2024: More of the Same
It was a good year for US large cap equities, while somewhat uninspiring for all other major asset classes. The biggest story for US equities is the continued outperformance of the Magnificent Seven, which in turn reflects investors’ infatuation with AI technologies. Despite calls for equity market performance to extend to the rest of the market, little progress was made. Here are the annual performance figures for the S&P 500 with, and without, Apple, Microsoft, Alphabet (Google), Amazon, Nvidia, Meta (Facebook) and Tesla. As you can see, the year-to-year variance is minimal.
2023 2024
S&P 500 24.30% 24.80%
S&P 500 Ex-Mag Seven 15.50% 16.70%
Valuations and Bubbles
After two years of tech-fueled equity appreciation, are we in a bubble similar to the late nineties? This has been a frequent question from our clients over the past few months. There are only vague definitions of what constitutes an asset bubble. The most precise view is you don’t know you’re in one until it bursts. In practice, the best signposts for spotting bubbles aren’t quantitative, like three standard deviations in price from the long-term trend, but qualitative. Bubbles brew when investors accept ridiculous assumptions needed to justify ever higher valuations. Silly assumptions about the growth of new technologies are hard to identify in real time because, by their very nature, they lack historical precedent.
You don’t have to look far to find similarities between now and the late nineties: 1) both have high valuations and narrow market leadership, 2) investors are paying a premium for a handful of companies based upon the yet unproven success of a promising technology.
On the other hand, compared to the tech bubble, valuations of the overall market today are lower, and technology companies’ earnings far more stable. But tech firms are such an important part of the market today that if they stumble, hindsight may declare this period to have been a bubble, even if other factors were at play.
Nvidia
Nvidia, the leader in AI related hardware, is the clear winner and focal point of the AI boom. Commentary about asset bubbles today generally centers around this firm. The company produces around 90% of all AI chips. Analyst consensus suggests they will remain the dominant provider in the space for the next two years. For one day, in June of last year, Nvidia was the largest public company in the world, overtaking Microsoft. Nvidia’s operating margins are over 60%. The operating margin of the full S&P 500 is 14%. The firm absolutely prints money and has grown its market cap at a pace rarely seen in market history. Nvidia is expensive at 55 times trailing earnings, but not a huge outlier to the overall market. Our beloved Costco trades at 54 times trailing earnings.
Nvidia is completely dependent on their customers’ willingness to continue to build AI infrastructure. And today, Nvidia benefits from a lack of competition. Neither dynamic is worrisome at present. Just the Magnificent seven alone are estimated to spend ~$400 billion on AI data centers running on Nvidia products in 2025, and any direct competition is years away. Still, neither competitive advantage will persist to the same degree over the long term.
AI Cost/ Benefit Analysis
Sequoia
The key question for AI is whether the investment in datacenters will be justified by the income they produce. That’s the best marker for spotting a bubble before it pops. David Cahn, of the venture capital firm Sequoia, attempted to quantify how much the major software companies (Microsoft, Amazon, Google, and Facebook among others) would need to earn to justify their AI data center investments. Assuming these firms need to earn similar margins on AI investments as they do in their traditional lines of business, Cahn found a revenue gap of roughly 500 billion dollars per year. This figure is “back of the napkin” math, but it does suggest that the AI investment gravy train will have a shelf life that investors would be wise to not ignore. AI as a technology will need to produce revenue to justify the massive investment at some point. We just don’t know when we’ll reach that point. For reference, Wal-Mart had revenue of $611 billion in 2023, which was the highest of any company in the US. $500 billion in revenue across all AI related firms globally doesn’t appear too outlandish if investors have several years of patience.
Five hundred billion in AI related revenue is difficult to fact-check, when so many related firms, outside of the Magnificent Seven, are private. We do have a glimpse into OpenAI’s books. This is the firm behind ChatGPT, the most used AI model based on 2023 revenue, with 300 million weekly users. OpenAI is private, but they recently raised capital, so some financial information is publicly available. This firm is one of the most valuable start-ups ever and lost (just) $5 billion last year. You can see clear tech bubble parallels here, but this example is in the private markets.
The firm predicts they will have $100 billion in revenue in 2029. Given that it currently generates 40% of all estimated revenue for AI models/platforms, and assuming that holds, the total revenue for all AI models as defined today would represent $250 billion, or half of what it needs to be based on Sequoia’s assumptions. Finding another $250 billion of revenue for new AI applications doesn’t seem unreasonable, given what is in the pipeline for related product development. The more striking observation from this exercise is that investors, known to be a finicky bunch, will need to wait nearly half a decade to see “normal” levels of profitability on their current investments.
McKinsey
Sequoia isn’t the only group making assumptions about the future of the industry. McKinsey, the big consultant, estimated AI’s potential impact on the global economy to be between 17 and 25 trillion dollars (no timeline provided, unsurprisingly). US GDP was ~27 trillion in 2024. We realize the McKinsey report is a marketing tool to justify their services to corporate America, but if you’re looking for examples of outlandish assumptions about the prospects of AI, finding one that suggests it will contribute the entirety of what the US economy produced last year certainly fits the bill.
