The mistake that NVDA investors are making beyond the AI bubble euphoria is that they believe NVDA Revenues (expected to be approximately $11 billion in Q2) will be up and to the right in perpetuity driven by demand for Generative AI capability. Generative AI will not be the gift that keeps on giving. Below we provide a brief AI primer and explain why the Generative AI hype is in fact little more than hype.

In reviewing NVDA’s IR preso, one sees how the company’s Data Center business has enjoyed (and will continue to enjoy, at least in Q2), a Revenue spike driven by CSP’s such as MSFT Azure as the largest CSPs invest in broad AI capabilities including Generative AI.

AI Primer:

Generative AI is of course a subset of AI. AI, at its core, is essentially automated Predictive Analytics. The two letters “AI” could refer to a number of things related to advanced automation. They could refer to Machine Learning (ML), Deep Learning (DL), Neural Networks, Natural Language Processing (NLP), Robotic Process Automation (RPA), Intelligent Process Automation (IPA, which knits together multiple “AI” elements), and more recently, Generative AI.

Generative AI is powered by Large Language Models (LLMs) that are deployed across very large data sets (most famously the public Internet in the cases of OpenAI’s ChatGPT and Google’s Bard). Once the LLMs are sufficiently trained on a given data set, users may pose questions or queries using natural language and receive a natural language response. When users engage with a Generative AI application such as ChatGPT or Google Bard, they are helping to train/refine the underlying LLMs similar to the way in which your Netflix viewing habits train the Machine Learning models that inform Netflix’s movie recommendation engine.

Back to Nvidia:

• Microsoft Azure’s investment in Generative AI was and will be the primary driver of NVDA’s Data Center business in my view. Microsoft is deploying LLMs across Bing search as well as across its Productivity products portfolio.
• Google’s Generative AI effort is a secondary driver as I am not sure that Google is as committed to partnering with Nvidia as Microsoft is. Google, like Microsoft, is deploying LLMs across search (Bard) but more importantly across its Productivity portfolio. It is on the Productivity side where I believe that Microsoft and Google will deliver the most value to users.
• AWS is the largest cloud provider and I am not sure of the pace in which it may produce and deploy the LLMs required for Generative AI.
• I am not sure of the pace in which Oracle may deploy LLMs / Generative AI. Oracle ought to sell itself to Google.

Azure (Microsoft), GCP (Google), AWS (Amazon) and perhaps Oracle will build the infrastructure required for Generative AI as they – especially the first three companies – have the required scale in the cloud and the balance sheets to make the necessary investments in LLMs (side note: Azure, AWS and GCP have offered AI capability in the cloud for years). So, when you read that ServiceNow is building out Generative AI capability, yes, but not really. ServiceNow’s offering will sit on top of GCP. GCP is doing all of the hard, valuable work whereas ServiceNow will build the customer interface. My point being that only a handful of companies are actually building the Generative AI infrastructure. They will do so in a step function over the years and thus the demand for NVDA chips coming from Generative AI use cases won’t be up and to the right in perpetuity.

I believe many companies will experiment with offering Generative AI services to their customers and many will fail because of lack of preparation. We wrote an article on why companies often fail with AI projects. You may read it HERE.

Thus, in the near term I do not believe that the customer value delivered by Generative AI will live up to the hype. Ancillary AI use cases are real, but they won’t result in sufficient revenue generation to justify NVDA’s nosebleed valuation. Heck, even if NVDA generates $50 billion of revenue this year, why pay 20x Revenue for a chip supplier?

Reach out at jmaietta@tek2day.com with your AI questions.

---

• Asset Management
• Investing
• Software
• Strategy
• Technology