TeraWulf
Power-first real-asset landlord: monetize energized, interconnect-queued, zero-carbon power by building HPC colocation shells and leasing critical IT capacity to a neocloud tenant (Fluidstack) on long-dated (10-yr) contracts. The economic engine is the spread between contracted, credit-backstopped lease revenue (~$1.5-2M/MW/yr) and the debt-funded capex to build the shells — a shift away from volatile, hash-price-linked BTC hosting toward higher-multiple, contracted landlord cash flows. The scarce asset is the site's power, not ASICs.
The thesis on this name
State of the AI Cloud
The cleanest full pivot in the cohort — HPC revenue (~$21M Q1'26) already overtook mining, Fluidstack counterparty backstopped by Google (a ~14% equityholder), on zero-carbon Lake Mariner power.
State of the AI Cloud
Cleanest pure-play AI-infra pivot — HPC revenue already overtook mining, ~360MW of Google-backstopped Fluidstack leases at Lake Mariner, and now a 20-year, ~401MW, ~$19B lease signed directly with Anthropic at Hawesville.
State of the AI Cloud
HPC revenue already > mining; ~360MW of Google-backstopped Fluidstack leases plus a directly-held ~401MW/~$19B Anthropic lease at Justified, on zero-carbon power — sized for the ~$428M Q1'26 loss and a 2H-2027 delivery.
Earnings, margins, COGS & capex
TeraWulf is mid-pivot from Bitcoin mining to AI/HPC colocation landlord. FY2025 revenue was $168.5M (+20.3%), still mining-dominated (digital-asset $151.6M vs HPC lease $16.9M). The inflection came in Q1 FY26: total revenue was ~flat at $34.0M but HPC lease revenue surged to ~$21M (~60% of the mix, +117% QoQ), overtaking BTC mining (<$13M) for the first time — the crossover in management's declared full exit from mining. The quarter also carried a ~$427.6M GAAP net loss (~$1.01/sh) including a ~$25.7M impairment, reflecting heavy capex and balance-sheet strain. The model is now a leveraged real-asset build: ~$5.8B debt (Mar 2026) funds Lake Mariner + new sites (Hawesville KY, Maryland) that are leased to Fluidstack under Google-backstopped 10-yr contracts. The bull case is contracted, credit-backstopped landlord cash flow at a re-rated multiple; the bear case is that the GAAP losses, the debt, and Fluidstack neocloud counterparty risk arrive before the contracted revenue scales.
Revenue trend
Margins
negative — capex + impairment + financing costs swamp early-ramp revenue
negative — construction + depreciation + SBC outrun revenue while HPC leases ramp
rising sharply — the crossover past mining, expected to keep climbing as Lake Mariner energizes
negative — structural while building multi-GW capacity on debt
COGS structure
Legacy mining COGS was dominated by electricity/power cost (the miner's edge was cheap, zero-carbon NYISO Zone-A power at Lake Mariner). For the HPC colocation model, direct operating cost is power + facility operations + staffing to run the leased shells; the far larger economic cost sits below the line as depreciation on the debt-funded build-out and interest on ~$5.8B of debt. During the transition a positive lease-level margin is masked at the GAAP level by depreciation, impairment, and financing costs — hence the large net loss despite growing contracted revenue.
Capex
Capex is the defining feature of the pivot. TeraWulf is spending multiples of its revenue to build HPC data-center capacity at Lake Mariner (NY) and newly acquired energy-advantaged sites — a $301.9M Hawesville, KY acquisition (former Century Aluminum smelter, ~480MW existing power) plus a Charles County, MD site, together adding ~1.5GW and more than doubling owned capacity to ~2.8GW; the Hawesville campus alone is an estimated ~$3-4B build (Fluor engaged for preconstruction), targeted for H2 2027. Funded by ~$5.8B of debt (convertibles + 7.75% senior secured notes due 2030) plus cash/restricted cash of ~$3.09B.
