TeraWulf Inc. (NASDAQ: WULF) — A Power Developer Priced as a Hyperscale Landlord, Funded on the Stock Itself

Date: June 12, 2026 Price (ref.): ~$26 / share · Market cap: ~$13.0B · Enterprise value: ~$15.2B (incl. ~$845M Google-warrant liability) Shares out: ~444.5M basic (12/31/25); ~700M+ fully diluted · 52-wk range: $3.40 – $27.78 · Beta: ~4.3 · Short interest: ~26% of float

⚡ The Author’s Take

This block is the author’s own independent opinion and general information only — not investment advice, and not a recommendation to buy or sell any security. The analysis that follows it takes no position and carries no price target; it discusses valuation only as embedded expectations. Do your own research.

Verdict: AVOID at ~$26 / HOLD-to-trim for existing holders. Not a short despite the valuation — the borrow is crowded (~26% of float), the story has real assets behind it, and a single hyperscaler headline can squeeze it 30% overnight. Constructive accumulation zone is the low-to-mid teens (≈$12–16), where you would be paying a more defensible multiple of the contracted 606 MW rather than an unsigned multi-gigawatt dream. Conviction: medium.

Tag: “The power guy at the door of the AI party — invited, but paying a cover charge sized for a building he hasn’t built yet.” TeraWulf is, at its core, a genuinely capable power and infrastructure developer — a 25-year power-plant operator (Paul Prager’s Beowulf lineage) that controls scarce, low-carbon, grid-connected sites at exactly the moment power is the binding constraint on AI. That is real, and it is why Core42, Fluidstack, and Google are at the table. But the market is not pricing the 606 MW of critical-IT load WULF has actually contracted; at ~$15.2B EV that is ~$25M per contracted MW — roughly double the $8–15M/MW that operating hyperscale capacity transacts at privately, and richer than even APLD or IREN screen. The price already capitalizes a large slice of the uncontracted ~1.75 GW development pipeline as if it were signed, funded, and stabilized — when in fact it is none of the three. Meanwhile the equity is being diluted hard: ~14% of the company handed to Google in penny warrants as credit support, $2.5B of converts struck from $8.48 to $19.94, and $101M of stock comp in a single quarter against $34M of revenue. The reported GAAP losses (−$661M FY2025, −$428M Q1’26) are mostly the non-cash, “good-news” markup of the Google warrant liability as the stock rose — so they overstate the operating bleed, but they understate the per-share dilution that is the real cost.

What the market is pricing correctly: power-secured, hyperscaler-credit-backed data-center capacity is scarce and valuable, and WULF has a credible team and a marquee anchor (the Google backstop is a genuine credit enhancement that pure neocloud landlords like APLD’s CoreWeave relationship lack). What it is pricing incorrectly: that the 606 contracted MW plus an unsigned pipeline plus ~$3B of remaining construction capex plus a negative-equity, serially-diluting balance sheet all convert cleanly into stabilized NOI accruing to today’s shares. They might. But you are paying the success price for a binary that still has three or four execution gates (CB-4/CB-5 energization, the Kentucky “Justified” anchor signing, Morgantown FERC, and refinancing the 7.75% project notes into IG debt) in front of it. Flip-to-bullish trigger: a signed investment-grade hyperscaler lease at Hawesville/Kentucky and the incremental 250 MW Lake Mariner interconnect, which together would convert the “pipeline” into contracted backlog and justify the multiple on signed MW. Flip-to-bearish trigger: any Fluidstack/neocloud counterparty wobble, a CB-4/CB-5 slip, or a capital-markets window closing while ~$3B of capex is still unfunded-to-completion — at which point a negative-equity, FCF-negative developer re-rates from “AI platform” to “leveraged contractor.”

1. Executive Summary

TeraWulf is a Maryland-based digital-infrastructure company in the middle of a deliberate, capital-intensive metamorphosis: from a pure-play Bitcoin miner into a landlord of power-secured, high-performance-computing (HPC) data-center capacity leased to AI cloud and hyperscale tenants. Its flagship asset is the Lake Mariner campus in upstate New York — a zero-carbon (hydro/nuclear-adjacent grid) site where it has energized 60 MW of critical IT load for Core42 (a G42/UAE entity) and is building ~378 MW more for Fluidstack under leases backstopped by Google. A 168 MW joint venture in Abernathy, Texas (50.1% WULF, pre-leased to Fluidstack) and new sites in Hawesville, Kentucky and Morgantown, Maryland round out a multi-gigawatt ambition.

The financial picture is two businesses fused at the hip. The legacy mining business (≈5–6 EH/s, ~$13M of quarterly revenue, winding down toward zero “by the next halving”) is a self-funding “melting ice cube” that bootstrapped the buildout. The emergent HPC-leasing business booked $21M of high-margin (~85% stabilized) lease revenue in Q1 2026 and is the entire reason the equity trades at ~90x trailing sales. Total contracted critical-IT load is ~606 MW; reported FY2025 revenue was $168.5M and GAAP net loss was −$661.4M, the bulk of it a non-cash markup of the Google warrant liability as the share price rose ~7x off its 52-week low.

The balance sheet has been transformed in twelve months: total assets from $787M to $7.0B, funded by $3.2B of 7.75% secured project notes and ~$2.5B of convertible notes, leaving ~$4.6B of debt, $3.1B of cash, and — after Q1 2026 — negative book equity. Capex ran $1.06B in FY2025 and $523M in Q1 2026 alone, with ~$3B more to complete the contracted builds. Operating cash flow is negative. The company is, in effect, a development-stage infrastructure platform whose realized economics are a fraction of its capitalized expectations.

This memo’s central question is not whether WULF’s assets can earn good returns — power-secured hyperscale capacity at ~$130/kW/month and ~85% NOI margins can — but whether those asset-level returns survive the trip to per-share value after ~14% warrant dilution to Google, $2.5B of convertible overhang, $100M/quarter of stock comp, a related-party-laden governance structure, and a multiple that already capitalizes capacity WULF has not yet contracted, built, or financed to completion. The verdict, argued section by section below, is that WULF is a credible operator attached to a demanding price.

2. Business Overview

What the company does. TeraWulf owns and develops digital-infrastructure sites in the United States, monetizing controlled electrical power two ways: (1) Bitcoin mining — running ASIC miners that convert electricity into Bitcoin, which the company liquidates almost immediately (it held just 3 BTC at 12/31/2025, fair value ~$0.3M); and (2) HPC hosting / data-center leasing — building out powered, cooled, fitted-out data-center “buildings” (the “CB” series at Lake Mariner) and leasing that critical-IT capacity to AI-compute tenants under long-term contracts. The strategic thrust, stated bluntly by management, is that mining “served its purpose” as a way to build infrastructure and monetize power, but “the future of this platform is contracted long-duration compute infrastructure” (Q1 2026 earnings call, May 8, 2026).

