Bitfarms Ltd. (BITF)

Executive Summary / Key Takeaways

• Bitfarms is executing a strategic metamorphosis from Bitcoin miner to North American HPC/AI infrastructure provider, transforming its 1.4 gigawatt power portfolio from a cost center into the primary value driver.
• The simultaneous Stronghold acquisition and Yguazu divestiture in early 2025 rebalanced the company toward U.S. power generation assets with immediate HPC conversion potential, creating a multi-year development pipeline in Pennsylvania.
• Bitcoin mining operations generate approximately $8 million in monthly free cash flow, funding the HPC buildout without equity dilution, while the Bitcoin 2.1 program monetizes treasury through derivatives to further support infrastructure investment.
• Management is leapfrogging current GPU architecture to build Vera Rubin-ready data centers, targeting 2027 lease rates of $150+ per kilowatt-month versus $120 for legacy infrastructure, capturing the supply-demand imbalance in next-generation compute.
• Execution risk on complex site conversions and construction timelines represents the critical variable: delays in energizing Panther Creek or Washington facilities could compress the valuation window before mining economics correct in 2026.

Setting the Scene: The Bottleneck Is Power, Not Silicon

Founded in 2017 in Toronto, Bitfarms spent its first six years executing a straightforward Bitcoin mining strategy: secure cheap power, deploy ASICs , accumulate Bitcoin. This model worked until the 2024 halving compressed margins and revealed a structural truth that would redefine the company’s trajectory. While the market obsessed over hashrate and Bitcoin holdings, Bitfarms recognized that its real assets—the power purchase agreements, generation facilities, and strategically located energy infrastructure—were becoming exponentially more valuable in an AI-driven world. The investment thesis hinges on a simple but powerful observation: compute demand is surging at 8.8% annually, GPU shipments will reach 100 gigawatts by 2030, but data center infrastructure cannot keep pace, creating a 45 gigawatt power shortfall. In this new paradigm, owners of power assets capture the economics, not just miners of cryptocurrency.

Bitfarms operates at the intersection of energy and compute, a position it deliberately engineered through two transformative transactions in Q1 2025. The acquisition of Stronghold Digital Mining (SDIG) brought three Pennsylvania power generation campuses with 665 megawatts of capacity and immediate high-performance computing (HPC) development opportunities. Concurrently, the sale of the 200-megawatt Yguazu site in Paraguay—purpose-built for Bitcoin mining with no HPC conversion potential—allowed the company to avoid $100+ million in capital expenditures while redeploying capital toward U.S. growth. This portfolio rebalancing shifted the asset base from 50% U.S. power to nearly 80%, reduced average power costs to $0.043 per kilowatt-hour, and created a development pipeline where every megawatt could eventually service AI workloads at 3-4x the revenue intensity of Bitcoin mining.

The company’s place in the value chain has fundamentally changed. Bitfarms is no longer competing solely in the commodity business of Bitcoin block rewards. Instead, it is building the physical infrastructure layer that hyperscalers and AI companies desperately need but cannot quickly source. Pennsylvania’s emergence as an AI hub—attracting over $90 billion in investments from Google (GOOGL), Meta (META), Blackstone (BX), and CoreWeave—positions Bitfarms’ Panther Creek campus at the center of a geographic cluster with robust fiber infrastructure (eight metro networks) and a cooler climate that enables industry-leading Power Usage Effectiveness (PuE) of 1.2-1.3 versus 1.4-1.5 in Texas. This translates directly into more revenue-generating compute per megawatt, a structural cost advantage that matters deeply when infrastructure owners are capturing the economic surplus from GPU operators earning 80-90% direct margins.

History with Purpose: The Strategic Architecture of a Pivot

Bitfarms’ transformation began in Q4 2023 with the Synthetic HODL program, a pilot initiative that sold Bitcoin to fund operations while preserving upside through call options. This evolved into Bitcoin One in Q1 2025 and subsequently Bitcoin 2.1, a quantitative multi-strategy program that CFO Jonathan Mir explicitly states is “not operating as a Bitcoin treasury company.” The goal is not accumulation but optimization: offsetting production costs and achieving higher value per Bitcoin sold to fund energy infrastructure investments. This demonstrates capital discipline at a time when many miners were hoarding Bitcoin at market peaks, and it provides a low-cost, low-risk funding mechanism that avoids the dilution of equity raises or the restrictive covenants of traditional debt.

