Wall street and its investment strategists are right to question whether today’s AI capital expenditures can continue and generate a timely, adequate return on invested capital. But perhaps the market is assigning too much attention to the assumption that a decline in AI spending must come at the hand of low revenue realization, and too little attention to the revenue opportunity beyond the existing infrastructure. A more competitive marketplace with revenues distributed among a broader array of AI developers and providers could create a more sustainable transition from AI spending to AI selling, potentially paving a path for new highs.
It’s nearly mid-October and the S&P 500 trades about where it did at the beginning of August. Since August 3, the S&P has traded within just a 4% range. But last Thursday brought a report from the Financial Times stating that annualized revenue for OpenAI was approaching $50 billion at month-end September, compared with a previously reported $70 billion figure. The news appeared to weigh on tech stocks which fell 1.33% on the day, re-igniting the question of whether the forward AI revenues could justify the extraordinary AI capital expenditures.
Much has been written about the sums of money being invested in artificial intelligence. Semiconductors and power providers have benefitted from the race to build increasingly sophisticated AI models. With trillions committed to future infrastructure, markets are understandably questioning whether those investments will generate a timely, sufficient return on the invested capital.
The dominant market concern rests its theory in “what happens when the spending slows?” If today’s demand for compute proves unsustainable, the resulting decline in capital expenditure could weigh heavily on the companies that have deployed capital.
It is a valid concern, but it assumes the reason for the reduction in spending is insufficient return on invested capital. But what if the declining capital expenditures are accompanied by accelerating aggregate AI revenues across the marketplace? I would propose the market is underpricing the probability of this upside case.
As previously mentioned, OpenAI’s annualized revenue through September was approximately $20 billion less than previously reported. Though the discrepancy is noteworthy, I think the more interesting question isn’t why OpenAI’s revenue was lower than previously reported, but whether the market is mispricing the probability of accelerating AI revenues across the entire space.
What if the future of AI revenues does not depend on a single company or model? As with other technological revolutions, competition increases, innovation expands, and revenues spread across a growing collection of businesses customizing and distributing technology. Rather than signaling weaker demand, perhaps a more competitive marketplace could represent a wider commercial opportunity.
Phase one of AI development has rewarded the companies that supply the infrastructure. Every dollar spent constructing a data center represents revenue for someone supplying chips, electric infrastructure, and/or construction services. But capital expenditures are a cost to those making the investments. The returns depend on what happens after the equipment is installed.
Most of the headlines I see are predicated in the fear that a wind-down in infrastructure capital expenditures is expressly due to low demand or underwhelming AI revenues. But what if the infrastructure proves to do as intended, and spending moderates because of sufficient use of the infrastructure and revenue generation from an expanding market for AI products and services?
Other than asking ChatGPT to organize my kid’s three different soccer schedules to one calendar link, the utility for large language models like Chat is under appropriate skepticism. However, not every application of AI requires a large language model. Increasingly capable small and mid-size models could perform specialized tasks at lower cost and increase productivity. As the models improve, businesses may find revenue production in creating, customizing, and distributing these smaller models for specific industries or customers, rather than relying on a general, large model. These smaller models may cost less in compute to run, potentialy improving margins for the AI seller, and increasing access for the AI buyer.
Proprietary data and intellectual property could accelerate this possible transition. For many businesses, customer data and intellectual property are its most valuable assets. Companies increasingly will need AI models compatible with existing security infrastructure allowing them to protect and control where that information sits. For example, Microsoft’s enterprise product suite provides an early test case, integrating across existing applications while maintaining organizational controls in who owns the data. A reduction in customer and company liability helps both the AI model buyer and seller. The result could support a wider collection of companies producing revenues from small model development, licensing, and distribution, utilizing existing infrastructure ultimately smoothing the transition from capital expenditure buildout to revenue generation.
David Waddell laid out the case last week for the possibility for new market highs, and the pessimistic view of AI seems always front and center. Capital expenditures continue to rise, infrastructure demand slows as the forward revenues from AI don’t materialize. But I suggest considering the alternative scenario, where capital expenditures moderate while revenues from model development accelerate into an increasingly capable and competitive marketplace. Revenues shift from the builders and suppliers to the creators and sellers of AI tools and models. And although this case may not materialize, the probability of it occurring, in my opinion, is underpriced in the market strengthening the case for future new highs.
Enjoy your Sunday!
- Matt
Matt Gentzkow CIMA®
Managing Director, Wealth Advisor
Matt Gentzkow is Managing Director and Wealth Advisor at Coastal Bridge Advisors, where he shares his insights with the investment committee and guides clients toward their financial goals. Before joining the firm, he held advisory and wealth strategy roles at UBS and Morgan Stanley. Matt earned bachelor's degrees in finance and economics from Xavier University and is a Certified Investment Management Analyst® (CIMA®). His market insights have been featured in publications including The Wall Street Journal and Kiplinger.
Sources: Ycharts, Financial Times
The views and opinions expressed are those of Matt Gentzkow, CIMA® as of the date of publication and are subject to change without notice. This material is provided for informational and educational purposes only and is not intended to constitute investment advice or a recommendation to buy or sell any security. The information and opinions expressed are based on sources believed to be reliable, including third-party sources, but are not guaranteed as to accuracy or completeness. Third-party information has not been independently verified by CBA. Any forward-looking statements or hypothetical examples are for illustrative purposes only and are subject to change based on market and economic conditions. Past performance is not indicative of future results.