Third Quarter 2025 AI: The Self-Perpetuating Machine
The Artificial Intelligence (AI) Boom Broadens and Hits the Self-Perpetuating Phase, Pulling Markets Higher Despite Weak Economic and Labour Market Trends
Current and future investments and planned investments into computing capacity for generative AI tools like ChatGPT, Grok and Perplexity ramped up in Q3. These new investments came on the heels of a period of doubt about just how big the AI market will become and how these generative AI tools will monetize their software to justify their sky-high valuations and the significant investments already made in computing capacity.
Although investment in AI computing capacity increased markedly in Q3, the majority of investment has come from those companies already operating within the AI industry, whose future sales depend on rapid growth and profitability of the AI industry. Most new investment seen in Q3 would be considered ‘circular financing’, whereby suppliers of components required to power the AI ecosystem are financing their customers. The biggest announcements in Q3 amounted to Nvidia making a $100 billion investment in OpenAI, OpenAI committing to purchase $300 billion in cloud computing services from Oracle (over 5 years), and Oracle buying $40+ billion in graphic processing units (GPUs) from Nvidia. So, Nvidia gives OpenAI money to give to Oracle who then buys Nvidia products. Other companies involved in this circle are Microsoft (which owns ~49% of OpenAI) and CoreWeave (which provides similar services as Oracle).
Subsequent to quarter-end, AMD secured a deal with OpenAI whereby AMD will supply OpenAI with GPUs, OpenAI will pay AMD in cash for these GPUs, but OpenAI will receive options to buy AMD stock based on AMD’s stock price. Yes, AI companies are now betting on their suppliers’ stock prices, and if recent history teaches us anything, they can simply announce an intent to buy GPUs from their suppliers, and this now directly benefits them via the bet on their suppliers’ stock prices. Arrangements like this may become a hallmark of what is increasingly looking like a stock market bubble in the making.
In a way, this is how the global economy works. You work an office job at Microsoft and use some of that money to pay the farmer for food and the farmer does the opposite, using some of his hard-earned cash on Windows 11. The issue with the Nvidia, OpenAI and Oracle circle is there isn’t much dispersion of capital outside of the circle meaning that the circular economy may need to absorb all of the losses of the most vulnerable party, OpenAI. OpenAI is expected to be unprofitable for many years to come and the circle’s ability to sustain itself will rely on Nvidia, Oracle, and now AMD’s ability to access capital for many years to come.
We have seen a similar type of circular financing in the past. In the late 1990s, telecom equipment providers like Cisco, Alcatel and Lucent provided credit to internet service providers so they could buy more of their routers. This period of telecom equipment providers financing their customers made revenue growth at Cisco, Alcatel and Lucent look great, and propelled their share prices higher, making them the darlings of the 2000 tech bubble. In hindsight, as we now know, providing ongoing financing to your customer effectively makes you the customer. Once the music stopped, Cisco, Alcatel and Lucent stopped financing their customer purchases and their stock prices collapsed.
It is widely accepted that the launch of ChatGPT in late 2022 and subsequent investment in global computing power helped jolt the global economy out of a post-pandemic zombie state. AI-related stocks have accounted for ~80% of U.S. earnings growth and ~90% of U.S. capital investment growth since ChatGPT launched in November 2022. Without AI investment, global economic growth and the performance of the global stock market would have likely been much less vigorous in recent years. Whether we like it or not, the fate of the global economy and the global stock market, at least in the short-term, is fully in the hands of the AI boom.
Many commentators have noted the similarity between the AI boom and the 1990s tech bubble. There are some aspects that are similar, such as investments in technology dwarfing investments in other sectors, but a key difference that may allow the current AI boom to endure for quite some time is that any AI-related losses are effectively underwritten by megacap technology companies with cash reserves and cash flows unlike anything the world has seen before.
For example, the circle of tightly linked companies above – Microsoft, OpenAI, Nvidia, Oracle, AMD and CoreWeave – have ~US$180 billion in cash on their balance sheets and generate profits of ~US$200 billion per year as a group (as of the last twelve months). Further, if we look at other companies that are a lesser part of the AI boom and/or may want to be a bigger part of the AI boom – Alphabet, Amazon, Meta, Apple, Broadcom and Tesla – these companies have ~US$340 billion in cash on their balance sheets and generate profits of ~US$390 billion per year as a group (as of the last twelve months). Together that is…
~US$520 billion in cash readily available and ~US$590 billion in annualized earnings power!
Plenty of capital waiting in the wings to pounce on any opportunities! All AI investments may not pay off as hoped, but there is certainly enough money to fund the development of a lot of AI projects and keep the global economy humming. This setup is incredibly different than the 1990s when the technology market leaders were cash-strapped and relied on debt and stock issuance, both of which are only accessible when investor confidence is high.
While investments in the ‘new’ economy are chugging along at full steam, investments in the ‘old’ economy’ are sitting near recessionary levels. U.S. trade wars and an extended period of high inflation and high interest rates have hindered new investment in capital and labour in more traditional economic sectors.
