From bitcoin to big data: miners’ $8b AI pivot redraws tech’s power map

In late 2025, major Bitcoin miners like TeraWulf and BitFarms initiated bold strategic pivots away from traditional cryptocurrency mining toward AI and high-performance computing (HPC) infrastructure, underpinned by record-breaking financing deals. TeraWulf secured a landmark $3.2 billion senior secured note at 7.75% interest, backed by infrastructure assets rather than mining equipment, while BitFarms raised $588 million through a convertible note with a low 1.375% coupon, leveraging its Pennsylvania coal-fired power plants for vertical integration. These moves reflect a broader industry trend amid rising sector debt exceeding $20 billion, as miners seek to reposition themselves as AI data center powerhouses rather than mere Bitcoin miners.

By early 2026, BitFarms had fully embraced its transformation, rebranding as Keel Infrastructure and relocating its headquarters from Canada to the United States to better align with its new focus on HPC and AI infrastructure. The company decisively exited the crypto mining business, committing 100% of its investments to AI and HPC while planning a gradual wind-down of existing Bitcoin mining operations throughout the year. This strategic clarity underscores the miners’ commitment to shedding legacy crypto assets in favor of AI-driven growth opportunities.

The momentum continued into mid-2026, with miners like Hut 8 raising a record $4.25 billion in triple-B rated senior secured notes to finance a massive 352 MW AI data center in Texas. This sophisticated financing structure, featuring non-recourse terms and interest-only payments until 2030, not only mitigates risk for Hut 8’s parent company but also signals growing market confidence in the sector’s AI pivot. The strong demand for investment-grade bonds and decreasing cost of capital highlight investors’ increasing trust in Bitcoin miners’ evolution into AI infrastructure leaders.

TeraWulf spearheaded the operational pivot from bitcoin mining to AI and HPC workloads with its Q3 2025 results, recording a landmark $7.2 million in AI revenue and achieving a remarkable 71.9% net operating income margin in its High Power Compute segment, which adjusts to about 82% pro forma excluding development costs. This financial success was underpinned by strategic partnerships with FluidStack and CoreWeave, securing substantial contracts totaling over 450 megawatts at Lake Mariner and additional capacity at Core 42, with ongoing expansions at sites like Cayuga signaling continued investment in AI infrastructure.

Other major bitcoin miners have followed suit, with Core Scientific converting roughly 500 megawatts of crypto mining infrastructure to AI workloads and signing 12-year contracts valued over $10 billion, propelling colocation and HPC revenues from $3.8 million in early 2025 to $31.3 million by year-end. Similarly, IREN transitioned to a vertically integrated AI cloud provider, growing AI revenue fivefold to $17.6 million with an impressive 86% gross margin, while Riot Platforms secured a 10-year, $311 million lease with AMD, marking a shift toward stable, contracted digital real estate income alongside mining operations.

Bitfarm, rebranded as Keel Infrastructure, exemplifies early-stage operational execution by repurposing its energy assets to host Nvidia-powered AI workloads, leveraging coal-fired plants for behind-the-meter power generation to circumvent grid constraints. While Keel has yet to sign tenants for its AI-focused sites, it targets initial AI revenue from expanded capacities at Panther Creek and Sharon sites by the second half of 2027, reflecting a deliberate and capital-intensive transition away from bitcoin mining toward AI and HPC services.

Hut 8’s ambitious AI campus buildout at Riverbend, anchored by a $7 billion, 245-megawatt deal with FluidStack and Anthropic, demonstrates the scale and early operational progress in AI workloads with the first data hall slated for delivery by early 2024. The company is also proactively addressing environmental concerns through investments in transmission upgrades and closed-loop chiller systems, underscoring a commitment to sustainable AI data center growth amid rising community scrutiny.

Wall Street has fundamentally restructured its valuation frameworks for bitcoin miners transitioning into AI infrastructure providers, moving away from traditional discounted cash flow models to sum-of-parts approaches that distinctly value Bitcoin holdings, mining EBITDA, AI co-location revenues, and power-site potential. This evolution reflects a broader investor acceptance of miners selling Bitcoin reserves to finance AI data center conversions, as exemplified by Core Scientific's early 2026 divestiture of most of its Bitcoin holdings—a move once deemed heretical but now viewed as pragmatic for infrastructure growth.

Investor enthusiasm for AI-driven revenue streams is palpable, with miners holding active AI contracts commanding roughly double the price-to-sales multiples of pure play miners—trading at about $6 million per planned megawatt compared to half that for traditional miners. Applied Digital's meteoric rise, including a 282% stock surge over the past year fueled by a $5.2 billion AI factory lease, underscores this shift, as analysts elevate price targets and reframe valuations around scalable AI infrastructure franchises rather than Bitcoin exposure.