Latest earnings
EPS MISS — loss wider than expected; revenue ~flat YoY. The market read the print as a wide loss offset by the strategic HPC crossover; the stock was volatile [fact/estimate]
Management reiterated intent to fully exit Bitcoin mining and scale contracted HPC capacity; Hawesville campus (~$3-4B) targeted for H2 2027. No clean revenue guide during the transition [fact/guidance]
- HPC lease revenue (Q1 FY26)
- ~$21M (~60% of total; +117% QoQ) — first quarter above mining
- Contracted critical IT (Fluidstack, Lake Mariner)
- ~360MW after CB-5 + JV expansions; JV reported at ~$9.5B committed revenue [company-claimed]
- Total debt / liquidity
- ~$5.8B total debt (Mar 2026); cash + restricted cash ~$3.09B
- Owned power capacity
- ~2.8GW after Hawesville KY + Charles County MD (+~1.5GW)
Growth drivers
- Google-backstopped Fluidstack 10-yr HPC leases at Lake Mariner — two initial leases (Aug 2025) for >200MW / ~250MW gross, ~$3.7B contracted over initial terms (up to ~$8.7B with two 5-yr extensions) — company-claimed contract values [company-claimed]
- Expansions — CB-5 (160MW) + a 168MW JV — lifting total contracted critical IT for Fluidstack at Lake Mariner to ~360MW; a Google-backed JV reported at ~$9.5B of committed revenue [company-claimed]
- Google equity participation — warrants for a stake rising to ~14% pro forma, plus a lease backstop scaled $1.8B → ~$3.2B — a counterparty-validation signal [fact]
- HPC lease revenue crossover — ~$21M in Q1 FY26 (~60% of revenue), now exceeding BTC mining — the declared full exit from mining [fact]
- New energy-advantaged sites (Hawesville KY ~480MW, Charles County MD) adding ~1.5GW → ~2.8GW owned capacity, extending the leasable-power runway [fact]
- Zero-carbon (Niagara hydro + regional nuclear — ~89% carbon-free NYISO Zone-A) power positioning as a differentiator amid grid/regulatory scrutiny [fact]
Reported financials — SEC EDGAR
Audited GAAP figures pulled from SEC filings · latest filing 2026-02-27. The audited primary-source spine — not financial advice.
Revenue — annual (GAAP)
Margins & balance sheet — FY’25
Bull & bear
TeraWulf is the cleanest pure-play AI-infrastructure pivot among the crypto-miners: HPC lease revenue has already overtaken mining, the counterparty is a Google-backstopped Fluidstack with Google taking a ~14% equity stake, and the underlying asset — energized, zero-carbon, interconnect-queued power (~2.8GW owned) — is exactly the scarce resource the AI build-out is bidding for.
- HPC lease revenue (~$21M, ~60% of Q1 FY26 mix) already exceeds BTC mining — the pivot is real, not a plan [fact]
- Google backstop scaled $1.8B → ~$3.2B plus Google warrants toward a ~14% equity stake is a powerful third-party validation of the Fluidstack counterparty [fact]
- Zero-carbon nuclear/hydro power at Lake Mariner (~89% carbon-free) differentiates on the exact axis — grid/regulatory scrutiny — where the AI build-out is most constrained [fact]
- ~360MW contracted critical IT for Fluidstack at Lake Mariner, with a Google-backed JV reported at ~$9.5B of committed revenue, gives multi-year contracted visibility [company-claimed]
- ~2.8GW of owned power after Hawesville (KY) + Charles County (MD) extends the leasable runway well beyond the current book [fact]
WULF is a leveraged, GAAP-loss-making bet that a debt-funded HPC build converts into contracted landlord cash flow before the balance-sheet strain and neocloud counterparty risk bite. The Q1 FY26 ~$427.6M net loss on $34.0M of revenue and ~$5.8B of debt frame the risk: the contracted revenue is still ramping, the tenant is a neocloud only partially backstopped by Google, and management has removed the mining cash-flow hedge.