Revenue segmentation. WULF reports two segments:

• Digital Asset Mining. ~$13M in Q1 2026 (down sharply YoY as capacity is repurposed). The economics are now driven less by Bitcoin price and more by the site’s flexible load profile and grid demand-response participation — in Q1 2026, $14.1M of demand-response proceeds were recorded as a reduction of cost of revenue, which is why mining produced “strong operating profit” even as it shrinks. Mining is run as a cash-generating runoff sleeve, expected to exit “by the next halving.”
• HPC Hosting / Leasing. $21M in Q1 2026 (up 117% from $9.7M in Q4 2025), the first quarter HPC leasing “is meaningfully reflected in our financials.” This is the growth engine and the valuation driver.

How it makes money — the landlord model (critical). WULF is a landlord, not a neocloud. It builds the powered shell and fit-out; the tenant (Core42, Fluidstack) brings and owns the GPUs. WULF collects long-dated, largely fixed lease payments (triple-net-style, with escalators) at ~85% segment profit margin once stabilized. This is structurally the Applied Digital (APLD) model — and crucially different from the IREN/CoreWeave “neocloud” model where the operator owns the GPUs and bears their depreciation and utilization risk. WULF therefore carries no GPU-obsolescence risk but full real-estate/counterparty/power risk. This distinction governs both the moat analysis and the valuation method.

Customer / end markets. Three anchor counterparties define the business today: Core42 (60 MW, a G42 subsidiary — UAE sovereign-adjacent, Microsoft-invested), Fluidstack (378 MW at Lake Mariner + 168 MW at Abernathy, an AI cloud startup whose lease obligations are backstopped by Google), and the joint-venture/development pipeline. End demand is the AI training/inference buildout by hyperscalers and AI-compute platforms.

Recurring vs. non-recurring. The HPC leases are long-dated (10-year base + two 5-year options at Lake Mariner; 25 years at Abernathy) and contractual — high-quality recurring revenue once commenced. But as of 12/31/2025 only the Core42 leases had commenced (18 of 60 critical-IT MW); the Fluidstack and Abernathy revenue is contracted but not yet recognized (RPO ≈ $0 on the GAAP lessor table beyond Core42). Mining revenue is non-recurring/commodity and shrinking by design.

Corporate structure. TeraWulf trades on Nasdaq under a legacy CIK (1083301) inherited from a 2021 reverse merger with IKONICS Corporation — which is why third-party data feeds mislabel it as a 1996-IPO “Financial Services / Capital Markets” company; ignore that classification. The HPC build is housed in subsidiaries (La Lupa for Core42, Akela for Fluidstack, WULF Compute LLC as the project-finance vehicle for the $3.2B secured notes), with the Abernathy JV (FS CS 1 LLC) held 50.1%.

Verdict. A real, power-anchored infrastructure business executing a credible pivot from a commoditized activity (mining) to a higher-quality one (contracted leasing). The business model is sound and the landlord structure is the right one (no GPU risk). The open question is not the model but its scale of realization relative to its valuation — today the recurring, recognized revenue is a rounding error against the market cap.

3. Industry Dynamics

The structural backdrop: power is the binding constraint. The AI buildout has shifted the bottleneck from chips to electricity and the infrastructure to deliver it. As management frames it — and the framing is correct — “the constraint is not GPUs. It is power,” with interconnection delays, transmission limits, and the multi-year lead time to bring new generation online creating a genuine scarcity of shovel-ready, power-secured, grid-connected sites. Hyperscaler capex is running at a ~$700B/year clip across the majors, chasing capacity that physically cannot be built fast enough. This is a real, durable tailwind and the single strongest pillar of the bull case.

Where WULF sits in the value chain. WULF occupies the “powered land + shell + fit-out” rung — between the utility/generator (upstream) and the GPU-owning cloud (downstream). Its differentiated claim is that it spans further upstream than most data-center developers: it can source, permit, and operate generation (Prager’s 25-year power-plant background), not merely lease grid capacity. In a world where utilities increasingly demand “bring your own generation” or surplus-generation commitments to grant interconnection, that upstream capability is a genuine edge. WULF’s three “paths to power” — immediate access (Hawesville), bring-your-own-generation (Morgantown’s gas + battery build), and utility partnerships — are a coherent strategy for a power-constrained market.

Market size and growth. The addressable market — contracted hyperscale/AI data-center capacity — is enormous and growing, but it is also drawing enormous capital. Lease rates for liquid-cooled hyperscale capacity run ~$130/kW/month with 15-year terms and annual escalators; NOI margins at the site level reach the high-80s%. On 1 GW of critical IT load that arithmetic implies ~$1.5B+ of gross rent and ~$1B+ of NOI — the math that animates the entire cohort’s valuations.

Competitive intensity and the capital cycle (Marathon lens). This is where skepticism enters. High, visible returns on power-secured capacity are attracting a flood of supply: the incumbent REITs (Digital Realty, Equinix), the hyperscalers’ own self-build, private capital (Blackstone/QTS, DataBank), the neoclouds (CoreWeave, Nebius, Crusoe), and the entire converted-Bitcoin-miner cohort (CORZ, CIFR, IREN, APLD, MARA, HUT, and WULF itself). Marathon’s capital-cycle framework warns precisely against extrapolating today’s IRRs in an industry where asset growth is vertical and capital is rushing in: high returns attract capital, capital builds capacity, capacity compresses returns. The power constraint slows the cycle (you cannot conjure interconnection), but it does not repeal it — the same constraint that protects today’s lease rates is being attacked by every well-capitalized developer racing to secure the next site.

Regulation and the social license. Data-center development faces rising NIMBY/political resistance over power prices, water, and grid strain. WULF’s Morgantown (MD) and Cayuga (NY) projects are in planning/FERC processes; Morgantown’s FERC decision is expected mid-summer 2026. Management’s “lead with transparency / educate the community” posture is sensible, but permitting and interconnection timing are genuine, externally-controlled risks that can slip the pipeline by quarters or years.

Verdict: a structurally attractive demand environment attached to an increasingly crowded supply response. The industry is good for those who already control scarce power; it is structurally less good as a place to deploy fresh capital at today’s prices, because the capital cycle is mid-to-late and the moat (power access) is replicable by anyone with capital and patience. WULF is on the right side of the demand curve but is paying — and asking shareholders to pay — late-cycle prices to expand supply.

