AI stocks hit by indiscriminate sell-off! Bernstein names 6 major oversold opportunities: Dell Technologies, Inc. Class C (DELL.US), SanDisk (SNDK.US) and others rated "Outperform"
As AI-related stocks have recently suffered a sharp sell-off, Wall Street has begun to re-examine the risks and returns of AI infrastructure investments.
As AI-related stocks have recently suffered a sharp sell-off, Wall Street has begun to re-examine the risks and returns of AI infrastructure investment. Bernstein analyst Mark Newman and his team noted in a newly released research report that market concerns about data center construction delays, political resistance, and financing risks at emerging AI cloud service providers have caused valuations of some AI-related stocks to fall excessively, while current evidence is not yet sufficient to indicate that AI infrastructure demand is deteriorating across the board.
Bernstein believes investors are applying similar valuation discounts to companies participating in different parts of the AI industry chain, while ignoring the significant differences among these companies in business structure, customer composition, and dependence on large data center construction.
The bank's quantitative analysis shows that Super Micro Computer, Inc. (SMCI.US) and SanDisk (SNDK.US) are the two companies whose valuations have fallen most excessively in the recent sell-off, while Dell Technologies, Inc. Class C (DELL.US), Seagate Technology Holdings PLC (STX.US), and Western Digital Corporation (WDC.US) have also seen some degree of oversold valuation. By contrast, although AI cloud service provider CoreWeave (CRWV.US) has already experienced a major share price correction, its valuation may still not fully reflect potential risks.
In terms of investment opportunities, Bernstein is particularly bullish on Dell Technologies, Inc. Class C, Seagate Technology Holdings PLC, Western Digital Corporation, and SanDisk, as well as data center operators Equinix (EQIX.US) and Digital Realty (DLR.US). All six companies carry the bank's "Outperform" rating.
Bernstein pointed out that an important problem with the current market sell-off is that investors are conflating the degree to which companies participate in the AI industry with their degree of dependence on data center construction by large cloud service providers and emerging AI cloud service providers.
In reality, different types of AI demand face different risks. Infrastructure construction by large cloud service providers and emerging AI cloud service providers is still relatively focused on model training and is directly affected by power supply, regulatory approvals, construction schedules, and customer deployment readiness, making it most vulnerable to data center project delays.
By contrast, enterprise-level AI demand is more concentrated in inference and AI agent applications, which can be deployed through enterprises' own servers, private clouds, storage, and networking equipment, with relatively lower dependence on newly built hyperscale data centers.
In addition, consumer-grade AI can also be commercialized through endpoint devices and ecosystems. For example, Apple Inc. (AAPL.US) is expected to benefit from on-device AI, monetization of its existing user base, and future device replacement demand without directly bearing the risks faced by large data center construction.
Bernstein believes that market concerns that the U.S. midterm election results may hinder data center construction are also somewhat exaggerated. The bank's bottom-up project analysis shows that after the midterm elections, regions in the U.S. that may face stricter data center construction restrictions could see planned data center power capacity rise from about 249 gigawatts (GW) to 306 GW. However, capacity from projects with relatively high confidence of being realized rises only from 78 GW to 105 GW, an increase of about 27 GW.
Of this 27 GW of capacity, about 10 GW has already entered the construction phase and is expected to retain existing approval qualifications, while the remaining approximately 17 GW may be affected by stricter regulatory measures.
Bernstein noted that even if all of these 17 GW of projects fail to materialize, that would amount to only about 9% of the U.S. national capacity of data center projects with relatively high confidence of being realized. In a more realistic scenario, the bank expects that new regulatory restrictions brought about by the midterm elections may ultimately affect only about 2% to 4% of relevant project capacity nationwide in the U.S.
In addition, local governments are more likely to require data center developers to bear additional grid infrastructure construction costs and strengthen protections for local communities, rather than completely blocking data center construction. Therefore, the relevant political risk may manifest more as higher project construction costs and delayed delivery timelines rather than a large-scale disappearance of computing power demand.
