Original author: Long Yue
Original source: Wall Street Journal
Nvidia's earnings season, the most important thing is no longer the numbers themselves.
On May 18, Bank of America securities analyst Vivek Arya's team released a preview report on Nvidia's Q1 earnings, which will be announced after the market closes on May 20, Wednesday, Eastern Time.
According to Nvidia's historical pattern over the past ten quarters, actual revenue is on average 7% to 8% higher than management guidance. The guidance for F1Q27 revenue previously provided by management was $78 billion, based on this, the actual revenue is likely to be in the range of $83 billion to $84 billion, while the current market consensus expectation is only $78.7 billion.
In other words, "surpassing expectations" is almost certain. But analysts believe that what truly affects the market after the earnings report is the following five questions.
Cash returns: Can Nvidia's "stinginess" change?
This is the topic that has the most coverage in the report, and they believe it is the core reason for Nvidia's long-term valuation discount.
Nvidia is currently the largest company by market capitalization in the S&P 500 index, accounting for a weight of 8.3%, surpassing both Apple (peak 7.9%) and Microsoft (peak 7.2%) at their respective historical highs. But the problem is that Nvidia's shareholder return strength is severely mismatched with its size.
The data is straightforward: from 2022 to 2025, Nvidia's free cash flow return rate (dividends + buybacks) averaged only 47%, while the average for similar companies in the industry was 80%, even Nvidia's own average level in the earlier ten years was also 80%.
Meanwhile, Nvidia's current dividend yield is only 0.02%, while the industry average is 0.89%. In equity income funds, only 16% of funds hold Nvidia, while 57% hold Microsoft and 32% hold Apple.
Where did the money go? Analysts pointed out that Nvidia has invested a large amount of capital into its ecosystem—OpenAI, Anthropic, and technology partners. These investments appear controversial to outsiders, with some voices suggesting it is "revolving financing," meaning Nvidia lends money to customers, who then use that money to buy Nvidia's chips.
How much is the valuation discount? Data shows that Nvidia's expected price-to-earnings ratios for 2026/2027 are 26 times/19 times, while the average for the other members of the "Seven Giants" is 49 times/42 times, indicating a discount of nearly 50%.
A more specific comparison is: analysts predict that Nvidia's combined free cash flow for 2026 and 2027 will exceed $430 billion, higher than the approximately $375 billion combined for Apple and Microsoft. However, Nvidia's market capitalization of about $5.46 trillion is about 28% lower than the combined $7.5 trillion for Apple and Microsoft.
Analysts believe that if Nvidia increases its dividends and buybacks, it is expected to attract more long-term funds that prefer returns, narrowing the valuation discount and also dispelling concerns about "revolving financing." They list this change as a "potential catalyst for the second half of the year."
Vera Rubin: When will the next-generation chip arrive?
Nvidia's current main product is the Blackwell series. The market is concerned: when will the next-generation Vera Rubin platform officially be ramped up?
The bank's judgment is the second half of 2026. Vera Rubin (codenamed R200) uses TSMC's 3-nanometer process and shares the "Oberon" rack architecture with Blackwell Ultra, so the product transition is relatively smooth, with limited expected impact on gross margins.
Looking further ahead, Vera Rubin Ultra (codenamed VR300) will be launched in the second half of 2027, adopting a completely new "Kyber" rack architecture, while the proportion of high-bandwidth memory (HBM) in costs will further increase.
The market also wants to hear Nvidia's latest statement on the "trillion-dollar revenue forecast" during the earnings call—Nvidia previously provided a cumulative revenue outlook of $1 trillion for 2025-2027, but contributions from LPU (language processing unit) racks, CPUs, and Vera Rubin Ultra have not yet been included. Will there be an update this time?
Gross margin: Can the 75% defense line be maintained?
Gross margin is one of the core supports for Nvidia's valuation.
Analysts’ judgment: in the short term, due to Vera Rubin using Blackwell's rack architecture, the gross margin during the product transition period is relatively stable. However, in the medium to long term, the rising cost proportion of HBM memory is a continuing source of pressure.
Market consensus expectations show that Nvidia's gross margin is expected to fluctuate within the range of 74% to 75%, and the bank has no objections to this, but emphasizes that any gross margin performance that exceeds expectations will be a positive catalyst.
How will the forecast for the AI accelerator market size be updated?
Bank of America previously provided a "trillion-dollar" forecast framework for Nvidia's AI market in 2025-2027. This earnings report, the market is paying attention to whether Nvidia will update this forecast, especially incorporating three previously unaccounted new growth points:
- LPU (language processing unit) racks
- Vera CPU (Nvidia's self-developed server CPU)
- Vera Rubin Ultra
The bank expects that by 2030, the overall market size of AI accelerators will reach approximately $1.17 trillion, and Nvidia will maintain a market share of about 68% to 70%.
Specifically, Nvidia's AI accelerator revenue is expected to grow from $102.2 billion in 2024 to $800 billion in 2030, while AMD's revenue is expected to grow from $5 billion to $80.1 billion, and Broadcom's from $9.3 billion to $181.9 billion.
Is the competitive threat from Google TPU and CPU exaggerated?
Recently, a rumor has been circulating in the market: as AI enters the "Agentic AI" era, the importance of CPUs will surpass that of GPUs, thereby threatening Nvidia's moat.
The bank clearly disagrees, giving two reasons:
First, Nvidia's self-developed "Vera CPU" will have new developments disclosed at the upcoming Computex conference, and its competitiveness in the independent CPU market should not be underestimated.
Second, in the large-scale deployed Blackwell and TPU clusters, the ratio of CPU to GPU has already been 1:2, which does not align with the narrative that "Agentic AI needs more CPUs."
The conclusion is: while the CPU market is large, it has many competitors (with both x86 and ARM architectures having strong rivals), and Nvidia's dominant position in the GPU/AI accelerator field is unlikely to be shaken in the short term. It is expected that by 2030, Nvidia will maintain about 70% of the revenue share in the addressable AI total market of over $1.7 trillion.
Valuation: The "Tech King" with a 50% discount
Finally, back to valuation. The report uses a set of data to directly point out the current valuation contradiction of Nvidia.
Based on the expected price-to-earnings ratio for CY26/27, Nvidia has a ratio of 26 times/19 times, while the average for the "Tech Seven Giants" (Mag-7) is 49 times/42 times—Nvidia is valued at nearly 50% less.
Based on EV/FCF (enterprise value/free cash flow), Nvidia has a ratio of 28 times/20 times, while the Mag-7 average is 83 times/59 times—this indicates a discount of over 66%.
Based on PEG (price-to-earnings to growth ratio), Nvidia has a ratio of 0.41 times, while the Mag-7 average is 2.61 times and the overall S&P 500 is above 1.3 times.
Bank of America maintains a "buy" rating with a target price of $320, based on an expected price-to-earnings ratio of 28 times for CY27 (excluding cash), which is in the lower-middle range of Nvidia's historical valuation range of 25 times to 56 times.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
