A Different Take on Tech

One of the many problems with Wall Street research is that more often than not, it is an exercise in group think, that is, it generally has the same opinion. It is rare that the other side of the coin is examined. There are many reasons for that, which we will leave unexamined here.

What we want to highlight is a report that came across our desk this week that is out of the ordinary. It questioned Wall Street’s current view of NVIDIA’s virtual monopoly in the AI hardware space. We focused on this particular report as it has to do with NVIDIA, one of the most highly capitalized companies in the world, and one that is a large weight in a number of the market indices. As a result, NVIDIA’s movement is important to virtually all US public equity investors.

The report was written by Marc Weiss, CIO of Open Field Capital. As some of you may recall, Marc is the underlying portfolio manager for the technology portion of the Palumbo Emerging Growth Portfolio. Here is a quick summary of Marc’s report:

Existing Wall Street View of NVIDIA:

- NVIDIA’s near monopoly position in AI chips is likely to be maintained going forward.
- NVIDIA has a “significant” barrier to entry due to its CUDA software.
- NVIDIA’s market opportunity will expand significantly as the market transitions from “training” to “inference”
- NVDIA’s newest chip, Blackwell, delivers a “30X” improvement in inference performance all but guaranteeing NVIDIA’s dominating the inference market going forward.

Open Field’s View:

- NVIDIA’s CUDA advantage helps mostly with training but not as much for inference.
- NVDA’s claim of Blackwell producing a “30X” improvement over their current Hopper chip for inference is closer to 6-7 X when you isolate the speed at which the chip is processing data. A 6-7 X improvement is not enough to fend off newcomers such as Groq (13X)1 and Cerebras (12X) 2 whose products are better than current NVIDIA GPUs. (Note: Groq and Cerebras are private companies.)

Why Does This Matter?

Let’s start with some definitions to better understand Marc’s points.

AI Training: AI model training is the process of feeding an AI model curated data sets to evolve the accuracy of its output. The process may be lengthy, depending on the complexity of the AI model, the quality of the training data sets, and the volume of training data. (Source: Oracle)

AI Inference: Inference is the process that a trained machine learning model uses to draw conclusions from brand-new data. An AI model capable of making inferences can do so without examples of the desired result. (Source: Cloudflare)

In layman’s terms, training is the process of feeding desired inputs and outputs, into a model. Inference is the ability of that model to draw conclusions based on new data that is not in the model.

Weiss’s position is that NVIDIA’s advantage is very strong training but is not as robust for inference. Other new tools in the market appear more cost effective than NVIDIA’s GPUs for inference.

He believes, Like most others, Weiss believes that the inference market could be 10 times the size of the training market. Thus, NVIDIA’s advantage today becomes less of an advantage as the AI market evolves into the much larger inference phase.

When that is combined with what Marc calls the NVIDIA tax (i.e., their very large profit margin) he believes that NVIDIA will be facing some difficult decisions over the next 12-24 months. Marc writes:

“It appears that 2025-2026 could be the time that NVIDIA has to make the tough choice between: Focusing on the training market in order to maintain its high gross margins (~80%) but losing future inference market share, or trying to maintain their dominant AI chip market share for both training and inference (~80%) by sacrificing 10-20 gross margin points in the process.

We believe the street is ill-prepared for either of the scenarios above and thus 2025-2026 could be a very challenging time for NVIDIA’s stock.”

To be clear, Weiss is not suggesting that it is time to sell or short NVIDIA. He believes that training spend will continue to grow strongly in 2025. However, the turn may come when inference spending becomes larger than model training, which he feels could happen in the second half of 2025 or 2026.

Perspective

This report takes us back 25+ years to the original tech bubble. At the time, the overwhelming opinion was that the Internet would be a massive and important change, and that opinion was undoubtably correct. What was incorrect were the ideas of precisely how that would happen. The large majority of those observations were off the mark. We view AI in much the same way. There is little doubt this will be an incredibly important technology in the future, it’s just very hard to see exactly how it will progress, which is why non-consensus views are so important.

If you would like a copy of Marc’s report, please contact us and we will be glad to send it along.

Have a great week!

What We’re Reading

Palumbo Wealth Management (PWM) is a registered investment advisor. Advisory services are only offered to clients or prospective clients where PWM and its representatives are properly licensed or exempt from licensure. For additional information, please visit our website at www.palumbowm.com.

The information provided is for educational and informational purposes only and does not constitute investment advice and it should not be relied on as such. It should not be considered a solicitation to buy or an offer to sell a security. It does not take into account any investor’s particular investment objectives, strategies, tax status or investment horizon. You should consult your attorney or tax advisor.

The views expressed in this commentary are subject to change based on market and other conditions. These documents may contain certain statements that may be deemed forward‐looking statements. Please note that any such statements are not guarantees of any future performance and actual results or developments may differ materially from those projected. Any projections, market outlooks, or estimates are based upon certain assumptions and should not be construed as indicative of actual events that will occur.

All information has been obtained from sources believed to be reliable, but its accuracy is not guaranteed. There is no representation or warranty as to the current accuracy, reliability or completeness of, nor liability for, decisions based on such information and it should not be relied on as such.

All information has been obtained from sources believed to be reliable, but its accuracy is not guaranteed. There is no representation or warranty as to the current accuracy, reliability, or completeness of, nor liability for, decisions based on such information, and it should not be relied on as such.

The views expressed in this commentary are subject to change based on the market and other conditions. These documents may contain certain statements that may be deemed forward‐looking statements. Please note that any such statements are not guarantees of any future performance, and actual results or developments may differ materially from those projected. Any projections, market outlooks, or estimates are based upon certain assumptions and should not be construed as indicative of actual events that will occur.

Past performance is no guarantee of future returns.

Different types of investments involve varying degrees of risk. Therefore, it should not be assumed that the future performance of any specific investment or investment strategy will be profitable.

What We’re Reading

Used vehicle prices expected to continue to stabilize in 2025

Services index shows big jump in prices for December as companies fear tariffs

Trump jolts Fed outlook

All information has been obtained from sources believed to be reliable, but its accuracy is not guaranteed. There is no representation or warranty as to the current accuracy, reliability, or completeness of, nor liability for, decisions based on such information, and it should not be relied on as such.

Past performance is no guarantee of future returns.

Different types of investments involve varying degrees of risk. Therefore, it should not be assumed that the future performance of any specific investment or investment strategy will be profitable.

AI, Artificial Intelligence, inerence, inference, Nvidia, training and inference

General News

By: Adam