The departure of NVIDIA Corporation (NVDA.US) Chief Executive Officer Jensen Huang brings another layer of uncertainty to the AI infrastructure industry. Bernstein noted that NVIDIA Corporation has become an important provider of financing and credit support for the expansion of emerging AI cloud service providers in recent years. According to reports, the company's various commitments and guarantees provided to support customers' data center construction total US$164.5 billion, of which about US$40 billion is used to support AI cloud computing partners, and these partners are mainly emerging AI cloud service providers.
However, Bernstein believes that the relevant agreements NVIDIA Corporation has already signed are contractual obligations and are not expected to be materially affected by management changes. What truly deserves attention are new financing and guarantee agreements that have not yet been signed.
If NVIDIA Corporation adopts a more cautious financing support policy in the future, the cost of capital for emerging AI cloud service providers to build a new round of data centers may rise. But the bank believes this is mainly a longer-term risk and is concentrated in companies most dependent on large AI data center construction.
For hardware suppliers, Bernstein believes that the current component supply shortage may pose a more direct business constraint than data center construction delays.
At present, AI infrastructure demand still exceeds available data center capacity and hardware supply capacity, with GPU, CPU, memory, storage, optical communications, and power-related components all facing varying degrees of supply constraints. Customers are therefore placing orders earlier to ensure access to scarce equipment.
With overall demand still exceeding supply, delays in individual data center projects are more likely to affect hardware delivery timing or supply allocation among customers rather than directly causing demand for the relevant equipment to disappear. Companies with diversified customer bases, scarce products, and flexible supply chains are expected to mitigate the impact of project delays by adjusting delivery schedules or supplying other customers.
Citing Super Micro Computer, Inc. as an example, Bernstein noted that the company recognized about US$2.3 billion of revenue in its most recent quarter that had previously been delayed because customers' data centers were not yet ready, indicating that some delayed deliveries of demand can ultimately still convert into actual revenue.
To further assess the risks of different companies, Bernstein conducted a quantitative analysis of 12 companies under its coverage, focusing on measuring the proportion of large cloud service provider and emerging AI cloud service provider data center business in each company's operating profit and operating profit growth, rather than simply comparing the scale of AI-related revenue.
The analysis results show that the higher the contribution of data center business to operating profit growth, the more pronounced the valuation decline of the relevant company tends to be. According to the bank's regression analysis, for every 10 percentage point increase in the contribution of data center business to operating profit growth, the decline in a company's forward valuation relative to its three-year average expands by about 7.2 percentage points.
Although the research sample is relatively small, this relationship can still explain about 64% of the variation in valuation changes, indicating that data center business risk exposure has become an important factor affecting the recent valuation performance of relevant stocks.
Among them, Super Micro Computer, Inc. has seen the most dramatic valuation adjustment. The company's forward price-to-earnings ratio for the next twelve months has fallen to about 4.2x, about 72% below its three-year average of 15.2x, while Bernstein's model, based on its data center business risk exposure, implies a reasonable valuation decline of about 34%.
This means that Super Micro Computer, Inc.'s actual valuation decline is about 38 percentage points greater than the model estimate, making it the company with the most obvious valuation overshoot among the bank's coverage universe.
However, Bernstein remains relatively cautious on Super Micro Computer, Inc. The bank expects that business related to large cloud service providers and emerging AI cloud service providers will contribute about 68.4% of Super Micro Computer, Inc.'s revenue, 47.7% of its operating profit, and 58.9% of its operating profit growth in fiscal 2027.
In addition, high customer concentration, uncertainty in converting backlog into revenue, business execution risk, and potential equity dilution also mean that the company deserves a higher risk discount.
Bernstein currently assigns Super Micro Computer, Inc. a "Market Perform" rating with a target price of US$42. The bank believes that if the company's business execution stabilizes and AI infrastructure demand remains strong, its low valuation will offer substantial potential for share price recovery, but investment risk remains significantly higher than that of Dell Technologies, Inc. Class C.
