TERAWULF INC.

TeraWulf is a US-based infrastructure company that owns and operates power-heavy computing campuses. It built its foundation on bitcoin mining but is actively pivoting toward high-performance computing (HPC) data center hosting for AI and machine learning workloads. It makes money by selling mined bitcoin and, increasingly, by leasing critical IT capacity to large technology tenants under long-term contracts.

Revenue for 2025 was $168.5M, made up of $151.6M in digital asset (bitcoin mining) revenue and $16.9M in new HPC lease revenue S3S4S5. That compares to roughly $140.1M in mining revenue and zero HPC revenue in 2024 S3S5, so total revenue grew about 20% year over year.

Gross profit was $66.0M against cost of revenue of $82.7M (up 32% from $62.6M in 2024) S4, giving a gross margin of about 39%. The cost increase was driven by higher power prices, partially offset by $17.7M in demand response credits (up from $8.6M in 2024) S5.

Operating loss was $186.2M [S4, XBRL]. Net loss ballooned to $661.4M, compared to $72.4M in 2024 S4. The vast majority of that deterioration is non-cash: a $429.8M charge related to Google Warrant liabilities and convertible note derivative revaluations, not operating deterioration per se S4. Interest expense alone jumped $60.4M to $80.2M in 2025 S4.

Operating cash flow was negative $123.2M in 2025 versus negative $24.4M in 2024 S4S5. Capital expenditures were $1.06B [XBRL], with $792M going to HPC infrastructure at Lake Mariner and $450M invested into the Abernathy joint venture S4S5.

Cash on hand looks large at $3.27B [XBRL], but almost all of it came from debt issuance: the company raised $5.1B in new debt proceeds in 2025 S4, including $3.1B in Senior Secured Notes at 7.75% due 2030 S7 and $1.97B in convertible notes S5. Long-term debt is $3.10B [XBRL], but total principal across all maturities runs to $5.725B S7.

Stock-based compensation was $50.9M in 2025, up from $30.9M in 2024 S4. The company repurchased $33.3M in shares [XBRL,S4] and paid $2.0M in dividends [XBRL], which is modest relative to the capital structure. The accumulated deficit stands at $993.7M, up from $332.3M S4.

Business segments: TeraWulf switched from a single segment to two segments in Q3 2025 S6. The Digital Asset Mining segment produced $151.6M in revenue and remains the larger segment for now S5. The HPC Leasing segment generated $16.9M in its first partial year S3S5. HPC Leasing is clearly the faster-growing segment and the one receiving essentially all new capital investment.

The strategic picture is genuinely improving. HPC lease revenue went from zero to $16.9M in less than a year of operations, with 18 MW energized at Lake Mariner as of mid-2025 S4. Contracted capacity has grown to 522 MW across primary campuses S1, including a 168 MW fully pre-leased Texas facility and a Fluidstack deal covering 378 MW at the Akela facility S1. The company received $90.0M in prepaid rent from a single HPC customer in 2025 S7, a concrete signal that tenants are committing real capital. Demand response revenue more than doubled to $17.7M S5, adding a useful margin buffer.

That said, several things are still challenges. Operating cash burn widened to $123.2M in 2025 from $24.4M in 2024 S4S5, reflecting the company still being in heavy build-out mode. The vast majority of revenue (still about 90%) comes from bitcoin mining S3S5, a volatile and cyclical source. Interest expense is now $80.2M annually S4 and rising as more debt was layered on. Most of the contracted HPC capacity does not generate revenue until 2026 deliveries begin S1S3. The path to profitability exists, but it runs through successful on-time delivery of multiple large construction projects, continued bitcoin price support, and effective management of a complex capital structure.

• The filing does not name specific bitcoin mining competitors by name, but management acknowledges operating in a market where mining difficulty and hash rate are determined by global participation, meaning any well-capitalized miner with cheap power is a competitor S2. The company's differentiation in mining was low-cost, high-uptime infrastructure at Lake Mariner, a hydro-powered campus in New York S1.
• In HPC data center hosting, TeraWulf is competing against established operators and hyperscaler-affiliated infrastructure providers for the same large AI compute tenants. Its differentiation is purpose-built, high-density power infrastructure at campuses that are already partially permitted and interconnected, plus the ability to convert former mining load to HPC load quickly S1S3.
• The Google relationship is a genuine competitive moat for the HPC side: Google's credit backstop on Fluidstack lease obligations enabled the $3.2B Senior Secured Notes financing at 7.75% S7, which would likely have been more expensive or unavailable without it. Google also pledged its warrants as collateral for that debt S7.
• A competitive risk the filing explicitly flags is that certain HPC customer agreements may restrict TeraWulf from serving competing tenants, limiting diversification S2.

