 4.2 GW | 1.4 GW | 2.8 GW |
| Operating Capacity | 207 MW | ~1,400 MW | 810 MW |
| HPC MW Contracted | 600 MW | 590 MW | Own GPUs |
| HPC Tenants | Google + Amazon | CoreWeave only | Self (GPU cloud) |
| HPC Revenue (FY2025) | $0 | $65M | GPU cloud revenue |
| Total Revenue (FY2025) | $224M | $319M | $757M |
| Net Income | -$822M | -$289M | $390M |
| Total Debt | $2,767M | $1,163M | $3,843M |
| Stockholders' Equity | $806M | -$963M | $2,510M |
| Employees | 66 | 325 | 257 |
| EV / Total Pipeline MW | $1.88M | $4.34M | $5.16M |
| Leopold Position | $155M (passive) | $419M (13D activist) | $329M (passive) |
| Leopold AGI Score | 9 | -- | 8 |
The correct way to frame 10x: start with the bull case terminal value, divide by 10, and compare to today's price.
| Scenario | MW Leased | Assumed $/MW | Bull Case EV | 10x Entry EV | 10x Entry Price* |
|---|---|---|---|---|---|
| Bear (only contracted MW) | 600 MW | $10M | $6B | $600M | ~$1.48 |
| Base (add Colchis + Ulysses) | 1,800 MW | $10M | $18B | $1.8B | ~$4.45 |
| Bull (75% of pipeline) | 3,150 MW | $12M | $37.8B | $3.78B | ~$9.33 |
| Full pipeline at premium | 4,200 MW | $15M | $63B | $6.3B | ~$15.56 |
*10x entry price = Bull Case EV / 10 / 405M shares. Current price: $13.71, current market cap: $5.55B.
Bull case (75% pipeline at $12M/MW): $37.8B. 10x entry = $3.78B market cap (~$9.33/share). Current price of $13.71 is 1.47x above the bull case 10x entry zone. Under the full pipeline scenario ($63B), 10x entry = $6.3B (~$15.56/share) -- current price is below this entry, meaning a full-pipeline bull case gives 10x+ from here. The question is whether you believe CIFR can develop and lease 4.2 GW at premium rates.
The bull case gets close to 10x if:
Revenue potential at full build-out: If 4.2 GW generates $150K-200K/MW/year in lease revenue (industry range), that's $630M-$840M/year in HPC revenue with 40-50% margins. At 15-20x EV/EBITDA, that's $3.8B-$8.4B EBITDA supporting $57B-$168B EV. Even the conservative end implies significant upside.
Confidence: LOW
Cipher is difficult to floor-price because:
| Method | Floor Estimate | Rationale |
|---|---|---|
| Tangible Book Value | $1.80/share | $728M tangible equity / 405M shares |
| Net Asset Liquidation (conservative) | $1.50-2.50 | Land + BTC inventory + cash - debt (heavy discount for project-specific debt) |
| BTC Mining Cash Flow | $1.00-1.50 | 207 MW mining at current BTC price, 28% gross margin, 3-year DCF |
| HPC Lease NPV (contracted only) | $3.00-5.00 | 600 MW at $150K/MW/yr, 15-year terms, discounted at 12% |
Rough floor: $2-3/share -- about 80% below current price. This is the price where the contracted HPC leases (Google + Amazon) plus net tangible assets minus debt provides a floor. Below this, you're being paid to own the pipeline optionality for free.
Why the floor is unreliable: The $2.77B of debt is the wild card. If it's truly non-recourse at the project level, the equity isn't at risk. But the 10-K says Cipher "typically" pursues non-recourse structures -- not that all of it is. If any recourse debt exists and projects fail, the floor drops.
Leopold's timing makes sense when you stack up what happened in Q3-Q4 2025:
July 2025: Google/Fluidstack lease signed for Barber Lake (300 MW). First hyperscaler validation.
October 2025: Amazon 15-year lease signed for Black Pearl (300 MW). Second hyperscaler. This is the signal -- TWO hyperscalers signing is not an accident.
November 2025: Colchis 1-GW site acquisition. This is a massive expansion -- pipeline jumps from ~2.2 GW to ~3.2 GW. Dual AEP interconnection is high-quality power.
