Executive Summary / Key Takeaways
-
Strategic Inflection from Volatile Mining to Stable AI Infrastructure: Core Scientific is completing a transformation from Bitcoin mining to high-density colocation for AI workloads, with $10+ billion in contracted revenue from CoreWeave providing a foundation of predictable, high-margin cash flows that alters the risk/reward profile from commodity price exposure to infrastructure services.
-
Proven Execution Engine Creates Competitive Moat: The company's ability to deliver 350 MW of energized capacity by end-2025, including 67 billable MW at Denton alone, demonstrates a repeatable execution capability that outpaces publicly traded peers and positions it to capture premium pricing in a supply-constrained market where "time to RFS is the trump card."
-
Fortress Balance Sheet Funded by Customer Capital: With over $830 million in cash, zero-interest convertible notes, and CoreWeave funding virtually all associated CapEx, Core Scientific has achieved a capital-light growth model that preserves equity value while building a $5-5.5 billion infrastructure asset base, creating substantial debt capacity for future expansion.
-
Concentration Risk Matched by Credible Diversification Path: While current colocation revenue derives from CoreWeave, management's target to reduce this to under 50% by 2028 is supported by a 1.5 GW pipeline, large enterprise demand for 50-100 MW deals, and hyperscaler conversations that restarted post-merger termination, offering a path to investment-grade credit metrics.
-
Valuation Disconnect in Pure-Play AI Infrastructure: Trading at 15.1x sales with a $4.8 billion market cap, the market prices CORZ as a miner rather than an AI infrastructure pure-play, creating potential upside as colocation revenue reaches an inflection point in 2026 with 75-80% gross margins on the CoreWeave contracts.
Setting the Scene: From Bitcoin Mining to AI Factory Builder
Core Scientific, founded in 2018 and headquartered in Austin, Texas, spent its first six years operating in the world of Bitcoin mining. This origin story matters because it forged the company's expertise in high-power digital infrastructure, utility relationships, and rapid deployment at scale—capabilities that became valuable when AI workloads began demanding exactly what Bitcoin miners had built: massive power capacity, high-density cooling, and operational excellence at the terawatt-hour scale. The April 2024 Bitcoin halving, which cut block rewards by 50%, coincided with the emergence of AI training clusters requiring 200+ kilowatts per rack, making Core Scientific's 500 MW of infrastructure more valuable for computing than for cryptocurrency.
The company's strategic pivot began in March 2024 with a series of 12-year agreements with CoreWeave, initially for 500 MW and expanding to 590 MW by early 2027. This transition represents a complete rewiring of the business model. Bitcoin mining revenue declined from $328.8 million in the first nine months of 2024 to $187.0 million in the same period of 2025, a 43% decline that reflects both the halving impact and intentional capacity conversion. Meanwhile, colocation revenue grew from $15.9 million to $34.1 million, a 115% increase that carries different margin economics. The mining segment's gross margin was 4% in the nine months ended September 2025, while colocation margins expanded to 26% in Q3 2025, demonstrating the earnings power of the new model.
This positioning within the industry structure creates a unique value proposition. Traditional data center providers struggle with AI's density requirements, while pure-play miners lack the engineering capability to convert facilities. Core Scientific sits at the intersection, having built a repeatable execution engine for AI infrastructure at scale. The company energized as many megawatts in 2025 as its closest publicly traded peers combined, delivering over 185 MW of billable capacity while competitors were still signing their first AI contracts. This execution advantage translates directly into pricing power, as customers facing GPU shortages and long lead times prioritize "time to RFS" over marginal cost differences.
Technology, Products, and Strategic Differentiation: The Execution Moat
Core Scientific's competitive advantage is the institutional knowledge of building AI factories at scale. The company has evolved its technology stack from H100s to NVIDIA (NVDA) Grace Blackwell platforms (NVL36s, NVL72s, GB200, GB300) and is already evaluating Rubin Vera architectures, demonstrating an ability to adapt to new GPU cycles faster than traditional data center operators. This matters because each architecture shift requires redesigning power delivery, cooling paradigms, and rack layouts. Companies that cannot keep pace lose 6-12 months to redesign, while Core Scientific's "Operation Forward Observer" strategy—advancing development through first commissioned data hall while securing long-lead equipment—enables sub-18-month delivery times, sometimes sub-12 months.
