
The Zhitong Finance App learned that although the market continues to worry about the debt burden, customer performance risk, and capital expenditure pressure brought about by Oracle (ORCL.US)'s massive investment in artificial intelligence (AI) data centers, Bernstein believes that Oracle's current stock price is lower than its theoretical valuation when it did not rely on the AI data center business at all.
Bernstein pointed out in a research report released on November 18 that even if all of Oracle's AI data center customers stop performing contracts, related business revenue is completely zero, and the purchased hardware assets of about 40 billion US dollars cannot be realized, the company still has to bear existing debts and interest expenses, and the value per share of its core business can still reach 133 US dollars.
Compared to Oracle's closing price of $115 on November 17, the valuation under this extremely pessimistic scenario is still about 16% higher. As a result, Bernstein reiterated Oracle's “outperforming the market” rating and maintained a target price of $325, which corresponds to a potential increase of about 183%.
Is Oracle still worth $133 if the AI business doesn't make a penny?
Bernstein pointed out that Oracle has become one of the most controversial stocks in the tech industry. Investors' concerns mainly focus on four areas: the economic return on AI data center investments, huge capital expenditure and financing needs, the company's ability to execute, and the long-term sustainability of existing businesses.
In particular, large-scale AI computing power contracts signed between Oracle and technology companies such as OpenAI, although they provide support for future revenue growth, have also raised questions from the market about whether customers can continue to fulfill their contracts.
To assess the potential downside risks faced by Oracle, Bernstein constructed an extremely pessimistic valuation scenario. In this model, AI data center customers such as OpenAI, Meta (META.US), Nvidia (NVDA.US), and AMD (AMD.US) all stopped performing their contracts. Even if the relevant contracts cannot be cancelled in principle, it is assumed that Oracle cannot obtain compensation for breach of contract.
The approximately $40 billion hardware assets Oracle has purchased, such as servers, network equipment, and power and cooling systems, are regarded as worthless, and cannot be continued to be used or sold for cash.
More conservatively, Bernstein assumes that Oracle will still have to bear the debt and related interest expenses previously raised to build an AI data center, and include the stock dilution caused by equity financing in the valuation.
In other words, this scenario not only completely excludes future revenue and profits from the AI data center business, but also preserves the financial burden already generated by related investments.
Even so, Bernstein estimates that a reasonable valuation of Oracle's core business still reaches $133 per share. This valuation is down from the agency's estimate of $137 in February 2026, mainly because Oracle has since increased debt, purchased more hardware, and expanded the AI data center business.
However, Bernstein stressed that this model does not include an “AI winter” scenario where global demand for generative AI completely collapses. Instead, it assumes that Oracle's existing AI customers have defaulted, but other cloud computing and software businesses continue to develop normally.
Traditional businesses still have potential for growth. FY2030 revenue is expected to be close to 100 billion US dollars
Bernstein believes that the market is paying too much attention to Oracle's AI data center investment risks while ignoring the long-term value of its traditional software and cloud computing business. The agency pointed out that Oracle's database business has high customer stickiness and profit margins, and software services such as enterprise resource planning (ERP) and human capital management (HCM) also have strong business stability.
At the same time, traditional computing services that do not rely on generative AI in Oracle Cloud Infrastructure (OCI), as well as sovereign clouds, private clouds, and multi-cloud services, are still expected to benefit from the trend of enterprises continuing to migrate to the cloud.
According to Bernstein's estimates, even if AI data center revenue is completely excluded, Oracle's core business revenue is expected to grow from about 61 billion US dollars in fiscal 2026 to 98 billion US dollars in fiscal 2030. Among them, core business revenue is expected to be approximately US$67 billion in fiscal year 2027, increase to US$75 billion in fiscal year 2028, and reach a further US$85 billion in fiscal year 2029.
As revenue grows and operating leverage effects gradually become apparent, the operating margin of Oracle's non-AI business is expected to increase from 35% in FY2026 to 39% in FY2030.
While continuing to include AI data center financing interest costs, Bernstein predicts that Oracle's core business net profit for fiscal year 2030 will still reach US$27.1 billion, with earnings per share of approximately US$10.8 billion.
In terms of valuation methods, Bernstein refers to the valuation levels of software companies such as Microsoft (MSFT.US), SAP (SAP.US), CRM.US (CRM.US), and ServiceNow (NOW.US). The average expected price-earnings ratio for these companies is approximately 24.4 times. Bernstein assumes that Oracle's core business will have a compound earnings growth rate of about 20% in the next two years, and a reasonable price-earnings ratio of 25 times, corresponding to a valuation of about 133 US dollars per share.
Bernstein believes that considering the competitive advantage of the Oracle database business and the high stickiness of enterprise software customers, its core business can theoretically even enjoy a certain valuation premium, but the agency did not include this factor in the model due to prudence.
The $288 billion data center lease becomes the focus of the market Bernstein believes the risk is exaggerated
In addition to the ability of AI customers to perform, the huge data center leasing contracts signed by Oracle are also an important source of market concern. Bernstein pointed out that Oracle previously signed a data center lease agreement with a total value of about $288 billion, with a lease period of 15 to 19 years. These lease promises are closely tied to the approximately $317 billion in irrevocable contract amounts the company received from customers such as OpenAI.