- ~$427.6M Q1 FY26 net loss (incl. ~$25.7M impairment) on $34.0M revenue — a wide GAAP loss during the most capital-intensive phase [fact]
- ~$5.8B total debt (Mar 2026) — convertibles + $3.2B 7.75% senior secured notes due 2030 — funds a build whose contracted revenue hasn't yet scaled; financing risk if AI-capex sentiment turns [fact/estimate]
- Counterparty risk sits with Fluidstack, a neocloud — Google's backstop covers lease obligations/project debt, not a direct Google tenancy; a Fluidstack stumble impairs the backlog [fact/estimate]
- Full mining exit removes the cash-flow hedge and residual BTC optionality, concentrating the equity entirely on HPC-build execution [estimate]
- Rapid multi-site, multi-JV expansion (Lake Mariner, Hawesville H2 2027, Maryland) multiplies grid-interconnection and construction-execution risk [estimate]
- Headline contract values (~$3.7B up to ~$8.7B; ~$9.5B JV) are company-claimed, extension-inclusive running totals — realized revenue depends on options being exercised and capacity being taken [company-claimed]
What it is worth
Real-asset landlord / contracted-backlog framing, not a near-term earnings multiple. WULF is GAAP loss-making (~$427.6M Q1 FY26) with ~$5.8B debt, so the equity (~$11.7B cap) prices the option value of converting ~2.8GW of energized, zero-carbon power + ~360MW contracted critical IT (Fluidstack/Google, ~$9.5B reported JV) into re-rated landlord cash flow. The price implies the pivot executes on schedule AND the ~$5.8B debt refinances smoothly AND Fluidstack (Google-backstopped) takes and pays — a stack of contestable assumptions.
A neocloud-demand air-pocket, a refinancing scare on the ~$5.8B debt, or interconnect/construction slippage compresses the multiple and forces the leverage + loss into focus [estimate]
Range-bound on each print: the HPC crossover is real and contracted, but large GAAP losses, ~$5.8B debt, and Fluidstack concentration cap the multiple until execution + refinancing are proven [estimate]
Re-rating toward a data-center-landlord multiple as HPC lease revenue scales past mining, the ~360MW (and beyond, on ~2.8GW owned power) converts to contracted cash flow, and Google's deepening backstop/equity de-risks the counterparty [estimate]
Headline contract values (~$3.7B up to ~$8.7B; ~$9.5B JV) are company-claimed, extension-inclusive running totals — not realized revenue; discount accordingly. The bull/bear spread hinges on execution + financing, not on a stable earnings stream that does not yet exist.
SWOT
Strengths
- Cleanest pure-play AI-infrastructure pivot of the miner cohort — HPC lease revenue (~$21M) already overtook BTC mining (<$13M) in Q1 FY26 [fact]
- Google-backstopped Fluidstack counterparty (backstop scaled $1.8B → ~$3.2B) plus Google as an equityholder rising to ~14% pro forma — a strong counterparty-validation signal [fact]
- Zero-carbon power at Lake Mariner (Niagara hydro + regional nuclear; ~89% carbon-free NYISO Zone-A) is a genuine differentiator amid grid/regulatory scrutiny of data-center power [fact]
- Energized, interconnect-queued, owned power (~2.8GW after Hawesville + Maryland) is the scarce asset in the AI build-out — hard to replicate quickly [fact]
Weaknesses
- Large GAAP losses now — ~$427.6M Q1 FY26 net loss (~$1.01/sh) incl. impairment, on just $34.0M revenue — balance-sheet strain is the top near-term risk [fact]
- Heavy leverage — ~$5.8B total debt (Mar 2026) — convertibles + $3.2B 7.75% senior secured notes due 2030 — funding a build whose contracted revenue is still ramping [fact]
- Counterparty concentration on Fluidstack, a neocloud — only partially mitigated by Google's backstop (which covers lease obligations/project debt, not a direct Google tenancy) [fact/estimate]
- Abandoning mining removes the cash-flow hedge and residual BTC upside, leaving the equity fully exposed to execution on the HPC build [estimate]
- Multi-site, multi-JV, compressed build timelines (Hawesville H2 2027) add grid-interconnection and construction-execution risk [estimate]
Opportunities
- Re-rating to a data-center-landlord / contracted-infrastructure multiple as HPC lease revenue scales past mining [estimate]
- Additional Fluidstack/Google or new-tenant leases on the ~2.8GW owned-power runway (Hawesville KY, Charles County MD) [fact/estimate]
