Who Was Wrongly Hit by the AI Infrastructure Panic? Bernstein Is Bullish on Riot (RIOT.US) and TeraWulf (WULF.US); IREN's (IREN.US) Power Asset Value May Be Underestimated
As AI infrastructure-related stocks have recently suffered a sharp sell-off, Bernstein believes the market may have overamplified the financing and operational risks facing emerging AI cloud service providers, while overlooking the scarce power assets and long-term contract value owned by some infrastructure developers.
As AI infrastructure-related stocks have recently suffered a sharp selloff, Bernstein believes the market may have overamplified the financing and operational risks facing emerging AI cloud service providers while overlooking the scarce power assets and long-term contract value owned by some infrastructure developers. The firm is particularly bullish on the investment opportunities in Riot Platforms (RIOT.US) and TeraWulf (WULF.US), while also believing that IREN (IREN.US), whose share price has fallen the most recently, possesses asset value that should not be ignored.
Bernstein analyst Gautam Chhugani and his team noted in a research report published on November 16 that over the past 24 months, the emerging AI infrastructure companies covered by the firm have signed a cumulative US$132 billion worth of AI-related contracts, corresponding to approximately 5.5 gigawatts (GW) of power capacity. However, over the past two months, the median decline in these companies' share prices has reached approximately 35%.
Bernstein noted that the initial trigger for this selloff was market concerns that data center construction could face more regulatory restrictions following the U.S. midterm elections. Subsequently, management changes at NVIDIA Corporation (NVDA.US) further intensified investor worries about the AI infrastructure industry.
Investors fear that after Jensen Huang's departure, NVIDIA Corporation's new management may no longer be as active as before in providing credit guarantees and strategic support to emerging AI cloud service providers, thereby weakening these companies' financing capabilities and affecting their data center expansion plans.
However, Bernstein believes the market has not sufficiently distinguished between emerging AI cloud service providers and data center infrastructure developers that own substantial power assets. The latter can not only commercialize through operating AI cloud services but also realize value by leasing data centers and selling assets that already have power access conditions.
Against the backdrop of power supply increasingly becoming a bottleneck for AI infrastructure expansion, the multi-gigawatt power assets owned by these companies themselves carry significant strategic value.
Bernstein's research shows that of the aforementioned US$132 billion in AI contracts, approximately 85% are data center colocation and leasing contracts, with the remainder being AI cloud computing service contracts. In terms of customer composition, AI labs are the largest customer group, accounting for approximately 42% of total contract value; emerging AI cloud service providers account for 23%, large cloud service providers 18%, chip suppliers 11%, and the remaining approximately 6% comes from enterprises and other customers.
However, the firm emphasizes that judging contract risk solely based on the identity of the end customer is inaccurate. Although some AI labs or emerging cloud service providers have limited credit quality themselves, their lease contracts have obtained credit guarantees provided by large technology companies.
Therefore, when assessing the financial risk of AI infrastructure companies, it is more important to determine who ultimately bears the payment obligation and whether the relevant contracts are backed by investment-grade creditworthy entities.
For example, TeraWulf currently has approximately US$27 billion in AI infrastructure contracts, of which approximately 96% of contract value is credit-supported by Alphabet Inc. Class C (GOOGL.US). The company's Justified Data data center project in Kentucky has signed a US$19 billion contract with an AI lab, corresponding to 401 megawatts (MW) of power capacity, and has obtained credit guarantees provided by Alphabet Inc. Class C. In addition, TeraWulf's Lake Mariner data center in New York has signed a US$7.2 billion contract with emerging AI cloud service provider Fluidstack, which also has credit support from Alphabet Inc. Class C.
Cipher Digital's (CIFR.US) contract structure is also relatively robust. The company currently has approximately US$16.5 billion in AI infrastructure contracts, of which approximately 55% comes from Fluidstack with credit guarantees from Alphabet Inc. Class C, and the remaining approximately 45% comes from AWS, the cloud computing division of Amazon.com, Inc. (AMZN.US).
Cipher's Barber Lake project in Texas has signed a US$9.1 billion contract with Fluidstack, with Alphabet Inc. Class C providing approximately US$1.7 billion in credit guarantees. The company's two contracts with AWS have a total value of approximately US$7.5 billion, with Amazon.com, Inc. providing guarantees for the related base rent and operating expenses.
Core Scientific (CORZ.US) currently has approximately US$24 billion in data center colocation contracts, of which approximately 42% comes from CoreWeave (CRWV.US), approximately 42% from AMD (AMD.US), and the remaining approximately 17% from emerging AI cloud service providers with AMD credit support.
CleanSpark (CLSK.US) has signed a data center contract worth approximately US$6.6 billion with Meta Platforms (META.US), involving 175 megawatts of data center capacity in Georgia, with Meta providing guarantees for the related rent and operating expenses.
Bernstein believes that the above contract structures demonstrate that some emerging AI infrastructure companies do not rely entirely on NVIDIA Corporation's credit support. Large technology companies such as Alphabet Inc. Class C, Amazon.com, Inc., Microsoft Corporation (MSFT.US), Meta, and AMD are also providing demand and credit support for the industry's expansion.
From a valuation perspective, Bernstein notes that after the recent sharp correction, the ratio of enterprise value to planned power capacity for emerging AI infrastructure developers has fallen to an average of approximately US$2.8 million per megawatt, down approximately 35% from the second quarter of this year.
Meanwhile, if an 8% to 9% capitalization rate is applied to value the future net operating income corresponding to signed contracts, and the additional debt required for related project construction is taken into account, the median discount of these companies' current enterprise value relative to contract-implied value is approximately 26%.
