The US stock market roared ahead in the third quarter of 2025, fueled by increasingly enthusiastic Artificial Intelligence (AI) investment. The S&P 500 gained 8.1% for the quarter, and is now up 14.8% for the year to date.
The investment in AI data centers has been historic. In one example from this quarter, OpenAI announced plans for $1 trillion in data center investment, and then committed to purchasing $300 billion in computing power from Oracle over five years sending Oracle stock up over 35% in one day. Then, OpenAI and Advanced Micro Devices reached a deal that could potentially give OpenAI a 10% stake in the chipmaker, expanding its supply chain in GPU chips.
There is little doubt that AI is an important innovation – a leap forward in computing capability. Still, many are starting to hear echoes of the late 1990s internet boom, when telecom companies spent over $100 billion building fiber optic cable networks in the belief that the internet’s growth would be explosive. Over time, the internet eventually did become an extremely important development, but the massive overbuilding in the early years wiped out companies like Global Crossing and WorldCom, and crushed the valuations of firms like Cisco and Level3 that had soared during the boom. Cisco’s stock price fell by 88% from its peak and Level3 declined 95%. Of course, eventually firms like Amazon, Netflix, and Google recovered from the dot-com bust to become some of the world’s largest companies.
“Booms start with some tie-in to reality, some reason which justifies the increase in asset values, and then – and this is the critical feature of speculative mood – the market loses touch with reality.”
– John Kenneth Gailbraith, 1955
Booms and busts are all part of the nature of the stock market. As Gailbraith implied in the above quote, it can be hard to identify where the shift happens from justified gains and when the market “loses touch with reality”. Will today’s AI investments produce enough revenue to justify them? The jury is out. On the one hand, the AI business of firms like Nvidia are hugely profitable, and OpenAI and Microsoft are posting rapidly increasing AI revenue. On the other hand, even a leader in the space like OpenAI is still expecting to burn $8.5 billion in cash this year as spending is well ahead of revenue. Are we at the “justified” stage or the “speculative” stage?
Netscape Navigator was released in 1994, making the World Wide Web easy to navigate, and setting up the internet boom. The NASDAQ hit a high in March of 2000, six years later. Although the current boom may not last quite as long given how quickly it has run up, it is certainly possible that the AI rally still has a long time to run, even if it eventually comes back to earth. Markets have a way of overestimating the profits from new innovations, but in the long run they do a pretty good job of incorporating a new reality.
Stocks
Smaller company stocks, as measured by the Russell 2000 small cap index, outperformed large company stocks overall in the third quarter. The Russell 2000 gained 12.4% over the past three months, as interest rates declined and as prospects for the US economy continued to look fairly strong. As of this writing, third quarter GDP is expected to be almost 4%, which indicates solid growth after a slower start to the year.
Larger stocks, as measured by the Russell 1000 large cap index, gained 8% in the quarter. Large stocks were once again led by the so-called “Magnificent 7” (a group of dominant US technology companies that include Nvidia, Microsoft, Apple, Tesla and others), as Tesla gained 40% and Alphabet (Google) gained 37% in the quarter.
International stocks are still well ahead of US stocks for the year, but in the third quarter the MSCI EAFE index of developed country stocks trailed the US indexes noted above, with a still solid return of 4.8%. The MSCI EM index of emerging country stocks outperformed developed countries and returned 10.6%. The emerging markets index was powered by a surging Chinese stock market, which gained over 20% in the quarter.
Bonds
The Federal Reserve reduced interest rates by a quarter point on September 17th. The cut had been widely anticipated, and although it was the first cut in 2025, more are expected in the coming months. The target range (short term) Fed Funds rate now stands at 4.00% – 4.25%, and the market is currently predicting it will be down to 3.0% by early 2027.
Longer term rates declined modestly as well, as the 10-year treasury fell from 4.24% at the end of the second quarter to 4.15% at the end of the third quarter. This has helped reduce 30-year fixed mortgage average rates, which started the quarter at 6.7%, but ended at 6.3%. Should mortgage rate declines continue, it will be a helpful tailwind for the housing market.
Bond values move inversely to rates, and so the declining rates have been a positive for bond indexes. The Bloomberg US Aggregate Bond Index, a broad index of US bonds, gained 2.0% in the third quarter, and is up 6.1% for the year to date.
Commodities
Gold has been a subject of much discussion recently, not only because it has rallied 45% this year through September (and doubled in approximately two years), but also because people are asking what this tells us about investor sentiment and the economy. The answer is – not as much as it first appears.
Gold has rallied while the dollar has fallen, and there has been a lot of talk about whether this reflects a loss of confidence in Fed independence or concern about the US debt level. But if either of those were true, we should see Treasuries declining significantly compared to other countries, and we do not see that.
The primary source of gold’s rally is central bank buying. China, as one example, has increased its gold holdings from a historically steady 3.5% of its reserves until 2022 up to almost 7% now (which is still low compared to developed countries, suggesting gold may have more room to run).
Looking Ahead
The artificial intelligence boom has pushed up stock valuations significantly over the past two years. As of the end of the quarter, the S&P 500 price to earnings (p/e) ratio based on 12 month trailing earnings (a measure of the price of the stock for each dollar of earnings) is up over 30. This is above the 50-year average of 21, but still well short of the peak of the internet boom, when p/e ratios briefly hit 47.
Has the market ‘lost touch with reality’ this time? Is it 1996 or 1999? Markets are impossibly difficult to reliably time. In the current runup, experts have been claiming the market is overvalued for years. But it’s not possible to simply say the market is “expensive” compared to long term average ratios – the market may be correct in forecasting rapid earnings growth.
If you have shorter term needs for funds, it makes sense to put aside those funds in low-volatility investments like money market funds or short-term bonds. However, for longer term investments that have time to ride out the ups and downs of volatile markets our overall guidance is that it is best to select a long-term allocation, aggressively diversify across asset classes, regularly rebalance to ensure your portfolio doesn’t drift toward an increasingly risky allocation, and stay invested through the ups and downs.