4. Competitive Position

Name the moat — and pressure-test it. Run WULF through Greenwald’s taxonomy (supply/cost advantage; demand/customer captivity; economies of scale + captivity):

• Cost/supply advantage — partial and erodible. WULF’s strongest claim. Lake Mariner’s low-cost, zero-carbon power and the team’s ability to source/permit/operate generation is a real cost-and-access edge versus a generic developer. But it is site-specific and replicable: it is not a structural cost advantage that competitors cannot match by securing their own power. Every miner-turned-landlord and every REIT is chasing the same scarce interconnection. A cost advantage that any well-funded rival can replicate at the next substation is a head start, not a moat.

• Customer captivity / switching costs — present once a lease commences, absent before. A 10–25 year signed lease is high switching cost — the tenant has co-designed the building around its hardware and cannot easily relocate GPUs mid-term. This is real captivity, but it is contractual, not franchise-based, and it only exists after signing. For the uncontracted pipeline (the bulk of the valuation), there is zero captivity — WULF competes site-by-site, deal-by-deal, against the whole field, as management candidly described (“we have one site we’re focused on at one time… it’s probably not 100% of the customers we want of what they’re doing”).

• Economies of scale + captivity — not yet. WULF is sub-scale versus REITs and hyperscalers; it has no network effect and no scale-based cost advantage that compounds. Its ~606 contracted MW is meaningful but small against the field.

The Google backstop — a genuine, differentiated credit enhancement. The single most distinctive feature of WULF’s competitive position is that Google guarantees Fluidstack’s lease obligations (Google will either pay the termination fee or assume the lease as tenant on a Fluidstack default), in exchange for which it received 73.58M penny warrants. This is a real mitigant to the counterparty risk that plagues the neocloud-landlord model — it effectively converts a fragile, FCF-negative AI startup’s credit into something closer to Google’s. Cipher Mining (CIFR) has the structurally-identical arrangement; APLD (reliant on CoreWeave) does not. This is the best argument that WULF’s contracted revenue is higher-quality than peers’ — but note it cost shareholders ~14% of the company.

Direct competitive comparison. Against CORZ (CoreWeave-anchored, ~590+ MW, post-bankruptcy, subject of a contested CoreWeave acquisition), CIFR (Google/Fluidstack/AWS-backed, but the thinnest cash-to-debt in the cohort), IREN (neocloud, owns GPUs, Microsoft deal), and APLD (CoreWeave landlord), WULF is mid-pack on contracted MW, best-in-class on counterparty credit support (Google), and the most expensive on EV per contracted MW. Its differentiator is the power-developer DNA — it is the most credible of the cohort at the upstream generation/permitting layer.

Verdict: a head start, not a durable advantage. WULF has a real, present edge in power sourcing and a genuinely better counterparty-credit structure than most peers. But under the Greenwald tests, this is not a franchise: the moat is replicable (power access), contractual (lease captivity exists only post-signing), and sub-scale. The competitive position justifies WULF being in the game and earning attractive project IRRs; it does not justify a franchise multiple. If the power advantage deteriorated tomorrow (a rival secures the adjacent interconnect), the contracted leases would survive but the growth — which is what the price pays for — would not.

5. Growth History and Forward Opportunities

Historical growth. Revenue grew from $69.2M (FY2023) to $140.1M (FY2024) to $168.5M (FY2025) — a respectable trajectory, but one driven by mining capacity additions and Bitcoin price, not yet by the HPC pivot. Tellingly, Q1 2026 revenue of $34.0M was essentially flat versus Q1 2025’s $34.4M and down from Q4 2025’s $35.8M — because mining revenue is shrinking faster than HPC leasing is ramping. The headline top line is in a transitional trough; the mix, not the level, is the story.

The HPC ramp — the only growth that matters. HPC lease revenue went $9.7M (Q4’25) → $21M (Q1’26), and management has laid out a building-by-building energization cadence: Core42’s 60 MW fully delivered (Q1’26); Fluidstack CB-3 energizing by end of Q2’26, CB-4 in Q3’26, CB-5 in Q4’26; Abernathy (168 MW) targeted 2H’26. As these commence, the ~606 MW of contracted critical-IT load converts to recognized, ~85%-margin recurring revenue. At ~$130/kW/month, 606 MW at full ramp implies on the order of ~$900M+ of annualized gross lease revenue and ~$750–800M of stabilized NOI — versus $168M of total revenue today. That gap is the growth the equity is paying for, and most of it is contracted and visible, not speculative.

Forward pipeline — large, but uncontracted. Beyond the 606 contracted MW, management points to: Lake Mariner’s potential expansion to ~750 MW plus an incremental 250 MW interconnect (ISO feedback expected mid-2026); Cayuga (NY) at up to 400 MW gross; Hawesville (KY) at 480 MW targeted online 2H’27 with a “highly confident” anchor-customer signing expected in Q2’26 (the “Justified” project, management implying an investment-grade hyperscaler); Morgantown (MD) — a large brownfield gas+battery site near the DC/Northern-Virginia corridor pending FERC; and a stated 1.75 GW portfolio of uncontracted development sites. Management guides to contracting 250–500 MW of new critical-IT capacity per year.

Quality of the growth. The contracted growth (606 MW) is high-quality: long-dated, high-margin, credit-supported. The pipeline growth (the multi-GW ambition) is lower-quality for valuation purposes — it is unsigned, requires ~$10M+/MW of fresh capex the company does not yet have, and competes against the whole field. The bull case treats the pipeline as nearly-certain; the bear case treats it as a series of options each gated on a customer signature, a FERC ruling, an interconnect award, and a financing.

Verdict: high-quality contracted growth wrapped inside a lower-quality speculative pipeline. The 606 MW ramp is real and largely de-risked on the demand side (signed, credit-backed). But the valuation requires the pipeline to convert at the guided 250–500 MW/year and be financed without crushing dilution — neither of which is yet evidenced. The growth is genuine but front-loaded with execution and funding gates.

6. Financial Quality

Revenue and margins. FY2025 revenue $168.5M; cost of revenue (ex-D&A) $82.7M. The Q1 2026 income statement is the cleaner read of the future mix: revenue $34M, cost of revenue ex-D&A just $2.4M (after a $14.1M demand-response credit), HPC segment profit margin ~50% as-reported but ~85% after adjusting for one-time tenant fit-out, pre-revenue WULF Compute costs, and development costs on the 1.75 GW uncontracted portfolio. The 85% stabilized lease margin is the key economic fact and, if sustained at scale, validates the landlord model.