Bernstein believes that in the current market environment of scarce power resources, some companies can realize value by selling data center assets with power access conditions even without further expanding their AI cloud computing businesses. Therefore, investors should not value these companies solely based on the operational risks of emerging AI cloud service providers.
In terms of specific investment targets, Bernstein is particularly focused on Riot Platforms and TeraWulf. The two companies have approximately US$38 billion and US$27 billion in AI data center contracts, respectively, ranking first and second among the firm's covered companies, with contract terms of 10 to 20 years.
Among them, Riot Platforms has the most significant valuation discount. The company has signed multiple data center contracts with AI labs and AMD, including a project in Corsicana, Texas, involving 910 megawatts of power capacity worth US$28 billion, and a project in Rockdale involving 191 megawatts of power capacity worth US$9.1 billion.
Bernstein estimates that the annual EBITDA corresponding to Riot Platforms' signed contracts is approximately US$2.2 billion. Using an 8.5% capitalization rate and factoring in approximately US$6.55 billion in additional debt expected to be needed for future project construction, the company's current adjusted enterprise value represents a discount of approximately 50% to contract-implied value, with its ratio of enterprise value to planned power capacity at approximately US$3.3 million per megawatt.
However, Bernstein also notes that approximately 98% of Riot Platforms' current contract value comes from AI labs, investment-grade credit guarantees for some contracts are still being finalized, and project financing details remain to be further disclosed.
Therefore, whether the company can obtain credit support from large technology companies and whether it can complete data center construction financing at attractive costs will become important factors affecting Riot Platforms' valuation recovery.
By contrast, TeraWulf has a clearer credit support structure. Bernstein estimates that the annual EBITDA corresponding to the company's signed contracts is approximately US$1.505 billion. Valued at the same 8.5% capitalization rate and factoring in additional debt needed for future project construction, its current adjusted enterprise value represents a discount of approximately 22% to contract-implied value.
TeraWulf's current ratio of enterprise value to planned power capacity is approximately US$2.6 million per megawatt. In addition, the company is actively seeking customers for its Muskie data center campus in Kentucky, which has 1 gigawatt of power capacity, providing potential opportunities for further expanding contract scale in the future.
Bernstein rates Riot Platforms and TeraWulf "Outperform" with price targets of US$30 and US$20, respectively. Based on the closing prices on November 13, the potential upside for the two stocks is approximately 71% and 58%, respectively.
Among the companies covered by the firm, IREN had the largest share price decline over the past two months, falling approximately 58% cumulatively. Unlike infrastructure developers that primarily generate revenue through data center leasing, IREN directly operates AI cloud computing services, so it not only faces data center construction and financing risks but also bears risks related to GPU procurement, equipment deployment, and fluctuations in computing power leasing prices.
Bernstein notes that investors are currently mainly concerned about whether IREN can smoothly execute its complex AI cloud infrastructure construction plan, especially given delays in delivering some computing capacity to Microsoft Corporation. In addition, most of the computing capacity the company expects to put into operation in 2027 has not yet locked in customer contracts, exposing it directly to fluctuations in GPU computing power leasing prices.
However, IREN currently has approximately US$20 billion in AI cloud service contracts, of which approximately 65% of contract value comes from investment-grade creditworthy customers.
Microsoft Corporation is IREN's largest customer, with related contract value of approximately US$9.7 billion, accounting for 48% of its total contract value; NVIDIA Corporation-related contracts are worth approximately US$3.4 billion, accounting for 17%; the remaining 35% comes from AI labs and other enterprise customers. IREN has also issued US$2 billion worth of equity subscription warrants to NVIDIA Corporation, tied to the milestone of cumulative installation of 600,000 GPUs.
Bernstein believes that although IREN has established a close partnership with NVIDIA Corporation, the company's core advantage distinguishing it from other emerging AI cloud service providers lies in its power asset portfolio of up to 5.8 gigawatts, of which more than 90% of capacity has not yet signed customer contracts.
This massive power asset portfolio gives IREN significant strategic flexibility. The company can adjust the pace of AI cloud computing business expansion based on market demand and financing conditions, and can also realize asset value by selling power-related assets, conducting data center colocation business, or establishing partnerships with technology companies other than NVIDIA Corporation.
Bernstein believes that with IREN's sharp share price decline, investors need to re-examine the potential value of its power assets rather than focusing solely on the short-term operational risks facing its AI cloud computing business. The firm currently rates IREN "Outperform" with a price target of US$55. Based on the stock's closing price of US$18.17 on November 13, the potential upside exceeds 200%.
In addition, Bernstein also rates Cipher Digital, Core Scientific, and CleanSpark "Outperform" with price targets of US$15, US$19, and US$11, respectively.
However, the firm also cautions that the AI infrastructure business remains capital-intensive, and low-cost financing, land and power resource acquisition, equipment supply, and on-time data center delivery capability will remain key factors determining whether relevant companies can convert long-term contracts into actual profits.
Related Articles

CITIC SEC: New-type energy storage "15th Five-Year Plan" introduced; returns on generation-side storage deployment expected to improve.

CITIC SEC: OpenAI Accelerates Embodied AI Deployment; Domestic Hardware Supply Chain May Usher in a New Round of Development Tailwinds

CITIC SEC: Short squeeze intensifies aluminum price volatility; focus on the sector's bottom value
CITIC SEC: New-type energy storage "15th Five-Year Plan" introduced; returns on generation-side storage deployment expected to improve.

CITIC SEC: OpenAI Accelerates Embodied AI Deployment; Domestic Hardware Supply Chain May Usher in a New Round of Development Tailwinds

CITIC SEC: Short squeeze intensifies aluminum price volatility; focus on the sector's bottom value