The economic impact of this capability is profound. The CoreWeave contracts project 75-80% non-GAAP profit margins, with power and utilities passed through to clients. This cost-plus structure eliminates commodity price risk while capturing premium rents for speed and reliability. At Denton, the flagship facility in the Dallas-Fort Worth area, Core Scientific delivered 67 billable MW by Q4 2025, supporting over 50,000 Grace Blackwell GPUs. The facility's design accommodates higher rack energy densities than legacy data centers, enabling it to host one of the largest GPU clusters in North America. This technical differentiation allows Core Scientific to command lease rates that are 5-10% higher than market, as tariffs and rising equipment costs continuously drive up economics for new builds.
The company's proprietary thermodynamic solution for managing heat and airflow, originally developed for Bitcoin mining, delivers uptime that translates directly to mining rewards and now to GPU utilization rates. In the mining context, this meant higher Bitcoin yields per terahash; in the AI context, it means higher GPU uptime and thus higher effective compute per dollar invested for customers. This creates switching costs beyond contract terms—customers who have experienced Core Scientific's operational excellence are less likely to risk migration to unproven competitors, even at slightly lower prices.
Financial Performance & Segment Dynamics: Evidence of Transformation
The financial results through 2025 tell a story of deliberate sacrifice and emerging strength. Total revenue declined 38% to $319 million for the full year, driven by the intentional runoff of mining operations. This decline is evidence of management's discipline in reallocating capacity. The company earned 719 Bitcoin in Q1 2025 versus 974 in Q4 2024, a 26% drop that reflects both network difficulty increases and capacity conversion. However, the direct cash cost to mine a Bitcoin rose to $51,035 in Q4 2024, making many machines unprofitable at current hash prices. Management's decision to not refresh the mining fleet until receiving 3-nanometer ASIC chips in H2 2025 demonstrates capital discipline.
The colocation segment's trajectory validates this choice. Revenue grew from $10.3 million in Q3 2024 to $15.0 million in Q3 2025, a 45% increase on a quarterly basis, while gross margins expanded from 13% to 26%. This margin expansion is structural, driven by the completion of data halls at Denton and Marble that enable full rental payments. The company delivered 8 MW of billable capacity at Denton in Q1 2025, with an additional 40 MW by Q2 and 67 MW by Q4, demonstrating linear progression toward the 250 MW target for end-2025. Each megawatt represents approximately $1.2 million in annual revenue at current market rates, implying a run-rate colocation revenue potential of $300 million by early 2026.
Balance sheet strength underpins this transition. The company finished 2025 with $530 million in total liquidity, having opportunistically sold 1,900 Bitcoin for $175 million in January at prices materially above current market levels. This cash position, combined with $625 million in 0% interest convertible notes issued in December 2024, provides a war chest for expansion without equity dilution. More importantly, CoreWeave's funding of virtually all associated CapEx means the company can grow its $5-5.5 billion infrastructure asset base while maintaining light leverage. The company has availability to borrow up to $4 billion against its contracted capacity, with management targeting net debt to adjusted EBITDA leverage of approximately 4x over time, consistent with infrastructure peers.
The hosted mining segment, while declining, provides a stable bridge. Revenue dropped from $71.1 million in the first nine months of 2024 to $18.1 million in 2025 as contracts sunset, but gross margins remained healthy at 27%. The top three hosting customers accounted for 100% of segment revenue in Q3 2025, but this concentration is manageable as the segment winds down. The cash flow from these legacy contracts helps fund the conversion process, effectively making the mining business a self-liquidating asset that finances its own obsolescence.