The GAAP losses are mostly non-cash and “good-news” driven. FY2025 GAAP net loss was −$661.4M and Q1 2026 was −$427.6M — figures that look catastrophic against $168M of revenue. They are not what they appear:

• The dominant driver is the Google warrant fair-value remeasurement. Because the 73.58M penny warrants are liability-classified, a rising stock price increases the liability and generates a non-cash loss. FY2025 booked a $429.8M warrant/derivative FV loss ($329.2M Google warrants + $100.6M on the 2031 convert’s conversion feature); Q1 2026 booked a further −$216.3M. The warrant liability sat at $844.7M at 12/31/2025. This is a “good-news loss” — it grows as the equity appreciates — and it does not touch liquidity.
• The second driver is stock-based compensation, which exploded to $101M in Q1 2026 (vs. $51M for all of FY2025), as price-hurdle PSUs ($6.00/$6.50, all achieved Sept 2025) and large 2025/2026 management grants vested.
• Underlying operating reality: Q1 2026 adjusted EBITDA was −$4.1M (improving from −$50.9M in Q4’25); operating loss was −$162M (inflated by the $101M SBC and a $25.7M impairment). The cash operating bleed is modest and narrowing as HPC revenue ramps; the GAAP loss is dominated by non-cash marks.

The quality-of-earnings nuance cuts both ways. Stripping the warrant mark flatters the loss — good. But SBC is a real economic cost (it dilutes shareholders), and “adjusted EBITDA” that adds back $100M/quarter of stock comp is a flattered number. The honest read: WULF is roughly cash-operating-breakeven today on a tiny revenue base, with real per-share dilution running through both SBC and the warrant/convert overhang.

Cash flow and capital intensity. Operating cash flow was −$24M (FY24), −$123M (FY25), −$18M (Q1’26). Capex was $268M (FY24), $1.06B (FY25), and $523M in Q1’26 alone, plus $450M into the Abernathy JV. This is an extraordinarily capital-hungry business mid-build: ~$3B of capex remains to complete the contracted projects (WULF Compute ~$2.2B remaining; Abernathy ~$0.9B remaining). Free cash flow is deeply negative and will remain so until the contracted MW stabilize.

Balance sheet — transformed and stretched. In twelve months: total assets $787M → $7.0B; cash + restricted $274M → $3.1B; total debt $0 → ~$4.6B ($3.2B 7.75% secured notes + ~$2.5B converts at carrying ~$1.58B); and stockholders’ equity went negative (−$78.8M at 3/31/26) as accumulated losses (mostly the warrant mark) overwhelmed paid-in capital. ROE/ROIC/P/B are meaningless here (negative equity). Liquidity is the relevant lens: ~$3.1B of cash at 3/31, of which only ~$300M was unrestricted at the parent (rising to ~$1.5B after the April 2026 equity raise); the rest is ring-fenced in project accounts (WULF Compute $2.8B gross, Abernathy $1.4B gross) with debt-service reserves.

Verdict: economics that improve with scale, attached to a balance sheet that depends on continuous capital-market access. The unit economics (85% lease margin, 12–15% NOI yield on cost) genuinely improve as the contracted MW come online — that is the bull’s strongest financial point. But the company is FCF-negative, has negative book equity, faces ~$3B of remaining capex, and is funded by a stack (7.75% project notes, converts, penny warrants, $100M/quarter SBC) that only works if the equity stays elevated and the debt windows stay open. This is high-quality project economics riding on a fragile corporate balance sheet.

7. Capital Allocation

The core allocation decision: convert mining cash flow + ~$5.5B of raised capital into a hyperscale-leasing platform. On its own terms this is defensible — redeploying a commoditizing activity (mining) into a higher-quality one (contracted leasing) is exactly the kind of pivot good capital allocators make. The execution, however, raises three flags.

Flag 1 — dilution as the primary funding currency. Shareholders have absorbed: ~14% of the company in Google penny warrants (73.58M shares at $0.01, received for credit support, not cash); $2.5B of convertible notes struck from $8.48 to $19.94 (a large future share issuance, only partially offset by capped calls); and stock-based comp that hit $101M in a single quarter. The Google warrants alone are a remarkable giveaway — necessary to secure the Fluidstack credit backstop, but a ~$1.9B economic transfer (at the current price) to obtain a counterparty guarantee. Capital was raised aggressively and well-timed (the converts and secured notes were placed as the stock ran), but the per-share cost is steep.

Flag 2 — related-party transactions with the CEO. This is the governance soft spot. CEO Paul Prager (8.5% owner, no super-voting stock) has a web of affiliate dealings the company has been steadily internalizing on terms it sets:

• Beowulf E&D (Prager-controlled) provided administrative/infrastructure services for ~$15M/year, then was acquired by TeraWulf in May 2025 from a Prager-controlled holdco for ~$53M ($3M cash + 5.0M shares + earnouts up to $19M cash and $13M shares), creating $55.5M of goodwill and an obligation to fund an employee trust at 2% of annual capex. The auditor flagged this as a critical audit matter due to the related-party relationship.
• Cayuga ground lease — the lessor (Cayuga Operating Co.) is CEO-controlled; WULF prepaid $95M in stock (18.6M shares) plus $3M cash on an 80-year lease with a $100 purchase option after year 50.
• Office leases were also acquired from Prager-controlled counterparties.

None of these is necessarily abusive, and internalizing a related-party service provider can be cleaner than leaving it outside. But the pattern — recurring affiliate fees converted into equity and goodwill on negotiated terms with the controlling CEO — warrants close diligence and is a real governance discount.

Flag 3 — compensation keyed to stock price and deal-making, not per-share value. Prager’s FY2025 total comp was $39.4M (incl. a $15M discretionary cash bonus and $23.2M of stock awards); the other three NEOs each took $20–30M. The incentive design rewards stock-price hurdles (the $6.00/$6.50 PSU hurdles, all cleared by Sept 2025) and discretionary “capital formation / strategic transaction” bonuses — i.e., it pays for raising capital and moving the share price, not for ROIC, FCF, or per-share value creation. In a company whose chief risk is value-destructive dilution, comp that rewards capital-raising and price momentum is misaligned with the long-term shareholder.

Buybacks — authorized then idle. A $200M repurchase program was authorized in October 2024; $0 was used in Q4 2025 (the program sits fully available), though ~$151.5M of treasury stock was acquired in FY2024/early 2025 before the pivot to massive capex. Capital allocation correctly flipped from returns to growth — buying back stock while facing $3B of capex would be indefensible — but it underscores that there is no return-of-capital pillar here for years.

Insider behavior. The available Section 16 signal tilts toward distribution, not accumulation: CEO Prager filed code-S open-market sales in May 2026 (near all-time highs), alongside option-exercise (M) and disposition (D) activity, and there are 7 Form 144 sale notices clustered across 2H2025–2026. No discretionary open-market purchases (code P) were identified. (Full Form 4 transaction-level quantification was not completed for this memo and is an open item — see.) Insiders selling into the AI-pivot rally is not damning on its own — much is mechanical PSU-vesting and tax — but it is not the conviction signal a bull would want.