Job creation in North America has ground to a halt and this trend is seen all across the world. The booming ‘new’ economy may create high demand for GPUs and particular types of tech workers, but the ailing ‘old’ economy is where most of us work…
To help limit job losses and stimulate economic activity in the ‘old’ economy, central banks have begun cutting interest rates again, with the U.S. Federal Reserve and Bank of Canada both cutting rates by 0.25% in September. It is expected that interest rates will continue to come down over the next 12-18 months.
In the late 1990s, falling interest rates are blamed for further inflating the tech bubble. This time around, considering much of the investment is in the form of cash on megacap technology company balance sheets, falling interest rates are less likely to ignite significantly more investment in AI.
How To Position Portfolios Going Forward?
The AI investment cycle appears apt to continue and falling interest rates should help lower the debt burden of non-AI-related companies and more easily allow them to fund new projects and hire more workers. We have not seen both AI-related and non-AI-related companies performing well in tandem and if this can be achieved, global economic growth may begin to trend upward after an extended period of weakness following the pandemic.
AI-related stocks appear highly valued relative to history while non-AI-related stocks appear more reasonably valued relative to history.
The AI spending spree may continue but it may be at the expense of lower earnings growth for U.S. megacap tech and potentially lower valuations for these companies over time as a result. We may see the beginning of a period of underperformance by U.S. megacap tech stocks as they invest in AI which may not begin to pay off for ten years or more, similar to the losses experienced by megacap technology companies in the 1990s tech bubble. It is already expected that OpenAI will not become profitable for five years or so and someone will need to incur these losses, and the losses of other unprofitable AI companies.
Non-AI stocks appear particularly attractive given their relatively low valuations and the tailwind provided by falling interest rates. In addition to that, governments around the world appear willing to run large budget deficits to sustain global GDP growth, which should be beneficial for cyclical stocks. AI-‘adjacent’ stocks like those in the energy, utilities, materials or industrial sectors may offer above average growth at reasonable valuations and could be where many opportunities arise in the coming months and years.
A rebound in AI investment helped U.S. stocks outperform international stocks in Q3. That said, there is good reason to believe that the rotation from U.S. (primarily tech) stocks to value-oriented international stocks, which began in early 2025, will resume in the coming months as the reality of AI spending and long payback periods of said spending sets in. If tech stocks begin to underperform similar to the 1990s, non-tech stocks, particularly those with commodities exposure, could begin to perform well after many years of underperformance. We have already seen gold, copper and uranium miners begin to perform well and this trend may continue if tech earnings growth begins to trend lower because of potentially inefficient AI spending.
Small- and mid-cap U.S. stocks as a group are predominantly value-oriented and should begin to benefit from trade war resolutions and falling interest rates. The valuation gap between U.S. small- and mid-cap stocks and large-cap stocks remains historically wide.
Canadian perpetual preferred shares remain attractive, boasting tax-efficient dividend yields of 5%-6% and should benefit from falling Canadian interest rates. The Canadian economy has been especially weak in recent months so Canadian interest rates should fall at a faster rate than interest rates of other nations.
Alternative investments continue to generate consistent yields with little volatility.
Although we see pockets of opportunity together with pockets of risk, maintaining a well-diversified portfolio tailored to your risk tolerance is the best way to avoid major negative surprises and participate in the upside provided by asset markets over time. You can be sure that we are tracking all market developments and are making portfolio adjustments to manage risk and pursue returns as investment opportunities arise.
If you ever have any questions or concerns about your portfolio or the investment markets in general, please feel free to reach out to us.
Sincerely,
Steele Wealth Management
The information contained in this report was obtained from sources believed to be reliable, however, we cannot present that it is accurate or complete. Information has been sourced from the RJL Bond Desk or RJ Private Client Solutions, unless otherwise noted. Index and sector returns represented in this commentary are measured using the S&P/TSX Total Return Index and S&P/TSX GICS Sector Indices as detailed in Raymond James Ltd.’s Insights & Strategies: Quarterly Edition. This report is provided as a general source of information and should not be considered personal investment advice or solicitation to buy or sell securities. The views expressed are those of the author and not necessarily those of Raymond James Ltd. (Member Canadian Investor Protection Fund).
This Quarterly Market Comment has been prepared by Steele Wealth Management and expresses the opinions of the author and not necessarily those of Raymond James Ltd. (RJL). Statistics and factual data and other information are from sources RJL believes to be reliable but their accuracy cannot be guaranteed. The performance outlined in the report is net of fees. The client account performance may vary from the model portfolio due to several factors, including the timing of contributions and dates invested in model. The performance reported is that of the account that represents the model, not a composite. Performance calculation for the models may be different than the index used as a reference point. It is for information purposes only and is not to be construed as an offer or solicitation for the sale or purchase of securities. This Quarterly Market Comment is intended for distribution only in those jurisdictions where RJL and the author are registered. Securities-related products and services are offered through Raymond James Ltd.
Commissions, trailing commissions, management fees and expenses all may be associated with mutual fund investments. Please read the prospectus before investing. Mutual funds are not guaranteed, their values change frequently and past performance may not be repeated.
Raymond James (USA) Ltd. advisors may only conduct business with residents of the states and/or jurisdictions in which they are properly registered. Raymond James (USA) Ltd. is a member of FINRA/SIPC.