SanDisk is another company Bernstein believes has seen its valuation severely over-compressed. The company's forward price-to-earnings ratio for the next twelve months has fallen to about 3.7x, about 65% below the historical average of 10.5x available since its separate listing, while Bernstein's model implies a valuation decline of only about 36%.
In addition to data center construction delays, the market is also concerned that new NAND flash capacity coming online from 2028 to 2029 may cause product prices and corporate earnings to quickly revert to normal levels. However, Bernstein believes the market may be underestimating the protective effect of SanDisk's long-term customer agreements on earnings, as well as the consumption of NAND capacity by new AI storage applications.
SanDisk's new business model agreements already cover about 50% of flash bit supply for fiscal 2027 and about two-thirds of supply for fiscal 2028. These agreements include customer purchase volume commitments, pricing mechanisms, and financial protection clauses, and are expected to reduce cyclical volatility in the company's future earnings.
At the same time, the development of enterprise-level AI inference and AI agent applications will drive enterprises to generate, process, and store more data, creating new demand for SanDisk's edge storage business outside its data center business.
Bernstein also noted that new applications such as single-level cell (SLC) NAND and high-bandwidth flash (HBF) may consume far more wafer capacity than traditional NAND products, thereby offsetting the impact of future new capacity on the industry's supply-demand landscape.
The bank estimates that if high-bandwidth flash accounts for 5% of NAND bit demand in the future, with its wafer consumption intensity about 6.5 times that of traditional triple-level cell (TLC) NAND, the industry may need to add about 28% more wafer capacity to meet related demand. This would help extend the period of tight supply and rising prices in the NAND industry.
Bernstein currently assigns SanDisk an "Outperform" rating with a target price as high as US$3,000. Based on the stock's closing price of US$937.46 on November 12, the potential upside is about 220%.
In the AI infrastructure hardware sector, Dell Technologies, Inc. Class C is Bernstein's top pick. The bank believes that Dell Technologies, Inc. Class C combines relatively limited earnings risk exposure to large AI data centers, strong enterprise-level AI business growth, high-quality backlog, and attractive valuation.
Bernstein expects that AI server business related to emerging AI cloud service providers will contribute about 26.9% of Dell Technologies, Inc. Class C's revenue in fiscal 2027, but due to the relatively low margins of the related business, its contribution to operating profit is only about 11.7%, and its contribution to operating profit growth is about 16.1%.
Most of Dell Technologies, Inc. Class C's earnings still come from traditional servers, enterprise infrastructure, and storage businesses, which are relatively less affected by delays in large data center construction. In addition, Dell Technologies, Inc. Class C's traditional server orders in the second quarter of fiscal 2027 grew 122% year over year, reflecting that enterprise infrastructure modernization, market share gains, and emerging AI inference demand are creating new growth drivers for its traditional business.
Dell Technologies, Inc. Class C currently has about US$95 billion in AI server backlog, and the relevant orders have undergone customer credit, financing, and technical design assessments. Company management said that the planned server deliveries match customer data center readiness by about 80% to 90%, and most of the backlog is non-cancelable.
Bernstein believes that Dell Technologies, Inc. Class C has more than 6,500 AI customers, covering emerging cloud service providers, sovereign AI, and enterprise customers, and that its diversified customer base and supply chain management capabilities will help mitigate the impact of delays in individual large projects.
Currently, Dell Technologies, Inc. Class C's forward price-to-earnings ratio for the next twelve months is about 12.6x, about 9.6% below its three-year average of 14x. Bernstein believes the market is overly focused on the data center construction risks of emerging AI cloud service providers while underestimating Dell Technologies, Inc. Class C's enterprise-level AI growth opportunities, diversified earnings sources, and business execution capabilities, and recommends that investors use the recent share price weakness to increase positions.
The bank assigns Dell Technologies, Inc. Class C an "Outperform" rating with a target price of US$650, about 75% above the November 12 closing price of US$371.17.
