The question of whether we are in an ‘AI bubble’ or not is one of great concern to many of our clients. Recently the phrase, “is the stock market a bubble,” has been searched more than any time in history on Google and, yes, ChatGPT. The word “bubble” alone seems designed to send chills up all our backs, as some of the most famous of recent times – The Great Nikkei Bubble (late 80s), the Southeast Asia Bubble (mid 90s), the Dot-com (Technology-Media-Telecom) Bubble (late 90s) and the Real Estate Bubble (2006-07) – all burst in spectacular fashion. They led, respectively, to 50%-plus market declines and broad economic pain, including Japan’s Lost Decade, Southeast Asia’s multi-year recession, the 2000-03 US economic recession and the Global Great Recession. Interestingly, all evolved out of a bull market with strong fundamentals.
What, if any, is the connection between a bull and a bubble? What signs can help us guess when a bubble is at risk of bursting and becoming a bear? Or more to the point, when should we start worrying whether the present bull is evolving into another bubble that will end in tears?
Our short answer to the last question is: “not yet.” Our longer answer is that most bull markets experience mini bubbles along the way. They usually only become a dangerous, system-wide bubble when a number of conditions are jointly present, including excessive liquidity, misallocation of capital/“over-investment,” excessive use of debt, regulatory errors, fraud and stock valuations that separate from reality. While the present bull, like others before it, has begun at times to evidence some of these elements, for now at least it appears to be early-to-mid innings, with plenty of market potential ahead. To us, there is no certain conclusion that a full-scale bubble will actually form and burst. The remainder of this memo presents the analysis that has led us to this conclusion.
The secular bull
At Federated Hermes, we have been of the view that the US stock market is in a generational long-term secular bull market (only the third since 1929) that would lead stocks up tenfold over 20 to 30 years. Extended bulls like this are powerful because of two things. First, they are born from a long, secular bear market, in which returns are negative-to-flat over a decade or more, and during which many investors are simply wiped out. This creates a scepticism about stocks that can last a generation and feeds the bull that follows. It also creates a legion of regulators who themselves are focused on preventing a recurrence of the blow-up that led to the bear market, along with a generation of corporate managers who remain sceptical about over-investing. All of this extends and protracts the positive conditions that feed the long bull run. Second, secular bull markets are fuelled by one or more new technological innovations that usher in an era of higher-than-normal productivity and economic growth, capturing the imagination of entrepreneurs who build new businesses around it. Ironically, the new innovation that drives the bull is often born out of the “creative destruction process” of the previous bear.
secular bull markets are fuelled by one or more new technological innovations that usher in an era of higher-than-normal productivity and economic growth
The present bull is rooted in the 13-year bear market that began with the collapse of the Dot-com Bubble, continued through the Great Recession and didn’t end until the last echoes of the euro crisis passed, allowing stocks to break through the old 1999 highs in 2013. With a deep reservoir of scepticism to build on, the emergence of the „Third Industrial Revolution” (the Internet, and now AI) has given this bull the new technology and growth needed to drive its long upward run.
Secular bulls can experience severe corrections (in fact, they are fuelled by them), even as deep as 35% as we experienced during the Covid meltdown. But importantly, these corrections tend to be short (less than six months), as the innovation and growth that fuelled the bull remain intact, policy makers are hyper-sensitive about allowing an extensive economic recession to recur, and the lower valuations following the correction attract investors back into the market. We have seen this phenomenon multiple times already in this secular cycle, most recently in the March “Liberation Day” selloff that saw the Nasdaq decline nearly 23%, and again in late October when many tech and AI-related names declined from 10%-20% (some fell even more). In both instances, improved valuations along with policy responses from the fiscal and monetary authorities brought investors quickly back into the markets.