Despite bullish narratives, investor sentiment remains nuanced and occasionally cautious due to execution risks, insider selling, and financial leverage concerns. Hut 8's experience illustrates this tension: after a strong rally propelled by $16.8 billion in AI lease contracts and a Benchmark price target hike to $165, the stock faced profit-taking, a 9.7% drop amid $12 million in insider sales, and wariness over persistent net losses and mark-to-market Bitcoin accounting. Similarly, Applied Digital and TeraWulf, though viewed as undervalued relative to their AI contract backlogs, confront scrutiny over customer concentration and debt-funded expansions that could pressure cash flows if demand softens or projects delay.

The market narrative has decisively shifted from Bitcoin price and hash rate metrics toward evaluating miners as hybrid tech infrastructure firms valued on power access, tenant quality, contract duration, and execution capability. This is evident in TeraWulf’s $19 billion, 20-year lease with Anthropic, which reframes its valuation around nearly $950 million in annual contracted AI revenue, and in Hut 8’s expanding AI portfolio now valued at $26.6 billion. While analysts remain bullish—Piper Sandler maintains a Buy on Hut 8 with a $127 target and consensus price targets average $144—the specter of regulatory risks, such as New York’s moratorium on hyperscale data centers, and potential project delays injects caution, underscoring the high stakes of this transformative market evolution.

Bitcoin miners-turned-AI data center operators have inked multi-billion-dollar lease agreements and embarked on massive campus buildouts, underscoring the scale and ambition of their AI infrastructure pivots. Hut 8 exemplifies this with a $7 billion, 15-year lease for 245 MW at its River Bend campus and a subsequent $9.8 billion, 15-year lease that fully commercializes its 1 GW Beacon Point campus, collectively pushing its contracted AI infrastructure portfolio to nearly $27 billion and 949 MW. Similarly, TeraWulf has transformed brownfield mining sites into AI hubs, securing a landmark $19 billion, 20-year lease with Anthropic for its Justified Data campus in Kentucky, alongside joint ventures and equity backstops from Google and Fluidstack, signaling a strategic shift from bitcoin mining metrics to infrastructure valuation focused on power access and contract duration.

Applied Digital and Iris Energy have emerged as major players scaling AI data center capacity through substantial lease agreements and strategic partnerships. Applied Digital's $5.2 billion, 15-year lease for 210 MW at its Delta Forge 2 campus expands its AI Factory franchise to five campuses, with total base-term lease revenue reaching $36 billion and potential renewals pushing that figure to $86 billion, propelling its stock to soar 282% over the past year—far outpacing Nvidia. Meanwhile, Iris Energy leverages a Microsoft-backed scaling plan that has contributed to its $14 billion valuation, highlighting the critical role of strategic alliances in accelerating AI infrastructure growth.

The broader sector's AI infrastructure expansion is marked by a flurry of multi-billion-dollar projects with tight execution timelines, as firms race to meet mid-2026 Ready-For-Service dates critical to fulfilling lease commitments. Terra Wolf exemplifies this urgency, aggressively scaling with multiple campus buildouts including a 480 MW site in Kentucky and a 170 MW joint venture with Fluidstack, driving AI and HPC revenue to 62% of its total and boosting its market cap to $11 billion. This rapid scaling, mirrored by Core Scientific and others, signals a transformative pivot where AI revenue is expected to constitute about 70% of public miners' income by the end of 2026, reshaping their identities from bitcoin miners to AI infrastructure operators.

Hut 8's ambitious $10 billion AI data center construction in West Feliciana, Louisiana, highlights both the promise and risks inherent in this large-scale pivot. Expected to be fully operational by Q3 2027 and supporting hundreds of construction and permanent jobs, the project builds on Hut 8’s prior $7 billion Fluidstack lease and is financed through $4.25 billion in triple-B rated senior secured notes with non-recourse terms, reflecting sophisticated capital deployment and strong investor confidence. However, the scale of this buildout intensifies execution risks around on-time delivery and tenant acquisition, even as Hut 8 projects 76.4% annual revenue growth to reach $1.6 billion by 2029, underscoring the high stakes of the AI infrastructure race.

The transition from Bitcoin mining to AI data center operations is fraught with substantial execution and capital intensity risks, as highlighted by Kevin O’Leary’s estimate that 50% of North American AI data center builds will fail due to operational challenges and the staggering cost of up to $7 billion for a 1.4GW infrastructure supporting 1GW compute. Companies like CoreWeave and Hut 8 exemplify these risks, grappling with multi-billion-dollar CAPEX plans, construction delays, and heavy debt burdens while striving to meet aggressive revenue targets amid a capital-intensive environment that dwarfs traditional mining costs.