The Stronghold acquisition and Yguazu sale represent the largest M&A deal between two public miners in industry history, but their significance extends beyond headlines. Stronghold’s power generation facilities in PJM Interconnection give Bitfarms something no pure-play miner possesses: control over energy pricing through generation, curtailment, and demand response programs. Ben Gagnon notes these assets are “worth significantly more” due to surging power demand and record capacity auctions, providing downside protection and optimization for free cash flow. This control transforms energy from a variable cost into a strategic asset that can be monetized multiple ways—selling power to the grid at peak prices, mining Bitcoin when prices are low, or leasing infrastructure to AI customers at premium rates.

The decision to shut down Argentina operations by November 2025, expected to yield $18 million in proceeds, further de-risks the portfolio. While the closure removes approximately 1 exahash of hashrate, it eliminates site remediation liabilities, recovers prepaid deposits, and reduces lease expenses—cash that can be redeployed to Pennsylvania. The move also improves fleet-wide KPIs: energy efficiency up 1%, average electricity price improved 2%, direct hash costs improved 5%, and uptime up 2%. This demonstrates that pruning underperforming assets enhances the quality of remaining operations, a level of portfolio management “unheard of in our industry,” according to Gagnon.

Technology and Strategic Differentiation: Building for Vera Rubin

Bitfarms’ HPC strategy rests on a contrarian but strategically sound decision: leapfrogging NVIDIA’s (NVDA) Blackwell architecture to develop infrastructure for next-generation Vera Rubin GPUs expected in Q4 2026. Vera Rubin infrastructure will be incompatible with facilities designed for current-generation hardware, creating a supply-demand imbalance where next-gen data centers command significantly greater economics. While competitors retrofit for 190 kilowatt-per-rack GB300s, Bitfarms is building for 370 kilowatt-per-rack VR200 systems, requiring different cooling and electrical distribution including potential 800-volt DC systems. The company is targeting 99% of its 2026-2027 development portfolio for Vera Rubin, positioning it to capture peak demand when others face obsolescence.

The Washington site conversion exemplifies this approach. The 18-megawatt Moses Lake facility, located in the largest West Coast data center cluster with sub-$0.03/kWh renewable power and a 10-year waitlist for new capacity, is being transformed into GPU-as-a-service infrastructure targeting PuE of 1.2-1.3. A fully funded $128 million binding agreement for critical IT infrastructure and building materials is already in place, targeting completion in December 2026. Gagnon states this single site “could be worth significantly more than the entire Bitcoin mining business that the company has been operating for multiple years” under a cloud monetization model. This implies that Bitfarms’ smallest site could generate more net operating income than its historical mining operations, demonstrating the transformative economics of the pivot.

Panther Creek serves as the flagship campus with 350 megawatts secured (50 MW by year-end 2026, 300 MW by 2027) and potential expansion beyond 500 MW. The recent decision to modify development for Vera Rubin GPUs marginally delays Phase 1 energization from December 2026 to first-half 2027, but management believes this will enable “significantly greater economics.” The partnership with T5, a top-tier data center developer, validates the development potential to investors and prospective customers. The $300 million Macquarie facility, converted to project-specific financing with an additional $50 million drawn in October 2025, accelerates development while enhancing financial flexibility by moving debt to the asset level.

Quebec’s 170 megawatts of low-cost hydropower represents a unique strategic opportunity to increase the province’s total data center megawatts by 25% from approximately 700 MW today while fulfilling national objectives around data sovereignty and scaling back crypto mining. The 96-megawatt Sherbrooke site is targeted for Vera Rubin conversion in 2027-2028, with the remaining 74 megawatts earmarked for potential expansion. This provides a pathway to convert legacy Bitcoin assets into higher-value HPC capacity with regulatory support, a conversion that is unprecedented but politically supported at multiple levels.

Financial Performance: Mining as a Utility for Transformation

Bitfarms’ Q3 2025 results from continuing operations provide clear evidence that the strategy is working. Revenue of $69 million represented 156% year-over-year growth, while adjusted EBITDA of $20 million (28% margin) increased from $2 million (8% margin) in Q3 2024 and $9 million (15% margin) in Q2 2025. The all-in cost per Bitcoin mined was $82,400, but after a $13.3 million net gain from derivatives, the effective cost fell to $55,200—demonstrating how Bitcoin 2.1 enhances cash generation.

Gross mining profit of $21 million at a 35% margin provides the $8 million monthly free cash flow that management explicitly states will fund HPC/AI development projects.