How bull markets can morph into bubbles
All of us dream of a market that simply goes higher gradually and rhythmically, that never overextends itself and never experiences a correction, much less a crash. Unfortunately, this dream never comes, primarily because the very performance of a bull attracts more investors. That usually leads to more outperformance, which leads to more investors and eventually irrational exuberance. Call it human nature. Fortunately, most bulls experience mini corrections along the way, such as in March and October of this year, which temper enthusiasm, create scepticism and restore the bull’s energy. But sometimes, bulls can find a momentum of their own that morphs into something more dangerous: a bubble. At Federated Hermes, we have studied (and/or lived through!) most of the big bubbles of the last 100 years and have concluded they are often characterised by the six conditions noted above that tend to occur in combination with one another. Because there is no science in measuring precisely when one or more of these conditions is pointing to bubble territory, we have developed a simple color-coding system to suggest where along the bull-to-bubble spectrum the current stock market appears to be, which we call our Bubble Condition Monitor.
Simply put: Green means relatively safe, Yellow something to watch and Red in the danger zone. When you blend the six indicators together, you get an overall diffusion indicator today of “green with a tinge of yellow.” Not perfect, but suggestive that we remain pretty far out of dangerous bubble territory, so far. Let’s go through each condition in turn:
Bubble Condition Monitor
Excessive liquidity: Green
Most bubbles are accompanied by excessive financial market liquidity, which results in too much money chasing too few good investment opportunities, often in leveraged structures. Fuelled by loose monetary policy, the inflation of 2022 ensured that the era of zero interest-rate policy (ZIRP) is well enough behind us. In fact, real interest rates are positive, creating a monetary policy that is modestly restrictive. This contrasts with historical bubbles, in which low real interest rates led to excessive liquidity only to cause the bubble to pop when the Federal Reserve shifted to a rate-hiking cycle. In an excessive liquidity environment, the market often makes no differentiation between sound and risky investment opportunities. This does not seem to be the case in the current environment. Take, for example, the difference in the market’s sanguine reaction to the capital expenditures (capex) investment by the Magnificent Seven (Mag 7) companies (largely free-cash flow funded) versus that of Oracle, which has issued a significant amount of debt to fund its expansion program. The market’s ability to be discerning between those two investment opportunities is more characteristic of a rational market than an irrational one funded by excess liquidity.
Massive misallocation of capital/“overinvestment”: Yellow
Another major condition of a bubble is misallocation of capital. Think too many fibre optic cables installed all over the world in 1999, or too many houses built on spec in 2007. The problem with this measure is it’s very hard to see clearly until after the fact–one entrepreneur’s dream can be another person’s nightmare. In the current bull, we find AI investment by the Mag 7 at impressive levels. By some estimates, these mega-cap US companies spent US$300-plus billion on AI alone in 2025, with another estimated US$400-plus billion committed for 2026. Whether this is a misallocation of capital or a brilliant ahead-of-the-curve commitment to a truly economy-wide productivity revolution will not be known for at least another couple of years. But there are ways to get your arms around it. For instance, JP Morgan estimates that the additional monthly fee the Mag 7 would have to charge to get a reasonable return on their collective AI investment so far would be about US$35/month payment per iPhone user in perpetuity. Given that plenty of ChatGPT users are already paying US$20/month for their premium service, this level of investment, so far, doesn’t appear “Red” to us. And throughout the world, our analysts at Federated Hermes are having multiple discussions with virtually all companies about how they are integrating AI into their daily work and production environments, on a real-time basis. Again, this doesn’t feel like a “Code Red” overinvestment to us. We’ve conservatively coloured it yellow because the return on this investment capital is not yet clear. While we expect AI to be transformative, there is some risk that the expected margin gains do not materialise, and that the capex could prove to be a misallocation of resources.
Excessive use of debt: Green
The use of debt is historically problematic because it can create financial contagion that spreads beyond the innovative sectors. Another source of potential contagion is circular financing arrangements, which raise the risk that weakness in one company spreads to the other players throughout the space, creating negative reflexivity. Current debt levels, however, are not growing at a pace that would be seen as problematic, or consistent with prior bubbles. Thus far, nonfinancial corporate debt and loans have grown at a mid-single-digit pace on a year-over-year basis. We are, however, watchful of future capex plans that may require more debt funding. Thus far, we are not seeing widespread concern being evidenced in the bond market. For example, credit spreads on investment grade and high yield currently sit near their lowest levels in history. Mag 7 capex is expected to grow at a US$400 to $500bn pace, but the group remains free-cash flow positive, with a net cash position on their balance sheets and growing free cash flow at a significant pace – even after the capex investment. Similarly, margin debt, a symbol of irrational investor exuberance, is at a multi-decade low as a percentage of the total US stock market value. Of course, there are examples of specific companies funding their AI-related capex with debt, Oracle perhaps being the most visible of this group. And the pounding that their tradeable bonds is receiving in the marketplace is impressive. But for now, the Oracle example seems to be the exception that proves the rule, rather than vice versa.