Verdict: a defensible strategic pivot executed with aggressive, well-timed capital raising but shareholder-unfriendly mechanics. Management has assembled real assets and real contracts. But it has done so by handing ~14% of the company to Google, layering on $2.5B of convertible dilution and $100M/quarter of SBC, transacting with CEO affiliates, paying itself $40M in a loss-making year on stock-price hurdles, and selling shares into the rally. The capital formation is impressive; the capital allocation to the existing shareholder is mediocre.

8. Changes and Headwinds — Last Two Years

The transformation timeline (2024–2026):

• Oct 2024 — Sold its entire 25% stake in the Nautilus Cryptomine JV (the Talen/Susquehanna nuclear-powered mining JV) for $102.1M, booking a $22.6M gain — exiting a passive nuclear-mining interest to focus capital on owned HPC infrastructure.
• Dec 2024 — Signed the Core42 lease (60 MW, 10+5+5 years) — the first HPC anchor.
• Aug 2025 — Signed the Fluidstack leases (378 MW at Lake Mariner) with the Google backstop and 73.58M penny warrants — the transformational deal that re-rated the stock.
• Sep 2025 — Stock-price PSU hurdles ($6.00/$6.50) achieved; the equity ran ~7x off its 52-week low.
• Oct 2025 — Issued $3.2B of 7.75% secured notes via WULF Compute (project finance for Lake Mariner); formed the Abernathy, TX JV (50.1%, 168 MW, pre-leased to Fluidstack).
• May 2025 — Internalized Beowulf E&D (related-party acquisition).
• 2025–2026 — Issued $2.5B of convertibles ($500M 2.75% '30, $1.0B 1.00% '31, $1.025B 0% '32).
• Q1–Q2 2026 — Core42 fully energized (60 MW); HPC leasing first material in the financials; added Hawesville (KY) and progressing Morgantown (MD, FERC mid-summer); raised ~$1.2B of equity YTD; secured a $250M revolver from 8 banks; repaid/terminated the $100M Kentucky bridge facility.

Headwinds and open fronts:

• Construction/energization risk on CB-4/CB-5 (Q3/Q4 2026) and Abernathy — management has already noted “customer-driven design refinements” and hardware-iteration-driven scope changes, a euphemism for the moving-target nature of building to evolving GPU specs.
• The “Justified”/Kentucky anchor is not yet signed — management is “highly confident” of a Q2 2026 signing with an “investment-grade, super high-quality customer,” but it remains a promise, not a contract.
• Morgantown FERC approval (mid-summer 2026) and the incremental 250 MW Lake Mariner interconnect (ISO feedback mid-2026) are externally-controlled gating items.
• Negative book equity and continuous funding need — ~$3B of remaining capex against a balance sheet that depends on capital-market access.
• Bitcoin runoff — as mining winds down, the self-funding cash bridge thins (though demand-response economics partly offset).
• Political/NIMBY resistance to data-center development is “slowly but surely picking up” (management’s words), a slow-building headwind to the pipeline.

Verdict: the changes overwhelmingly strengthen the business and complicate the equity. Two years converted a struggling miner into a credibly-financed HPC-infrastructure platform with marquee counterparties — a genuine strategic success. But the same period loaded the share count and balance sheet with dilution and debt, and left the thesis dependent on a sequence of not-yet-completed energizations, signings, and approvals.

9. Risk Analysis (Risk Matrix)

Net risk read: the dominant, correlated risk is funding-meets-execution — a multi-billion-dollar build, financed continuously, dependent on energizations and signings landing on schedule while capital markets stay open. The Google backstop materially de-risks the demand/credit side; nothing de-risks the funding/construction side except continued access to capital, which is itself a function of the share price staying elevated — a reflexive loop that works on the way up and bites on the way down.

10. Valuation Discussion (Embedded Expectations)

No price target and no recommendation in this section — embedded-expectations and scenario framing only.

Why trailing multiples are meaningless. At ~$15.2B EV on $168M of revenue, WULF screens at ~90x EV/Sales and ~60x P/S (96th percentile of its own short history). EV/EBITDA is negative/non-meaningful. P/B is negative (negative equity). None of these is a valuation — they merely confirm that the market is paying entirely for future contracted and pipeline NOI, not current results. The only coherent way to value WULF is to capitalize its contracted (and prospective) lease economics and subordinate the common to the capital stack.

The contracted-MW lens. WULF has ~606 MW of contracted critical-IT load (Core42 60 + Fluidstack 378 + Abernathy 168 × 50.1% ≈ 84 net, or 168 gross). On gross contracted MW (~606), EV/contracted-MW ≈ $25M/MW — roughly double the $8–15M/MW that operating hyperscale capacity transacts at in private markets, and richer than APLD (~$12–18M/MW) or the CoreWeave/Core Scientific deal comp. On a net-to-WULF basis (counting only 50.1% of Abernathy) the figure is higher still. The market is plainly not paying ~$25M/MW for operating assets — it is paying for the uncontracted pipeline (1.75 GW) as if a large fraction were already signed and built.

The stabilized-NOI lens. Capitalize the contracted book: 606 MW × $130/kW/month × 12 ≈ $945M gross rent; at ~85% margin ≈ $800M stabilized NOI. Against ~$15.2B EV, that is a ~5.3% NOI yield / ~19x NOI — fair-ish versus data-center REITs at 4–5% cap rates (20–25x), but only if (a) all 606 MW reaches stabilization, (b) ~$3B of remaining capex is funded without further dilution, and © the NOI accrues to the EV holder ahead of the ~$248M/year of 7.75%-note interest and the convert/warrant claims. Net of those frictions, the contracted book alone does not comfortably support today’s EV — which means the pipeline must deliver.

Embedded expectations — what must be true for ~$26. Reverse-engineering the price: the market is underwriting that WULF (i) energizes the full 606 MW contracted book on schedule at ~85% margins, and (ii) converts a large slice of the 1.75 GW pipeline — call it another ~600–1,000+ MW — into contracted, financed, stabilized capacity over the next several years at attractive NOI yields, and (iii) refinances the 7.75% project notes into cheaper investment-grade debt to “unlock” equity free cash flow, and (iv) does all of this without the share count ballooning beyond what is already embedded. In effect, the price capitalizes ~1.2–1.6 GW of stabilized, financed capacity when only ~0.6 GW is contracted and a fraction of that is energized. That is a demanding, multi-year, multi-gate set of assumptions.