- Zero-carbon positioning could win power-constrained, ESG-sensitive AI demand and ease regulatory/interconnect approvals [estimate]
- Deepening the Google relationship (equity + backstop) toward a stickier, more investment-grade-adjacent counterparty profile [estimate]
Threats
- AI-capex air-pocket / neocloud-demand unwind — if Fluidstack's demand or credit softens, contracted backlog quality and renewals are impaired while debt stays fixed [estimate]
- Financing risk — rising rates or a risk-off market makes rolling / refinancing ~$5.8B of debt expensive as capex continues to outrun revenue [estimate]
- Execution / interconnect risk on multi-GW builds (Lake Mariner, Hawesville, Maryland) delivering late or over budget [estimate]
- Grid / regulatory scrutiny of data-center power demand — a two-edged sword the zero-carbon story only partly offsets [estimate]
- Peer competition for the same power-monetization trade (IREN, Core Scientific, Applied Digital, Cipher) compresses the re-rating premium [fact/estimate]
Moats, dependencies & bottlenecks
Moats
zero-carbon, interconnect-queued owned power (~2.8GW; Lake Mariner ~89% carbon-free NYISO Zone-A) grid interconnect + carbon-free power is genuinely scarce and slow to replicate, but not exclusive to WULF The real moat is the site's power, not the buildings or ASICs; the zero-carbon angle differentiates amid regulatory scrutiny [fact]
Google-backstopped Fluidstack contract structure + Google ~14% equity stake a strong validation and switching cost, but the backstop covers obligations, not a direct Google tenancy; contractual not technological Counterparty validation is granted by Google, not owned — the tenant is still a neocloud [fact/estimate]
Long-dated (10-yr) contracted HPC lease backlog (~360MW critical IT at Lake Mariner) contractual switching cost; realized value depends on options exercised and counterparty ability to pay Backlog quality is only as good as Fluidstack's (Google-backstopped) willingness/ability to take and pay [company-claimed]
First-mover among miners on the decisive full pivot (HPC revenue already > mining) an execution lead, capital-replicable by well-funded peers (IREN, CORZ, APLD, CIFR) Being furthest along is an advantage, not a structural barrier [estimate]
Dependencies
The single anchor tenant for the HPC pivot; its own demand and credit ARE the thesis, only partially covered by Google's lease backstop [fact/estimate]
Backstop scaled $1.8B → ~$3.2B and warrants toward ~14% equity underpin the counterparty's credibility; a change in Google's posture would materially re-rate the thesis [fact]
~$5.8B debt (convertibles + $3.2B 7.75% senior secured notes due 2030) funds a build whose revenue is still ramping; a risk-off market makes refinancing/continued issuance costly [fact/estimate]
new KY/MD sites) Energized, carbon-free, interconnect-queued power is the scarce asset and the gating constraint on activating contracted capacity [fact]
Multi-GW builds on compressed timelines (Hawesville H2 2027) depend on EPC delivery on time and budget [fact/estimate]
Advantages
- Furthest-along full pivot in the miner cohort — HPC lease revenue (~$21M) already exceeds BTC mining (<$13M) [fact]
- Google-backstopped Fluidstack counterparty ($1.8B → ~$3.2B backstop) plus Google as a ~14% equityholder — best-in-cohort counterparty validation [fact]
- Zero-carbon nuclear/hydro power at Lake Mariner (~89% carbon-free) differentiates amid grid/regulatory scrutiny [fact]
- ~2.8GW of owned, energized power after Hawesville KY + Charles County MD — a long leasable-power runway [fact]
- ~360MW contracted critical IT for Fluidstack at Lake Mariner gives multi-year contracted revenue visibility [company-claimed]
Weaknesses
- Large GAAP losses now: ~$427.6M Q1 FY26 net loss (~$1.01/sh) on $34.0M revenue [fact]
- Heavy leverage: ~$5.8B total debt (Mar 2026) funding a build whose contracted revenue is still ramping [fact]
- Anchor-tenant concentration on Fluidstack (neocloud), only partially backstopped by Google — not a direct Google lease [fact/estimate]
- Full mining exit removes the cash-flow hedge and residual BTC optionality [estimate]