Customer concentration emerges as a double-edged sword in this evolving landscape, providing credibility and financial backing through long-term contracts with hyperscalers such as Microsoft, OpenAI, Meta, and Fluidstack, yet simultaneously heightening vulnerability to shifts in demand or contract terms. Bitzero’s 18 customers and CoreWeave’s billion-dollar backlogs illustrate this dynamic, while Applied Digital and Hut 8 face market skepticism and share price volatility fueled by reliance on a handful of large clients, underscoring the precarious balance between securing stable revenue streams and managing concentrated counterparty risk.

Regulatory and community opposition risks are increasingly prominent as AI data center projects scale, with incidents like the failed Granbury, Texas vote to expel Marathon Digital Holdings and local concerns over energy prices, water usage, and environmental impact at Hut 8’s Louisiana and Illinois sites. These challenges compel companies to invest in transmission upgrades and closed-loop cooling systems to mitigate water consumption, while navigating moratoriums on hyperscale data centers in states like New York and municipalities such as Tulsa, reflecting a complex regulatory environment that could materially affect operational costs and project timelines.

The competitive landscape is marked by large-scale, long-duration deals and ambitious site developments that underscore both opportunity and risk in the AI data center sector. Hut 8’s $7 billion Fluidstack lease and $10 billion West Feliciana project, alongside Applied Digital’s $7 billion CoreWeave leases and multiple gigawatt-scale sites across the U.S., highlight a capital-intensive race to scale AI infrastructure. However, these massive commitments amplify execution risk, financial strain, and market volatility, as investors weigh lofty valuations—often with price-to-sales ratios far exceeding industry norms—against the uncertainties of meeting construction deadlines, securing tenants, and managing evolving regulatory pressures.

Bitcoin miners are fundamentally reshaping their business models by evolving into hybrid AI-crypto infrastructure providers, a transition that decouples their valuation and cash flow stability from the volatile crypto price cycle. This shift is exemplified by Bernstein's November 2025 adoption of a sum-of-parts valuation model that separately values Bitcoin holdings, mining EBITDA, AI co-location revenue, and power-site potential, reflecting the growing importance of stable AI cash flows. Companies like Core Scientific and Marathon Digital Holdings have pragmatically adjusted treasury strategies, selling Bitcoin reserves to fund AI infrastructure buildouts, signaling a departure from the traditional 'HODL at all costs' mentality and underscoring the strategic prioritization of AI-related revenue streams.

The pivot to AI data center hosting introduces a new economic paradigm for Bitcoin miners, transforming them from speculative, price-sensitive crypto operators into utility-like providers of contracted high-performance computing services. This transition leverages miners’ locked-in, long-term renewable energy contracts and large-scale infrastructure, such as Riot Platforms’ Rockdale facility, enabling operating margins exceeding 70%–80% compared to 55%–65% for optimized mining. However, this shift also replaces one cyclical risk—crypto price volatility—with another tied to AI market sentiment, as over 14 GW of miner-controlled power capacity migrates to AI hosting, exposing companies to potential drawdowns if AI demand falters.

Leading miners like TeraWulf and Hut 8 illustrate the long-term potential and challenges of this hybrid model by securing multi-billion-dollar, multi-decade AI contracts that significantly enhance revenue visibility and investor confidence. TeraWulf’s 20-year lease with Anthropic PBC for approximately 401 MW of HPC capacity and Hut 8’s $9.8 billion, 15-year lease at its 1 GW Beacon Point campus exemplify how these companies are locking in stable, contracted AI revenue streams that could comprise up to 70% of total revenue by late 2026. Nonetheless, this capital-intensive pivot demands careful management of debt burdens, commissioning timelines, and tenant performance, as evidenced by concerns that capital spending might outpace cash generation or that AI tenants could underdeliver on lease commitments.

While the hybrid AI-crypto infrastructure approach offers a promising path to stabilize and diversify revenue, it also introduces significant operational and strategic challenges that will shape the industry's future. Companies like Keel are leveraging on-site coal-fired power plants to mitigate grid interconnection issues and power cost volatility, highlighting innovative solutions to support AI data center operations. However, the transition requires new talent, supply chains, and cooling technologies distinct from traditional ASIC mining, and the long-term viability hinges on sustained enterprise AI adoption and compute investment. Moreover, as miners divert megawatts from Bitcoin mining to AI workloads, the industry faces a potential rebalancing of Bitcoin hashrate concentration and geographic mining profiles, underscoring the broader implications for crypto network security and the global AI infrastructure race.