Scenario sketch (illustrative, not targets):

• Bull: Kentucky anchor signs (IG hyperscaler), 250 MW interconnect awarded, pipeline converts at the guided 250–500 MW/year, 7.75% notes refi to IG — the contracted book grows to 1.5+ GW of high-margin NOI, the multiple holds on a “scaling AI-infrastructure platform,” and per-share value compounds despite dilution. The stock works from here.
• Base: the 606 MW ramps roughly on schedule but the pipeline converts more slowly and more dilutively; stabilized NOI lands meaningfully below the capitalized expectation; the multiple compresses toward an infrastructure (not platform) rate; per-share returns are mediocre as dilution absorbs the asset-level value creation.
• Bear: a counterparty wobble (Fluidstack stress, even with backstop friction), a CB-4/CB-5 or Abernathy slip, a denied interconnect or FERC delay, or a capital-markets window closing mid-build — and a negative-equity, FCF-negative developer re-rates hard from “AI platform” to “leveraged contractor,” with the converts and 7.75% notes sitting ahead of a heavily-impaired common.

The per-share discipline. The single most important valuation point — shared with the IREN and APLD analyses — is that asset-level IRRs can be excellent and the equity still expensive, because ~14% warrant dilution, $2.5B of converts, $100M/quarter of SBC, and $4.6B of senior debt sit between the project returns and per-share value. WULF’s projects may well earn 12–15% unlevered NOI yields on cost; that does not mean the common at ~$26 is cheap.

Verdict (embedded expectations): the price discounts near-flawless conversion of an unsigned pipeline on top of a contracted book that is itself only partly energized and not yet fully financed to completion. The contracted assets are real and valuable; the equity is priced for the pipeline, the refinancing, and the absence of dilution surprises — three things not yet in evidence.

11. Variant Perception

Consensus belief. The sell side is overwhelmingly bullish (≈10 buy/strong-buy vs. 1 hold; mean target ~$35). The consensus narrative: WULF is a scarce, power-secured, Google-backed AI-infrastructure platform with a multi-gigawatt runway, early in a multi-year contracted-revenue ramp, and the stock is a way to own “AI picks-and-shovels” with a credit-enhanced demand side. Simultaneously, ~26% of the float is sold short — so the market is sharply divided, with momentum/growth longs against valuation/dilution shorts. This is a battleground stock.

Strongest bull case. Power is the binding constraint of the AI era; WULF controls scarce, low-carbon, grid-connected power and has the rare team that can source, permit, and operate generation — the most defensible upstream position in the converted-miner cohort. It has converted that into 606 MW of contracted, long-dated, high-margin leases with the best counterparty credit support in the group (the Google backstop). With $3.1B of cash, a fully-funded near-term pipeline, and a 1.75 GW development portfolio, it can compound contracted NOI at 250–500 MW/year, refinance into IG debt, and “unlock” a large equity FCF stream “only a couple of years away” (CFO). At a 4–5% REIT cap rate on stabilized NOI, the multi-GW vision supports a far higher equity value than today’s.

Strongest bear case. This is a FCF-negative, negative-book-equity developer with ~$3B of remaining capex, funded by 7.75% project notes and serial dilution, trading at ~$25M per contracted MW — double the price of operating capacity — because the market is capitalizing an unsigned pipeline as if it were built. The “moat” (power access) is replicable; the captivity is contractual and only post-signing; the cohort and the REITs are racing to add supply into a late-stage capital cycle that will compress the very lease rates the bull extrapolates. Per-share value is being eroded by ~14% Google warrants, $2.5B of converts, and $100M/quarter of SBC, while the CEO sells shares and pays himself $39M on stock-price hurdles in a loss-making year. Any execution, counterparty, or funding stumble re-rates the common violently.

The 3–5 assumptions that matter most:

1. Does the contracted 606 MW energize on schedule at ~85% margins? (Largely de-risked; the nearest-term, highest-confidence variable.)
2. Does the uncontracted pipeline convert at 250–500 MW/year — starting with the unsigned Kentucky anchor? (The swing factor for the valuation; unproven.)
3. Can WULF fund ~$3B of remaining capex without value-destructive dilution, and refinance the 7.75% notes into IG debt? (The reflexive funding question.)
4. Does the capital cycle compress lease rates / NOI yields before WULF scales? (The Marathon risk; unknowable but directionally adverse.)
5. How much does per-share value leak to warrants, converts, and SBC even if the assets perform? (The dilution drag; quantifiable and material.)

What would falsify each side. Bull falsified by: an unsigned/slipped Kentucky anchor, a Fluidstack/neocloud counterparty event, a CB-4/CB-5 delay, or an equity raise at a depressed price. Bear falsified by: a signed IG hyperscaler lease + the 250 MW interconnect award (pipeline → backlog), plus an IG refinancing of the 7.75% notes — which together would convert the speculative premium into contracted, financed NOI and validate the multiple.

Verdict: the variant perception is not “is the business real” (it is) but “is the equity worth the pipeline-and-refinancing price after dilution.” The bulls own the demand story and the power scarcity; the bears own the balance sheet, the dilution, and the capital cycle. At ~$26, the price sides decisively with the bulls.

12. Fact vs. Interpretation Table

13. Open Questions

1. Full Form 4 insider read. This memo sampled recent Form 4s (showing Prager code-S sales and option exercises) but did not quantify the complete 2025–2026 buy/sell record across all 314 filings. A definitive P-vs-S tally should be completed before any position decision.
2. The marketed “$X billion backlog.” WULF cites a large contracted-revenue backlog in investor decks; the 10-K discloses only $307.8M of GAAP lessor minimum receipts (Core42-commenced only). The reconciliation between the deck figure and the GAAP-recognizable amount — and the assumptions inside it — is unresolved.
3. Kentucky (“Justified”) anchor. Will the “highly confident” Q2 2026 IG-hyperscaler signing actually close, and on what economics (NOI yield, term, credit structure)?
4. Remaining capex and funding path. Exact total remaining capex to complete all contracted projects, the funding sources, and the dilution implied — plus the feasibility/timing of refinancing the 7.75% notes into IG debt.
5. Fully-diluted share count. The true fully-diluted count (basic + Google warrants + converts net of capped calls + RSUs/PSUs) and the resulting per-share economics under each scenario.
6. Core42/G42 and Fluidstack credit. The standalone creditworthiness of the tenants behind (and beyond) the Google backstop, and the precise mechanics/timing of the backstop’s effectiveness.
7. Morgantown & interconnect timing. FERC ruling (mid-summer 2026) and the NYISO 250 MW interconnect decision (mid-2026) — both externally gated and thesis-relevant.