December 2025: Ulysses Ohio acquisition. Geographic diversification into PJM, de-risks the Texas story.
Q4 2025: Stock was $12-18, down from $24.71 peak. Market was focused on the BTC mining losses, not the HPC pivot catalysts.
Leopold's pattern: He buys into companies where the market is pricing the old business (BTC mining, declining) while the new business (HPC data centers, explosive growth) is being built but hasn't yet shown up in the financials. He did this with CORZ (bought before HDC revenue ramped) and IREN (bought before GPU cloud scaled). CIFR in Q4 2025 was the same play: buy the infrastructure before the revenue shows up.
The convertible notes position (Q3 2025, $72M) suggests Leopold initially took a safer position via debt, then added common equity ($155M) after the Amazon lease was signed. This is a confidence escalation pattern.
Cipher scores 9/10 because it addresses the single biggest constraint on AGI development: physical infrastructure for compute. AGI doesn't happen without massive GPU clusters, and GPU clusters don't run without energized data centers. Cipher's 4.2 GW pipeline makes it one of the largest providers of this bottleneck resource.
| Demand Boost | 10/10 | AGI requires unprecedented compute infrastructure. Cipher builds exactly this. |
| Margin Expansion | 4/10 | Data center margins are good but not exceptional. Cost discipline matters. |
| Strategic Assets | 9/10 | 4.2 GW of power interconnections take 4-5 years to replicate. Scarce. |
| Disruption Risk | 1/10 | Physical infrastructure needed regardless of AI algorithm changes. |
| Innovation Risk | 2/10 | Data centers don't become obsolete. Power and cooling always needed. |
AGI bull case: If AGI arrives by 2030, compute demand will be orders of magnitude beyond current capacity. Every MW of data center capacity will be needed. Cipher's 4.2 GW pipeline could be worth $60B+ in this scenario. The power interconnections are the moat -- you can't replicate them quickly.
| Floor (tangible book basis) | ~$2-3 | Where you can't lose if HPC leases hold |
| Current Price | $13.71 | Market is pricing ~$1.9M/MW of pipeline |
| Fair value (base case, 1,800 MW leased) | $25-35 | 2-3x from here |
| Bull case (3,000+ MW, AI infra valuation) | $60-90 | 5-7x from here |
| Full execution (4.2 GW at $15M/MW) | $120+ | ~10x |
CIFR is a high-risk, high-reward bet on AI infrastructure demand. It's cheaper than CORZ and IREN on EV/pipeline MW, has better tenant diversification (Google + Amazon vs CoreWeave), and Leopold's $227M position provides smart-money validation. But it's pre-revenue on HPC, carrying $2.77B in debt, and 86% of its pipeline is uncontracted. This is not a "little chance of losing money" situation -- it's a "massive upside if right, significant downside if wrong" setup.
Compared to CORZ: CIFR is earlier, cheaper per MW, and less levered (positive equity vs negative). But CORZ has $65M in HPC revenue already and an activist investor pushing for shareholder value. CIFR is the "if you like CORZ, here's the next one" trade.
Action: Watch for Q4 2026 rent commencement at Barber Lake (Google) and Black Pearl (Amazon). That's the proof point. If they deliver on time and within budget, the stock likely re-rates significantly. If construction delays materialize, the debt load becomes the dominant concern. Not a floor-price buy at $13.71. A floor-price buy would be closer to $2-4, which would require either a BTC crash or market-wide panic. Add to watchlist; revisit if price drops meaningfully or if new HPC leases are signed for the 3.4 GW pipeline.
| SEC EDGAR CIK | 0001819989 |
| 10-K Filing (FY2025) | Filed 2026-02-24 (accession: 0001819989-26-000009) |
| 10-K Filing (FY2024) | Filed 2025-02-25 (accession: 0001819989-25-000005) |
| XBRL Financials | SEC EDGAR Company Facts API |
| Market Data | yfinance (as of 2026-03-12) |
| Leopold Holdings | Situational Awareness LP 13F filings (Q3-Q4 2025) |
| AGI Score | AWB AGI Impact Scoring (Claude Sonnet 4.5, scored from 10-K) |
| Sector Knowledge | bitcoin-miners-ai-pivot.md |