Dramatic policy errors: Green
From a policy perspective, the administration is fully supportive of the AI buildout and wants the US to be the global AI leader for both economic and national security reasons. This support extends to the overall regulatory regime for which the administration is trying to create a consistent framework throughout every state. In addition, the immediate expensing of capex from the One Big Beautiful Bill creates powerful tax incentives that support the buildout of AI infrastructure. While this is very supportive of AI development, it does carry with it some risk that the government is incentivising a misallocation of resources. The Fed, for its part, has reduced interest rates from multi-decade highs. But it has done so in a careful and deliberate manner, with an eye to both its full employment and inflation mandates. If anything, there are many (us included) who see the Fed as loosening too slowly, rather than too quickly. With a Fed leadership change forthcoming in 2026, we expect the incoming chair to be more willing to assume robust productivity gains from AI than the current one. This could lead to structurally easier monetary policy, raising the risk of a policy error in the future. But it is way too early to predict one.
Fraud: Yellow
In prior bubbles both the novelty of new technology and the prospects of financial gain create incentives for fraudulent activity. So far, there have been high-profile allegations of fraud in private credit including Tricolor and First Brands. These cases received significant media attention and sparked concerns about the possibility of additional “cockroaches” emerging in the near term. Fortunately, the impact of these incidents has been limited. If anything, they underscore the importance of robust underwriting standards in private credit. We remain vigilant, mindful of potential risks that may be lurking beneath the surface and continue to monitor for any future issues. Our study of prior bubbles shows a tendency for these one-off instances to grow into a domino effect. But so far, several weeks have already past and no other “cockroaches” have crawled out of the woodwork. We’ll see. The good news is that, so far at least, the limited examples of fraud have occurred in the private-debt sector rather than in the banking system, about which regulators, still scarred by 2008-09, have remained vigilant. This lowers the risk considerably of a system-wide contagion that would spark a Code Red.
Stock valuations that separate from reality: Green
The epicentres of prior bubbles routinely produce stock multiples that are completely divorced from economic reality. Oftentimes P/E ratios hit 60-90 times earnings for individual companies. While the valuation of AI-focused firms, and in particular the Mag 7, are certainly not cheap, with the exception of Tesla, they are nowhere near those extremely elevated levels and are supported by robust EPS growth, strong balance sheets and significant free-cash flow generation. In particular, EPS growth has surprised to the upside in 2025 and we project US$310 for 2026 and US$355 for 2027, a growth rate of 14% to 15% over the next two years. This is reflective of a market that is growing with its valuations, not away from it. Outside the Mag 7, there have been examples for sure of smaller firms, fresh IPOs and so-called “meme stocks” that have temporarily experienced sharp valuation spikes. But comfortingly, most of these spikes have been hard to sustain and many of these stocks have subsequently already experienced significant pullbacks.
Conclusion
Although the present secular bull market is 12 years old, that is only middle-aged by historic standards and itself is not a cause of concern. It has also experienced a number of dramatic 20% to 30% declines along the way that have refreshed it and restored the investor scepticism needed to keep it running. Against our monitor of common bubble indicators, the present bull still looks pretty grounded in reality: call it “green with a tinge of yellow.” We are sticking with our optimistic outlook for 2026 and 2027, expecting the S&P 500 to advance to somewhere in the 8,600 neighbourhood–fuelled by solid economic and earnings growth, itself supported by AI-driven productivity gains. That said, corrections along the way are likely, particularly next year with the midterm elections upon us and the AI future still uncertain. Look to us to be adding further to stocks when those inevitable corrections come. This bull may someday morph into a bubble, but it hasn’t yet.
Share:
BD016989