14. What Must Be True

Bull case — what must be true:

• The contracted 606 MW energizes on schedule (CB-3 Q2’26, CB-4 Q3’26, CB-5 Q4’26, Abernathy 2H’26) at ~85% stabilized margins.
• The Kentucky anchor signs (IG hyperscaler) and the 250 MW Lake Mariner interconnect is awarded, converting “pipeline” into contracted backlog.
• The pipeline continues converting at the guided 250–500 MW/year, and ~$3B of remaining capex is funded without value-destructive equity issuance.
• The 7.75% project notes are refinanced into cheaper IG debt, unlocking equity free cash flow within “a couple of years.”
• The capital cycle does not compress lease rates/NOI yields faster than WULF scales.

Falsification test (bull): A single material counterparty event (Fluidstack stress), a multi-quarter energization slip on CB-4/CB-5 or Abernathy, a failure to sign the Kentucky anchor by mid-2026, or a dilutive equity raise at a depressed price — any one falsifies the “clean, funded, scaling platform” thesis.

Bear case — what must be true:

• EV/contracted-MW at ~2x operating-asset prices proves to be paying for a pipeline that converts slowly, partially, or dilutively.
• Dilution (warrants + converts + SBC) and the senior debt stack absorb most of the asset-level value creation, leaving mediocre-to-poor per-share returns even if the projects perform.
• The capital cycle compresses lease rates/NOI yields as the cohort and REITs flood supply.
• A funding-meets-execution stumble forces a re-rate from “platform” to “leveraged contractor.”

Falsification test (bear): A signed investment-grade hyperscaler lease at Kentucky plus the 250 MW interconnect award plus an IG refinancing of the 7.75% notes — delivered without a dilutive equity raise — would convert the speculative premium into contracted, financed NOI and falsify the “priced for a pipeline that won’t fund cleanly” thesis.

15. Source Appendix

See the Source Appendix below for the full citation list. Primary sources relied upon:

• TeraWulf Inc. FY2025 Form 10-K (filed 2026-02-27), Notes 8 (lessor/leases & Google warrants), 10 (debt), 13 (commitments), 17 (segments & SBC).
• TeraWulf Inc. Form 10-Q Q1 2026; Q1 2026 earnings call transcript (May 8, 2026); Q4 2025 call (Feb 26, 2026).
• TeraWulf Inc. DEF 14A (filed 2026-04-28) — executive compensation, related-party transactions, beneficial ownership.
• SEC EDGAR XBRL company facts (CIK 0001083301) — revenue, net income, operating income, assets, equity, cash, debt, capex, SBC, OCF, shares outstanding.
• Form 4 filings (sampled) and Form 144 notices (insider activity).
• Peer market data (yfinance via internal tooling, 2026-06-12): WULF, CIFR, CORZ, IREN, APLD, MARA, RIOT, BTDR, CLSK, HUT.
• Prior the author research: IREN (2026-06-10), APLD (2026-06-12) — peer scaffolding and valuation method.

This is an independent analyst’s article for general information only. The body of this piece takes no position and carries no price target; the sole exception is the labeled “Author’s Take” block at the top, which is the author’s own subjective opinion and not investment advice.

APPENDIX A — Standard Diligence Questionnaire

TeraWulf Inc. (NASDAQ: WULF) · As of June 12, 2026 · Independent research

Answers are labeled [F] Fact / [I] Interpretation / [A] Assumption where it matters.

General

What thoughtful questions have other investors asked about this company? The recurring institutional questions (from the Q1 2026 and prior earnings calls): (1) Will the Kentucky/“Justified” anchor customer sign, and at what credit/economics? (2) How do contract terms compare to a year ago — have hurdle rates/yields held as the market crowds? (3) The cadence of the Bitcoin-mining wind-down (hash rate) vs. HPC ramp; (4) the path to refinancing the 7.75% project notes into investment-grade debt to “unlock” equity FCF; (5) new-site additions and the utility-partnership model; (6) NIMBY/political risk to the pipeline. [F/I] Notably, the buy side has focused on backlog and pipeline, while the short side (~26% of float) focuses on dilution, negative equity, and the gap between contracted MW and the price.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? [F/I] Neither — they are at a transitional trough that is also a GAAP-loss peak. Revenue is flat-to-down (Q1’26 $34M ≈ Q1’25 $34.4M) as mining shrinks faster than HPC ramps; GAAP losses are at an artificial peak driven by the non-cash Google-warrant markup (rising as the stock rises). The underlying earnings power is pre-inflection: the contracted 606 MW has barely begun to energize.

Driven by the external environment or internal actions? [I] Both. The demand tailwind (AI/power scarcity) is external; the pivot, contracting, and capital raising are internal/deliberate. Mining cash flow is partly external (BTC price, demand-response economics).

How stable are revenues? [F/I] Currently unstable (mining/BTC-linked, demand-response-dependent), transitioning to highly stable (10–25 year contractual leases) as HPC commences. The stability is contracted but not yet recognized.

Outlook for products/services? [F] Strong contracted demand; management guides to 250–500 MW/year of new contracted capacity. The product (power-secured HPC capacity) is in structural shortage.

How big will this market be — growing, shrinking, domestic or international? [F/I] Large and growing — US-domestic data-center capacity for AI, with ~$700B/yr hyperscaler capex chasing it. WULF’s sites are all US (NY, TX, KY, MD).

Business Quality & Competitive Moat

Is the industry getting more or less competitive? [I] More. The converted-miner cohort (CORZ, CIFR, IREN, APLD, MARA, HUT), REITs, hyperscaler self-build, and private capital are all racing to secure power and sign tenants — a mid-to-late-stage capital cycle (Marathon lens).

How profitable is the business (ROIC, ROE)? [F] Currently negative — negative book equity makes ROE meaningless; FCF is deeply negative mid-build. [I] Project-level economics are attractive (~85% lease margin, ~12–15% NOI yield on cost), but corporate returns are pre-inflection and burdened by SBC and interest.

How profitable is the industry — how many competitors, barriers to entry? [I] Barriers are real but replicable: securing power/interconnection takes capital and time (a head start), but is not a franchise. The binding scarcity (power) is the barrier; it favors whoever moves first at a given site.

Can the business be easily understood? [I] The model yes (build powered shells, lease to AI tenants). The financials no — penny-warrant liabilities, convert derivatives, project-finance ring-fencing, related-party deals, and SBC distortions make the GAAP statements hard to read without normalization.

Can it be undermined by foreign low-cost labor? [F] No — it is domestic power-and-real-estate infrastructure, not labor-arbitrageable.