Outlook, Management Guidance, and Execution Risk
Management's guidance for 2026 and beyond reveals both ambition and pragmatism. The company targets 250 MW delivered to CoreWeave by end-2025 and the full 590 MW by early 2027, representing $725 million in average annual revenue at 75-80% margins. This implies a potential $540-580 million annual EBITDA contribution from the CoreWeave contract alone by 2027. The path to this target is visible: Denton North at 90 MW operational, Denton South at 40 MW commissioning, Marble at 36 MW delivered, and Muskogee Phase 1 (70 MW) and Dalton Phase 1 (30 MW) fully energized and progressing through commissioning for Q2 2026 delivery.
The diversification strategy is where execution risk meets opportunity. Management aims for CoreWeave to represent less than 50% of billable capacity by end-2028, supported by a 1.5 GW pipeline. The company has two sites under short exclusivity arrangements and expects colocation leasing agreements in the near future. Large enterprise demand is emerging for 50-100 MW deals, with customers in the "$75 billion market cap plus" range offering higher returns than hyperscale deals. This matters because enterprise customers typically have faster contracting timelines and less stringent guarantee requirements, providing a quicker path to diversification.
However, the failure to sign a new customer by the Q4 2025 earnings call highlights the rigorous due diligence process. Hyperscaler conversations restarted following the merger termination, but deals require strong balance sheet standing behind the contract, often meaning investment-grade guarantees from chip manufacturers or other backers. The company is evaluating how chip manufacturers might begin providing guarantees to secure their customers, which would unlock neocloud demand. This delay is a sign of market maturation—buyers are scrutinizing power agreements and execution capability more carefully after recent cancellations of contracts with companies that overstated available power.
The "Operation Forward Observer" strategy mitigates this risk by advancing sites through first commissioned data hall before securing customers. The Hunt County, Texas acquisition (285 MW leasable, closing Q1 2026, initial delivery H2 2027) and Pecos conversion (200 MW leasable, RFS within 12 months) demonstrate this approach. By investing $1.5 million per megawatt in the Denton expansion while receiving full rental payments during the first two years, the company de-risks development and provides customers with certainty around RFS timelines. This creates a competitive advantage: the closer a site is to RFS, the better the pricing power.
Risks and Asymmetries: What Could Break the Thesis
The most material risk is customer concentration. For the nine months ended September 30, 2025, CoreWeave accounted for 100% of colocation revenue and Customer G accounted for 78% of total revenue through hosted mining. If CoreWeave were to default or demand renegotiation, the $10 billion contract value would be at risk. However, the 12-year term and CoreWeave's funding of CapEx suggest strong commitment. The company's response is to prioritize counterparty strength over volume, refusing to compromise on credit quality even if it slows diversification. This discipline may delay growth but reduces long-term credit risk.
Execution risk at scale remains significant. The company delivered 350 MW energized by end-2025 but must deliver an additional 240 MW by early 2027 to meet CoreWeave commitments. Supply chain bottlenecks for switchboards, static transfer switches, and generators could delay timelines. The company has secured all equipment necessary for 2025 goals and is insulated from recent tariff impacts, but future builds face 5-10% CapEx inflation that must be passed through to lease rates. The ability to maintain sub-18-month delivery times while scaling to 1+ GW by end-2027 will determine whether the company can capture the full $8.7 billion contract value.
The material weakness in internal controls identified in 2025 is a governance red flag. The restatement of financials due to improper capitalization of assets committed to demolition resulted in a $124.7 million overstatement of property, plant and equipment. While the impact had no effect on cash flows or revenue, it reveals gaps in technical accounting expertise during rapid transition. Management's remediation plan includes additional training and external resources, but the weakness persisted through September 2025. For a company seeking investment-grade credit and targeting hyperscale customers with rigorous procurement standards, this must be resolved to avoid competitive disadvantage.
Regulatory risk, particularly Texas Senate Bill 6, could impact operational flexibility. The bill imposes operational and interconnection requirements, demand response programs, and potential load curtailment during grid conditions. While Core Scientific's utility relationships, cultivated through its Bitcoin mining business, provide some protection, the company could face allocation of upgrade costs that delay project economics. Given that Texas represents a significant portion of gross power capacity, any adverse regulatory changes could materially affect returns.
