Nvidia doesn't have a monopoly in AI - though it sure seems that way

By Daniel Newman

Pricing power and lack of competition are driving Nvidia's growth, but rivals are gaining traction

Nvidia is the poster child for investing in artificial intelligence. But Chief Executive Jensen Huang has been quick to point out that rather than taking market share or fighting for share, Nvidia is all about creating markets that didn't exist before.

The advent of ChatGPT and the ensuing wave of generative-AI investment created a gold rush for Nvidia's products. Demand is so robust and steady that the company has seen its margin expand to an eye-watering 70%+.

Yet with such incredible growth, questions have arisen about whether Nvidia (NVDA) has a monopoly in AI. With estimates of Nvidia holding more than 97% of the data-center GPU market, it's understandable why the company maintains that it is an end-to-end, or full-stack, solution provider for enterprise AI - including the hardware, software, frameworks and libraries developers need to build AI applications. This powerful combination is how Nvidia exhibits monopolistic behavior.

CUDA, the company's parallel computing and programming model, is often touted as the reason developers and users are locked into Nvidia. It's difficult to move AI applications to run on other hardware, so even if AMD (AMD), Intel (INTC), Microsoft (MSFT) and Alphabet (GOOG) (GOOGL) offer homegrown AI chips, it doesn't unlock the chains of using Nvidia.

Therefore, most companies aren't merely buying chips from Nvidia; they're buying entire systems - a software-optimized bundle that incorporates all of the necessary computing, networking and even cabling. In the rapidly growing AI market, the Nvidia component of the bill of materials can now be around 85% - leaving very little for the competition.

Don't call the DOJ just yet

With the U.S. Department of Justice's recent lawsuit against Apple (AAPL), antitrust action is back in the mainstream. Most tend to struggle with what constitutes anti-competitive behavior that would drive regulators to pursue antitrust action. Is this applicable to Nvidia?

The answer is complicated, but for simplicity purposes, I like to focus on two areas that tend to set the bar for an antitrust case. First, how do the company's monopolistic actions harm consumers? Second, how is the company's market monopoly used to stifle competition? Additionally, there is something referred to as the durability of the monopoly as it relates to how long the company has been in this position in the market. What most people don't know is that it is not, in fact, anti-competitive to have a monopoly market share.

In the case of Nvidia, it is hard to identify any consumer harm. A broad set of customers are seeking to build applications and offer AI-based computing services, and Nvidia has the most comprehensive set of solutions to accomplish this. The technology is new and innovative and expensive to produce, and Nvidia's pricing power has come more from scarcity and demand rather than monopolistic behavior.

Anti-competitive behavior requires real marketable competition, and right now, we are only seeing the beginning of any meaningful competition for Nvidia at scale. Competition is coming, and anti-competitive concerns should subside as Nvidia cedes market share and more capable hardware and software alternatives become available.

As these competitive offerings come to market, it will be interesting to see whether Nvidia makes it more difficult for developers to migrate workloads onto other hardware if desired. This will be the big anti-competitive litmus test. It was always about software lock-in, after all. If there is anywhere that Nvidia will get tripped up in an antitrust dispute, it will be the software and Nvidia's insistence on it being optimized only for its hardware.

In many ways, Nvidia is doing for AI what Apple did to make consumers love it so much: building vertical integration that enables simplicity, reliability and improved user experiences. Additionally, AI is so new that it isn't Nvidia's fault that meaningful competition has been slow to market.

From a competitive perspective, Nvidia's future is more opaque. Nvidia will likely have to continue working with others and making other hardware accessible to its software. This process will likely include opening up APIs and simplifying compilers that can take software running on Nvidia and allow it to run on Intel, AMD or another AI chip.

For now, Nvidia seems safe from antitrust action. Yet its incredible growth in the most critical technology sector, coupled with its pricing power and lack of competition, will keep Nvidia and its business model in the crosshairs.

Daniel Newman is CEO and chief analyst at The Futurum Group, which provides or has provided research, analysis, advising or consulting to ServiceNow, Intel, Nvidia, Microsoft, Amazon.com, IBM, AMD and other technology companies. Neither he nor the firm have any positions in any of the companies cited. Follow him on Twitter @danielnewmanUV.

Also read: Nvidia shows it's still on track for AI dominance

More: Nvidia is the 'Magnificent One' now - but these rivals are closing in

-Daniel Newman

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

04-13-24 0923ET

Copyright (c) 2024 Dow Jones & Company, Inc.