Do brands matter? [I] Not consumer brands; counterparty credibility and execution track record matter (the “you’re only as good as your last job” point). Reputation with hyperscalers and utilities is the relevant intangible.

Nature of competition? Customers’ switching costs? [F/I] Competition is deal-by-deal for power-secured sites. Switching costs are high post-signing (10–25 yr leases, co-designed buildings) but zero pre-signing.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? [I] The uncontracted 1.75 GW development pipeline and the option value of power positions are not on the balance sheet — they are the bull’s “hidden value.” Conversely, the contracted-but-uncommenced Fluidstack/Abernathy leases are not yet revenue.

Off-balance-sheet liabilities? [F/I] The Abernathy JV (50.1%, equity-method) carries ~$0.9B of remaining capex partly off WULF’s consolidated balance sheet; the Beowulf employee-trust funding (2% of annual capex) is an ongoing obligation; capped-call and warrant structures are complex. Operating-lease/ground-lease obligations (Cayuga 80-yr) are largely capitalized.

How conservative is the accounting? [I] Mixed. Early adoption of ASU 2023-08 (BTC fair value) is current-standard. The “adjusted EBITDA” add-back of $100M/quarter of real SBC is aggressive. The warrant-liability marks make GAAP net income unusable without normalization. A critical audit matter was flagged on the related-party Beowulf acquisition.

How CapEx-hungry is the business? [F] Extremely — $1.06B FY2025, $523M in Q1’26 alone, ~$3B remaining. This is among the most capital-intensive business models in the public markets right now.

Capital Allocation & Management

How much FCF does the business generate; how does management use it; philosophy? [F] None currently — FCF is deeply negative. Philosophy: raise aggressively (converts, secured notes, equity, penny warrants), deploy into contracted/commercializing assets, “contract first, deploy capital second.” [I] Disciplined on what it builds (no speculative builds), undisciplined on per-share cost (heavy dilution).

Significant acquisitions recently? [F] Yes — the related-party Beowulf E&D internalization (~$53M, May 2025) and the Abernathy JV formation (Oct 2025). Sold the Nautilus 25% stake (Oct 2024, $102M, +$22.6M gain).

Buying back shares? [F] No — $200M authorization (Oct 2024) sits unused in 2H’25; ~$151.5M treasury bought FY24/early-25 then idled as capital pivoted to capex.

Issuing large amounts of new shares to insiders? [F] Effectively yes via SBC ($101M in Q1’26 alone) and price-hurdle PSUs (all $6.00/$6.50 hurdles hit Sept 2025). Plus ~14% of the company to Google in penny warrants.

Compensation policy of directors/management? [F] CEO Prager $39.4M FY2025 (incl. $15M discretionary cash bonus, $23.2M stock); other NEOs $20–30M. [I] Keyed to stock-price hurdles and capital-formation/deal-making, not ROIC/FCF/per-share value — misaligned for a company whose chief risk is dilution.

Motivations of management? [I] Build a large, IG-rated AI-infrastructure platform (the stated Cheniere analogy). Founder-led (Prager 8.5%, power-developer pedigree). The related-party dealings and momentum-based comp suggest founder-economics are well-protected; minority-shareholder per-share value is a secondary consideration.

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer? [F] No — a US-domestic C-corp (Nasdaq: WULF), standard 1099 reporting. Legacy CIK from the 2021 IKONICS reverse merger (causes third-party-feed mislabeling as “Financials”).

Dividend policy? [F] None and none expected — capital is fully directed to the buildout.

How profitable is the business? [F] Unprofitable on GAAP and FCF; project-level NOI economics attractive but pre-stabilization.

Is net income diverging from cash from operations? [F] Massively — FY2025 net loss −$661M vs. OCF −$123M; the ~$540M gap is dominated by the non-cash warrant/derivative marks and SBC. [I] This is the single most important normalization: GAAP net loss vastly overstates the cash bleed, but understates per-share dilution.

Risks & Downside

What factors would cause the stock to decline? [I] Failure to sign the Kentucky anchor; a CB-4/CB-5 or Abernathy energization slip; a Fluidstack/neocloud counterparty event; a denied interconnect or FERC delay; a dilutive equity raise; capital-cycle lease-rate compression; or a capital-markets window closing mid-build. The ~26% short interest means positive surprises can also squeeze it sharply.

Risk of a catastrophic loss? [I] Low-probability but real: a negative-equity, FCF-negative, levered developer could see the common heavily impaired in a counterparty-failure-plus-funding-freeze scenario, with $4.6B of debt ahead of it. Asset coverage limits a literal zero, but the levered common is the risk locus.

Chance of a total loss? [I/A] Low in the base case (real assets, contracted leases, Google backstop, $3.1B cash), but non-trivial in a severe-stress tail given the leverage and negative equity. Not a “safe” balance sheet.

Recent News & Events

Has the business environment changed recently? [F] Yes, transformationally over 18 months: Core42 (Dec’24) and Fluidstack/Google (Aug’25) leases, $3.2B secured notes + $2.5B converts (2H’25), Abernathy JV (Oct’25), Hawesville/KY and Morgantown/MD added (2026), ~$1.2B equity raised YTD 2026, $250M bank revolver. The AZI news feed returned no scored items for WULF (a known feed gap, not a tell); the event timeline was built from 8-Ks and transcripts.

Significant acquisitions? [F] Beowulf E&D (related party, May’25); Hawesville site purchase agreement (Feb’26); Nautilus stake sold (Oct’24).

Change in accounting policies? [F] Early adoption of ASU 2023-08 (BTC fair value) effective Jan 1, 2024.

Recent changes — new markets, facilities, management? [F] New facilities/markets: Abernathy (TX), Hawesville (KY), Morgantown (MD), Cayuga (NY) added to the Lake Mariner (NY) base. Internalized Beowulf management team. No major C-suite turnover (Prager/Khan/Fleury stable).

APPENDIX B — Source Appendix

TeraWulf Inc. (NASDAQ: WULF) · Research as of June 12, 2026

Primary sources (SEC filings, company disclosures, regulatory data) are prioritized over secondary. Every non-obvious fact in the memo traces to one of the entries below.

1. SEC Filings — TeraWulf Inc. (CIK 0001083301)

2. Earnings Call & Event Transcripts

3. Peer / Market Data

4. Analytical Frameworks

Data notes: All financial series were taken from SEC EDGAR XBRL and the 10-K/10-Q directly rather than third-party aggregators, whose three-statement data is unreliable for this recently-restated, heavily-financed filer. Common aggregator metadata (a “Financials / Capital Markets” GICS tag, a 1996 IPO date, a 2004 split) are legacy-CIK artifacts from the 2021 IKONICS reverse merger and were disregarded.