The terminated CoreWeave merger, while freeing the company to pursue diversification, revealed valuation sensitivity. CoreWeave CEO Michael Intrator called the acquisition "a nice to have, not a need to have," while shareholder Sina Toussi of Two Seas Capital opposed the $16.40 per share valuation as too low. This suggests the market may be undervaluing Core Scientific's infrastructure assets, but also that a future takeover premium may be limited. The $270 million termination fee that could have been payable under certain circumstances highlights the risks of deal-based strategies.
Valuation Context: Pricing a Transformation
At $15.30 per share, Core Scientific trades at a $4.8 billion market cap and 15.1x trailing sales, a multiple that reflects its mining heritage rather than its AI infrastructure future. The enterprise value of $5.7 billion represents 17.8x revenue, elevated compared to traditional data center REITs but justified by the 62.9% projected three-year revenue CAGR. For context, Marathon Digital (MARA) trades at 3.4x sales with negative margins, Riot Platforms (RIOT) at 7.4x sales with -102% profit margins, and Hut 8 (HUT) at a high sales multiple due to its own transformation. Only Iris Energy (IREN) commands a similar multiple at 14.9x sales, but with 68% gross margins from its renewable-powered AI services.
The valuation multiple becomes more reasonable when viewed through the lens of contracted revenue. The CoreWeave contract alone represents $8.7 billion in potential revenue over 12 years, or $725 million annually at stabilization. At 75-80% gross margins, this implies $540-580 million in annual gross profit from a single contract. The company's ability to raise up to $4 billion against this contracted capacity at stabilization, with 60-85% advance rates on build costs, means the equity value is backed by real assets and predictable cash flows.
Balance sheet strength provides downside protection. With $530 million in liquidity at year-end 2025, a current ratio of 1.15, and no near-term debt maturities on its 0% convertible notes, the company has a 2-3 year runway even if colocation ramp delays. The $380 million in net operating loss carryforwards will reduce future cash taxes, enhancing free cash flow conversion. The negative book value of -$3.06 per share is a legacy of the bankruptcy restructuring and asset impairments, not a reflection of current asset value, given the $5-5.5 billion infrastructure base.
The key valuation driver is the inflection point management projects for 2026. As colocation revenue scales to cover operating costs and drive margin expansion, the market should re-rate the stock from a miner multiple to an infrastructure multiple. Peers in the AI data center space trade at 20-25x forward revenue given growth rates and margin profiles. If Core Scientific executes on its 590 MW delivery by early 2027 and diversifies its customer base, a similar multiple would be justified, implying significant upside from current levels.
Conclusion: Execution Premium in AI Infrastructure
Core Scientific has engineered a dramatic strategic pivot in the infrastructure space, converting a distressed Bitcoin mining operation into a pure-play AI colocation platform with $10+ billion in contracted revenue and proven execution at scale. The company's ability to deliver 350 MW energized in 2025 while competitors struggled to sign first contracts demonstrates a moat built on speed, operational excellence, and deep utility relationships forged in the crucible of crypto mining.
The investment thesis hinges on two variables: the pace of customer diversification and the maintenance of execution velocity. If management can reduce CoreWeave concentration below 50% by 2028 through large enterprise and hyperscale wins, the company will achieve investment-grade credit metrics and a lower cost of capital, unlocking its 1.5 GW pipeline. If execution remains on track for 590 MW by early 2027, the 75-80% gross margins will drive EBITDA from negligible levels to $500+ million annually, justifying a re-rating from mining to infrastructure multiples.
The stock's 15.1x sales multiple prices in execution risk but not the potential margin inflection. With a fortress balance sheet, customer-funded CapEx, and a first-mover advantage in high-density colocation, Core Scientific is positioned to capture a meaningful share of the AI infrastructure buildout. For investors, the question is not whether AI demand will materialize—it already has—but whether Core Scientific can continue delivering capacity faster than peers while diversifying its customer base. The evidence from Denton, Marble, and the development pipeline suggests it can, making the current valuation an attractive entry point for those willing to hold through the 2026 inflection.