At this point AI is fueling the S&P 500’s assent to higher valuations that vaguely resemble what the market looked like in the late nineties. The key difference between the two periods is that the market today has solid earnings and marginally lower valuations. Remember, the S&P 500 logged five years of double-digit gains in the nineties before the bubble burst. We’ve had two. The market probably has further to run if left to its own devices, but there are important changes coming to the economy this year that may cause investors to reassess the current valuations of AI-related firms sooner rather than later.
Interest Rates: Then and Now
What ultimately stopped the booming market of the late 90’s was the Federal Reserve increasing interest rates to slow down the economy and fight inflation. The higher cost of capital eventually forced investors to reassess their assumptions about the profitability of tech investments. A similar pattern may develop now.
In 1998 the Fed cut rates after holding them at 5.5% for a year, only to see inflation reverse course and pick up speed. This forced the Fed, starting in the Summer of 1999, to increase rates to 6.5% by the following year.
In 2024 the Fed cut rates after holding them at 5.5% for roughly a year. So far inflation has moved from 2.5% to 2.7%. It’s an open question as to whether the Fed will need to reverse course and follow the late 90’s road map of fighting inflation with higher interest rates. The threat may be enough to slow investors’ enthusiastic embrace of AI investments.
The new administration’s potential tariff, tax and immigration policies are all, if passed as currently understood, inflationary. Even without these policy changes, the US economy may see above target inflation rates for a while longer. This is due to a surprisingly resilient labor market and resulting consumer spending. The “easy advances” for lowering inflation rates are now in the rear-view mirror. Some datapoints in the ISM survey, such as rents and producer prices paid suggest the next few inflation reports will disappoint. This gives the Federal Reserve less room to lower interest rates. Lower interest rates encourage investment. Today, the bond market thinks we’ll get one rate cut from the Federal Reserve in 2025, down from the five 0.25% cuts expected by the market before the election. Interest rate expectations have been highly volatile. It won’t take much new data to push those expectations of cuts into expectations of hikes for the early part of 2026 if inflationary pressures reappear due to policy changes from Washington.
Corporate Earnings Expectations
For the first time in the past three years, the earnings growth currently priced into the market is above what investors expect. For example, in 2024 market pricing implied 0.8% earnings growth and the market delivered 8%. This low hurdle from implied market pricing made it more likely the equity market would outperform. For 2025, market implied earnings growth is 23%. This level of earnings growth generally only happens when the economy is recovering from a sharp recession. Implied market pricing is just a simple model based on the historical relationship between certain bond and equity market values. It’s a way to estimate what level of earnings growth firms need to produce to justify their valuation relative to historical relationships, which brings us to the next point.
The Bond Market
The bond market has been through a rough patch over the past few years, but it remains an important signal to equity investors. The US Treasury market is particularly important as it sets funding costs for the rest of the financial world and provides a less risky, (in theory), alternative to equities. Currently, the Treasury market is very sensitive to inflation expectations. There has been a great deal of progress made bringing inflation down from its peak of 9% in June of 2022, but there has been mild acceleration in CPI readings since August. As mentioned, the possibility of tariffs, tax cuts and immigration policy changes from the new administration have investors concerned that inflation may move higher.
Bond investors aren’t waiting for the new administration to announce their policies before acting. The US Ten-year Treasury hit a yield of 4.75% in the first week of the New Year, up from 3.6% in mid-September. Yields this high, if sustained, have proven to be problematic for the equity market in the past. As a reminder, yields on bonds move inversely to price. When yields rise, the prices of bonds fall. The message from the bond market is that yields need to be higher to compensate investors for the risk of inflation and the growing amount of new debt needed to fund the US Government. Don’t be surprised if we see 5% yields on the Ten-year Treasury over the coming months.
Bond yields spiked in 2022 as we dealt with inflation that proved to be not-so-transitory, and the results were painful. Equity markets suffered, with the S&P 500 falling by 25% at one point during the year. Technology companies led the way down. The important difference between then and now is that in 2022 corporate America was dealing with a profit recession, with overall earnings falling by 5% into early 2023. That type of earnings contraction is unlikely to happen in 2025 as the policies that could stoke inflation will also improve corporate earnings. But it’s an unpleasant reminder that equity markets are sensitive to much more than investor enthusiasm around AI.
Summary
We honestly can’t say with authority if AI stocks are in a bubble today or not. The new technology may quickly prove to be just as important as its cheerleaders are claiming, justifying the massive investment in AI data centers. Being financial people, rather than tech people, we argue that the increase in the price of AI stocks is, at least in part, the byproduct of a strong economy and assumptions about relaxed monetary policy from the Federal Reserve moving forward. Should those assumptions prove incorrect, investors may reassess. We anticipate that 2025 will be characterized by the give and take between what the new administration wants to achieve and what the financial markets will tolerate. AI related tech companies and the bond market will be the center of this debate.
Should you have questions about this commentary, or anything else, please contact your PIM financial advisor. We are grateful, as always, for the opportunity to serve you.