The segment performance tells a crucial story. Bitcoin mining is explicitly managed as a “low-risk cash flow foundation with minimal CapEx needs,” not a growth engine. The company completed all Bitcoin mining growth plans in Q2 2025 and consistently states “no plans for large miner purchases in 2025 or 2026.” This focuses capital allocation on higher-return HPC infrastructure where Bitfarms can capture 12% annual lease rate growth versus 3% historical data center rates. The mining segment’s purpose is to service G&A and debt for Panther Creek while funding conversion capex, a disciplined approach that avoids the growth-at-all-costs trap that plagued competitors.

Competitive context reveals Bitfarms’ strategic clarity. Marathon Digital (MARA) operates 60.4 EH/s versus Bitfarms’ 12.3 EH/s, generating $252 million in Q3 revenue but with higher energy costs and no clear HPC conversion pathway. CleanSpark (CLSK) runs 46.6 EH/s with $198.6 million revenue, emphasizing rapid ASIC deployment rather than infrastructure pivot. Riot Platforms (RIOT) operates 33.2 EH/s from Texas, where 1.4-1.5 PuE and grid volatility create structural cost disadvantages. Core Scientific (CORZ), the only peer with a meaningful AI pivot, generated $15 million in AI revenue (+45% YoY) but remains burdened by mining declines and a $146.7 million net loss. Bitfarms trades at 2x adjusted EBITDA versus 3-5x for these peers, reflecting market failure to price the HPC optionality.

The balance sheet transformation underpins the entire thesis. The upsized $588 million convertible note offering in Q3 2025, structured with cash-settled capped calls to offset dilution up to $11.88 per share, demonstrates sophisticated capital markets execution. Combined with over $1 billion in liquidity ($820 million cash and Bitcoin plus $200 million available from Macquarie), Bitfarms possesses what Gagnon calls “the strongest balance sheet and most available capital in the company's history.” This eliminates the dilution risk that plagues capital-intensive pivots, allowing management to fund multi-year development from a position of strength.

Outlook and Execution Risk: The 2027 Inflection

Management’s guidance reveals a methodical, milestone-driven approach to the HPC transition. Panther Creek Phase 1 energization is marginally delayed to first-half 2027 to accommodate Vera Rubin specifications, while the Washington site targets December 2026 completion. The Sharon site’s full 110-megawatt substation is expected online by year-end 2026, with revenue beginning in first-half 2027. These timelines align with NVIDIA’s Q4 2026 Vera Rubin shipment schedule, positioning Bitfarms to capture first-mover economics before competitors can react. The company’s exahash is expected to “stay relatively consistent in Q4” for continuing operations, then decline gradually as sites convert—Washington potentially impacting 1 exahash by mid-2026, followed by other facilities. This orderly transformation ensures mining cash flow continues funding development until HPC revenue materializes.

The 2026 mining economics correction that Gagnon anticipates—driven by network hashrate growth and a Bitcoin price correction from late-2025 all-time highs—actually supports the thesis. It creates a natural exit ramp from mining just as HPC revenues begin, avoiding a painful cash flow cliff. The company’s guidance that “direct mining margins should range between 50% and 75%” if Bitcoin stays above $100,000 provides a clear marker: mining remains profitable enough to fund the transition but not so attractive that it deters conversion.

Critical execution factors center on construction management and regulatory approval. Gagnon acknowledges that “potential bottlenecks in construction are a little hard to forecast,” but mitigates risk through partnerships with owner’s reps, general contractors, and internal project managers tracking critical paths. The Vera Rubin infrastructure challenge—where “even NVIDIA hasn't completed their validated reference designs”—requires building for technology that doesn’t yet fully exist. This is mitigated by spreading facilities across multiple years and “building to the technology that's coming, not to the technology that already exists.” The Quebec conversion’s regulatory pathway, while “unprecedented,” has “broad political support” and “initial indications of support at various levels,” suggesting manageable political risk.

Risks and Asymmetries: What Can Break the Thesis

Three material risks threaten the investment case. First, the capital intensity of HPC/AI—$30-40 million per megawatt versus $1 million for Bitcoin mining in 2024—creates execution risk. The $400 million envisioned for full Panther Creek buildout through 2026, while funded by the $1 billion+ war chest, represents a massive bet on a business model where Bitfarms has limited operating history. If construction costs exceed estimates or lease-up is slower than projected, returns could compress significantly.