Investors worry that once major customers such as OpenAI fail to deliver on their computing power procurement promises, Oracle may be forced to bear rent for idle data centers for a long time, causing free cash flow to continue to be pressured.
However, according to Bernstein, this concern overlooks the reuse value of data center assets.
First, Oracle's data center production capacity is not fully put into use at once, but will be gradually delivered over the next few years. Therefore, even if some customers are unable to fulfill their contracts, the company still has time to find other customers or adjust uses. Second, the $288 billion lease commitment is spread over a 15-19 year contract period, which doesn't mean Oracle needs to pay the full amount in the short term.
Bernstein estimates that annual spending on related leases peaked at around $15 billion to $19 billion, and is not expected to reach this level until around fiscal year 2030. Furthermore, the agency believes that as long as the generative AI industry does not completely collapse, the market will maintain a high demand for data centers with sufficient electricity supply.
Even if the original customer withdraws, Oracle may reduce losses caused by idle assets by subleasing data centers or using related facilities for traditional cloud computing services.
Bernstein estimates that the potential annual size of the global infrastructure as a service (IaaS) and platform as a service (PaaS) market is about $1.2 trillion to $1.4 trillion, yet the actual market is currently less than half of this potential size.
Therefore, the data center demand brought about by the continuous migration of enterprises to the cloud is expected to provide additional utilization space for Oracle's existing assets.
$40 billion in hardware assets or not sunk costs
Regarding the GPU and server procurement risks that investors are concerned about, Bernstein also believes that Oracle has more flexibility than expected by the market. The agency estimates that for every dollar of annual AI revenue generated by Oracle, it needs to invest approximately $2.5 to $3 in computing assets, and the estimated useful life of related assets is about 6 years.
However, Oracle usually only procures the relevant hardware about 3 to 6 months before the data center is put into operation, rather than complete all purchases many years in advance. This means that even if AI customers suddenly stop fulfilling their contracts, Oracle may not have borne the full hardware procurement costs for the next few years.
Bernstein estimates that Oracle currently has about $40 billion in hardware assets in the construction or procurement process, including not only GPU servers, but also storage, networking, memory, and power and cooling equipment.
Some of these assets can be repurposed for traditional cloud computing workloads, or the value can be recovered by selling, leasing, etc. Notably, in the pessimistic valuation model of $133 per share described above, Bernstein did not take into account any recovery benefits that these hardware assets might generate. According to the agency, this further reflects the conservative nature of its valuation assumptions.
The return on AI investment is expected to gradually deliver free cash flow in FY2030 or reach $80 billion
Under a normal business scenario, Bernstein remains optimistic about Oracle's growth prospects in the next few years. The agency predicts that as AI data center contracts are gradually converted into actual revenue, Oracle's total revenue for fiscal year 2030 is expected to reach about US$234 billion, which is higher than the company's previously proposed target of US$225 billion. Among them, AI cloud infrastructure business revenue is expected to reach about US$136 billion, which is an important driving force for the company's overall growth.
However, the gross margin of the AI data center business is relatively low. As its share of revenue increases, Oracle's overall gross margin may drop from about 66% in FY2026 to 47% in FY2030.
Despite this, Bernstein expects Oracle's free cash flow situation to improve markedly as data centers are gradually put into operation and capital expenditure growth slows. According to the financing and cash flow plan previously disclosed by management, after completing the additional financing of approximately $20 billion, the company does not expect to raise further capital, and is expected to achieve positive free cash flow in the second half of fiscal year 2028.
Bernstein expects Oracle's free cash flow to reach approximately US$30 billion in fiscal year 2029 and further increase to approximately US$80 billion in fiscal year 2030. Looking at the longer term, the agency's model shows that Oracle's revenue for fiscal year 2035 is expected to be close to 396 billion US dollars, operating profit is about 160 billion US dollars, and free cash flow may reach 233 billion US dollars.
However, these long-term forecasts are highly dependent on the continued growth in AI data center demand, existing customer fulfillment, and the company's smooth execution of expansion plans, and are not sure of the operating results that can be achieved.
Stock prices fell sharply from the high level during the year, Bernstein is optimistic about the risk-reward ratio
By the close of trading on November 17, Oracle's stock price was reported at 115 US dollars. The cumulative decline over the past 12 months was about 47.7%, and the year-to-date decline was about 41%.
Bernstein believes that Oracle's current stock price not only does not fully reflect the value that the AI data center business may create in the future, but is even lower than its core business valuation under extremely pessimistic assumptions. The agency pointed out that the market may be undervaluing Oracle's traditional software business, or may have taken into account the risk of demand collapse in some AI industries ahead of time.
However, Bernstein believes that Oracle's core database and enterprise software business are still highly competitive, and AI data center investments will bring significant additional value to the company once they begin to generate the expected cash flow.
The agency expects Oracle's stock price to gradually recover as more data centers are put into operation in the next few quarters, the execution of customer contracts is verified, and investors' confidence in OpenAI's long-term business prospects improves.
Bernstein said that the shift in market sentiment could take 18 months, or sooner. Based on the support provided by core business valuations and potential long-term growth opportunities for the AI data center business, Bernstein reaffirmed Oracle's “outperforming the market” rating and maintained a target price of $325.