By contrast, although Hewlett Packard Enterprise Co. (HPE.US) has lower direct dependence on large AI data center construction, its valuation has already fairly fully reflected this defensive advantage. Bernstein believes that compared with Hewlett Packard Enterprise Co., Dell Technologies, Inc. Class C has more attractive fundamentals and valuation upside.
In the hard disk drive sector, Bernstein remains bullish on Seagate Technology Holdings PLC and Western Digital Corporation, believing that the hard drive price upcycle is still in its early stages and that future price increases are expected to accelerate further.
Strong demand, limited industry capacity, and cautious capacity expansion by suppliers will continue to support hard drive price increases. At the same time, the spread of enterprise-level AI inference and AI agent applications is expected to drive enterprises to generate and store more data, thereby expanding hard drive demand and gradually reducing the industry's dependence on large cloud service providers.
Bernstein is particularly bullish on Seagate Technology Holdings PLC's leading advantage in heat-assisted magnetic recording (HAMR) technology. This technology can increase hard drive storage density, enabling the company to provide more storage capacity without increasing hard drive output proportionally, thereby strengthening its cost-per-unit-of-storage advantage.
Currently, Seagate Technology Holdings PLC's forward price-to-earnings ratio for the next twelve months is about 11x, about 46% below its three-year average; Western Digital Corporation's corresponding valuation is about 9.9x, about 55% below its historical average.
Bernstein believes that although both companies currently rely relatively heavily on large data center demand, the hard disk drive industry is still in a price upcycle, and coupled with expanding enterprise-level AI storage demand, both companies offer attractive risk-reward. The bank assigns Seagate Technology Holdings PLC and Western Digital Corporation "Outperform" ratings with target prices of US$1,100 and US$600, respectively, about 102% and 93% above their November 12 closing prices.
Bernstein is similarly bullish on Equinix and Digital Realty, believing that the market should not treat traditional data center operators and emerging AI cloud service providers as equivalent.
Equinix's business is mainly concentrated in enterprise-level data center colocation and interconnection services, with relatively low direct dependence on large AI model training projects. The company's normalized adjusted monthly recurring revenue in the third quarter achieved double-digit growth for the fourth consecutive quarter, and it raised full-year guidance for the third consecutive quarter.
Digital Realty has more data center business serving large cloud service providers, but its customer quality is high, and its existing cloud computing portfolio has a large number of lease contracts with terms of 15 to 20 years, with direct exposure to emerging AI cloud service providers close to zero. The company's leasing business set a record in the third quarter and it raised its full-year core funds from operations (FFO) guidance.
Bernstein assigns Equinix and Digital Realty "Outperform" ratings with target prices of US$1,270 and US$226, respectively.
By contrast, Bernstein believes CoreWeave still faces the risk of further valuation downgrades. The company's business relies almost entirely on AI data centers and needs to rapidly convert signed computing contracts into actual operating capacity and achieve profitability after deducting depreciation and financing costs. Although its forward enterprise value-to-EBIT ratio has fallen about 61% from its previous peak, Bernstein's model shows that its valuation may still have room for further compression. The bank assigns CoreWeave an "Underperform" rating with a target price of US$54.
In addition, although Apple Inc. has played a relatively safe-haven role in the recent AI stock sell-off, its forward price-to-earnings ratio for the next twelve months has reached 37.3x, about 24% above its three-year average. Considering rising memory costs, pressure on gross margins, and slowing growth in the services business, Bernstein believes that once market risk-aversion subsides, Apple Inc.'s share price may also come under pressure in the near term.
Related Articles

MANULIFE-S (00945) plans to issue subordinated debentures with a principal amount of C$750 million.

WuXi AppTec (02359): Court approves dismissal of 1260H lawsuit
.png)
SNACK EMPIRE (01843) subsidiary intends to sell Singapore property.
MANULIFE-S (00945) plans to issue subordinated debentures with a principal amount of C$750 million.

WuXi AppTec (02359): Court approves dismissal of 1260H lawsuit

SNACK EMPIRE (01843) subsidiary intends to sell Singapore property.
.png)