Second, technology obsolescence risk is acute. The rapid evolution from 150kW/rack GB200s to 190kW GB300s to 370kW+ Vera Rubin systems means infrastructure built today could be obsolete by completion. Bitfarms mitigates this by targeting Vera Rubin for 99% of 2026-2027 development, but if NVIDIA delays shipments or alternative architectures emerge, the company could be left with expensive, empty data centers while competitors adapt faster.

Third, the mining-to-HPC conversion in Quebec requires regulatory approval for “converting our crypto mining megawatts to traditional data center megawatts,” a process with “no clearly defined pathway.” While politically supported, delays could push conversion beyond 2027-2028, leaving Bitfarms overexposed to the anticipated 2026 mining correction and missing the peak Vera Rubin demand window.

The asymmetry lies in lease rate dynamics and asset scarcity. If Bitfarms executes on timeline, it enters a market where rates have grown from $120/kW-month to $150+ in months, with hyperscalers willing to sign multi-year agreements at premium prices to secure capacity. The Washington site alone could exceed the historical value of the entire mining business under a GPU-as-a-service model, suggesting potential for meaningful re-rating. Conversely, if execution falters, the company retains a profitable mining operation generating $8 million monthly free cash flow and $820 million in liquid assets, providing substantial downside protection versus peers with higher operational leverage.

Valuation Context: Priced for Mining, Positioned for AI

Trading at $3.10 per share, Bitfarms carries a $1.85 billion market capitalization and $1.84 billion enterprise value. The TTM financial ratios reveal a company in transition: -46.38% profit margin, -28.77% operating margin, and -480.42 million in free cash flow reflect one-time transformation costs, Argentina impairments, and heavy HPC development capex. These backward-looking metrics are less relevant than the Q3 2025 adjusted EBITDA margin of 28% and the forward-looking asset base.

On a revenue multiple basis, Bitfarms trades at 6.86x TTM sales versus Marathon at 4.57x, Riot at 8.73x, CleanSpark at 5.69x, and Core Scientific at 15.01x. The premium to pure-play miners reflects the HPC optionality, while the discount to Core Scientific reflects CORZ’s earlier AI revenue recognition. More meaningful is the enterprise value per strategic megawatt: at 461 MW energized today growing to 1.4 GW by 2028, Bitfarms trades at approximately $4,000 per kilowatt of future capacity, significantly below the $30,000-40,000 per kilowatt replacement cost for HPC infrastructure.

The balance sheet strength—3.20 current ratio, 0.12 debt-to-equity, $820 million cash and Bitcoin—provides a valuation floor that pure-mining peers lack.

Marathon’s 0.70 debt-to-equity and Riot’s 0.25 leverage expose them to margin compression in a mining downturn, while Bitfarms can fund its pivot from internal resources. The forward P/E of 155.00 reflects expectations of profitability recovery as HPC revenues replace mining cash flows, but this metric is less relevant than the path to contracted, multi-year lease revenue that commands 20-30x EBITDA multiples for data center REITs.

Conclusion: The Infrastructure Premium Is Earned, Not Given

Bitfarms has executed a strategic pivot that redefines its earnings power from commodity Bitcoin mining to scarce AI infrastructure. The simultaneous acquisition of Stronghold’s generation assets and divestiture of non-strategic Latin American mining created a U.S.-centric power portfolio where every megawatt can be monetized at 3-4x mining economics through HPC colocation. The $8 million monthly mining cash flow and $1 billion+ liquidity provide a self-funding mechanism that eliminates dilution risk while management builds toward 2027 Vera Rubin capacity.

The investment thesis succeeds or fails on execution against construction milestones and lease-up velocity. If Panther Creek energizes in first-half 2027 and secures $150+/kW-month leases for Vera Rubin infrastructure, the company will have transformed 350 MW into annual revenue potential exceeding $600 million—nearly matching its entire current market capitalization. If construction bottlenecks or regulatory delays push timelines beyond the 2026 mining correction, the company still retains a profitable mining business and ample liquidity to navigate the transition.

The market’s 2x adjusted EBITDA multiple reflects a failure to price the HPC optionality embedded in 1.4 gigawatts of strategically located power. Competitors trade at higher multiples but lack Bitfarms’ portfolio quality and conversion pathway. The critical variables to monitor are construction progress at Panther Creek, regulatory approvals in Quebec, and lease rate negotiations for Vera Rubin capacity. If management delivers on these fronts, the valuation gap between Bitcoin miner and AI infrastructure owner will close, creating substantial upside for investors who recognize that power, not hashrate, is the true scarce resource in the AI era.