The filing highlights that the CEO controls the entities that lease land to TeraWulf at both the Lake Mariner campus (157-acre lease with Somerset Operating Company, 35-year term) and the Cayuga site (183-acre lease with Cayuga Operating Company, 80-year term) S3S6. TeraWulf paid $68.8M in stock plus $12M in cash to restructure the Lake Mariner ground lease in 2025 S6, and prepaid $95.0M in stock plus $3.0M in cash for the Cayuga lease S6. These are material related-party transactions that require careful scrutiny by investors.

The Beowulf E&D acquisition in May 2025 for $54.6M S6 was also a related-party deal: the seller was controlled by a Company management member S6. All contingent earnout milestones were triggered and paid in 2025 S6.

RSM US LLP replaced the prior auditor in August 2024 S9, a change worth noting.

Specific insider ownership percentages are not disclosed in the sections provided. The $200M share repurchase program authorized in October 2024 had zero shares repurchased through December 2025 despite being extended indefinitely in May 2025 S3.

• The company is betting that HPC data center hosting for AI workloads will become the dominant revenue source, displacing bitcoin mining. Management is targeting 250 to 500 MW of new critical IT HPC capacity contracted annually S1.
• They are betting that the Abernathy (168 MW) and Akela (378 MW) facilities deliver on time in 2026 and that the Fluidstack/Google tenant relationship holds. Those two facilities represent the bulk of the contracted pipeline S1S3.
• Management is betting that the Cayuga site (up to 400 MW) and newly acquired Hawesville, Kentucky site become the next phase of growth, with the Morgantown generating station acquisition (210 MW operational capacity, closing Q2 2026) providing on-site power generation to reduce grid dependency S3.
• They are betting that the capital structure, centered on $3.2B in Senior Secured Notes at 7.75% S7 and $2.5B in convertible notes S7, can be serviced and refinanced as HPC lease revenues ramp, and that bitcoin prices hold well enough to keep the legacy mining segment cash generative through the transition.

• HIGHThe capital structure is extremely complex and highly leveraged. Total principal across all debt maturities is $5.725B S7, including $489.8M of the 2030 Convertible Notes now classified as short-term S4. Annual interest expense is already $80.2M S4 and will grow as more debt is drawn. Operating cash flow was negative $123.2M in 2025 S4, meaning the company is entirely dependent on raising new capital or generating HPC cash flows to service this debt. A construction delay or tenant default could trigger a liquidity crisis.
• HIGHThe $844.7M warrant liability on the balance sheet relates to 73.58M Google Warrants exercisable at $0.01 per share S7S6. These warrants auto-exercise on a net basis if not exercised and the stock is above $0.01 S7, representing near-certain massive dilution. The fair value swings alone drove $329.2M in losses in 2025 S6, and the warrants expire in August 2030 S7.
• HIGHSignificant related-party transactions: the CEO controls both major land lessor entities, and the company paid a combined $163.8M-plus in stock and cash for the Lake Mariner and Cayuga leases to CEO-controlled entities in 2025 S6. The Beowulf E&D acquisition from a management-controlled entity added $55.5M in goodwill S6 that generated zero revenue post-acquisition S6. These arrangements create genuine conflict-of-interest risk.
• MEDShares outstanding rose 9% to 444.5M in 2025 S4, stock-based compensation was $50.9M against $168.5M revenue (30% of revenue) [S4,XBRL], and the convertible notes (totaling $2.525B in principal) can convert into equity at prices ranging from $8.48 to $19.94 per share S7, implying hundreds of millions of additional shares at conversion.
• MEDRevenue is still roughly 90% dependent on bitcoin mining S3S5, and nearly all of that flows through a single mining pool operator (Foundry USA Pool) S5. Bitcoin miners are purpose-built for SHA-256 and cannot be repurposed S2. The next halving is expected April 2028 S2, which will cut mining rewards.
• LOW$3.649B of bank deposits exceeded FDIC insurance limits as of December 31, 2025 S5, concentrating nearly all liquidity at uninsured institutions.

TeraWulf is a company in mid-transformation, converting a bitcoin mining asset base into a large-scale AI data center platform, with real contracted capacity, a credible Google-backed anchor tenant, and a genuine first-mover position in power-dense HPC infrastructure. But it is doing so with a highly leveraged balance sheet, a complex web of related-party transactions, near-certain dilution from Google Warrants, and an operating model that will not be cash flow positive until multiple large construction projects deliver on time in 2026 and beyond. An investor needs to believe that the construction executes on schedule, that Fluidstack and its other tenants remain creditworthy, that bitcoin prices hold during the transition period, and that management's related-party dealings are priced fairly and governed appropriately.