- Multi-site, multi-JV, compressed-timeline builds add grid-interconnect and construction-execution risk [estimate]
- Headline contract values are company-claimed, extension-inclusive running totals, not realized revenue [company-claimed]
Bottlenecks
- Power + grid interconnect — activating contracted critical IT is gated by energized, interconnect-queued, carbon-free MW at Lake Mariner and new sites [fact]
- Capital access — the build cannot outrun the ability to raise/refinance debt (~$5.8B already) plus cash on hand (~$3.09B); rate/risk-off shocks throttle growth [fact/estimate]
- Construction / EPC throughput — standing up multi-GW HPC shells on compressed timelines (Hawesville H2 2027) [fact/estimate]
- Counterparty ramp — converting ~360MW of contracted critical IT into cash depends on Fluidstack taking capacity on schedule [company-claimed]
- Skilled operations talent to run large HPC colocation campuses at high utilization [estimate]
Top signals & trends
Top signals
Q1 FY26 crossover (~$21M, ~60% of revenue) is the key validating data point; watch each print for continued ramp [fact]
~$427.6M Q1 loss on $34.0M revenue; the bull case needs the loss to narrow as contracted revenue scales [fact]
Rising debt to fund the build is the top financing risk; watch issuance cost and any covenant/maturity pressure [fact/estimate]
Deeper Google involvement is bullish validation; any softening of Fluidstack demand or Google's posture is the bear trigger [fact]
On-time energization converts backlog to cash; slippage compounds financing risk [fact/estimate]
Any hyperscaler/neocloud capex pullback hits contracted-backlog quality and the equity's multiple first [estimate]
Trends
Converts volatile hash-price revenue (~$3-5M/MW/yr) into contracted, credit-backstopped 10-15yr leases at higher multiples — WULF is the cleanest pure-play [fact]
Drives neocloud (Fluidstack) demand for the contracted capacity and the value of energized power [fact]
Scarcity supports lease economics for those with energized power (positive), but interconnect + regulatory approval gate activation (negative). Zero-carbon positioning partly offsets [fact]
Backlogs resting on one neocloud (Fluidstack) whose credit is the real thesis raise confidence-shock fragility even with a Google backstop [estimate]
Cheap-enough capital enables the pivot (positive) but leaves large GAAP losses + ~$5.8B debt exposed to a rate/risk-off turn (negative) [fact/estimate]
Lake Mariner's ~89% carbon-free hydro/nuclear power is a durable differentiator as ESG + grid pressure rises [fact]
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
GPUs deployed by the tenant (Fluidstack) in WULF's colocation shells — the compute inside the leased critical IT
EPC / preconstruction partner for the ~$3-4B Hawesville, KY campus build-out
NYISO Zone-A grid (Niagara hydro + regional nuclear) Zero-carbon power source for Lake Mariner (~89% carbon-free) — the scarce input
Thermal, power-distribution, and rack infrastructure for high-density HPC colocation shells
Anchor HPC colocation tenant (neocloud); ~360MW contracted critical IT at Lake Mariner on 10-yr leases, Google-backstopped
Lease backstop ($1.8B → ~$3.2B) + warrants toward a ~14% equity stake — a strategic backer/validator, not a direct tenant
Structural twin — entire 300MW Barber Lake (TX) leased to Google-backstopped Fluidstack, full project ownership retained, but a more hybrid model that keeps meaningful mining [fact]
Australia-domiciled (non-US), US-listed peer with a stronger direct anchor — Microsoft (~$9.7B) + NVIDIA (~$3.4B) — and an owned-GPU AI Cloud model rather than shell leasing [fact]
AI-factory landlord anchored by CoreWeave (~400MW / ~$7B) + a separate ~$5B IG-hyperscaler lease at Polaris Forge ND; pursuing a REIT conversion [fact]
Largest contracted HPC backlog of the pure pivots (~590MW / >$10B to CoreWeave); stayed independent after shareholders rejected CoreWeave's all-stock buyout (Oct 2025) [fact]
The competing neocloud counterparty behind CORZ/APLD backlogs; sets the alternative deal structure to WULF's Fluidstack/Google cluster [fact]
Private energy-first AI-factory analog (OpenAI/Stargate, Abilene TX); NOT publicly tradable, but the pure-play private benchmark for the power-first model [fact]