In an increasingly short-term world where market indices are at or near record levels of concentration in the buildout of artificial intelligence (“AI”) technology, we believe that our ability to think long term enables us to think independently from the crowd and position our client portfolios for superior risk-adjusted returns.

Last year, at a time when uncertainty surrounding tariffs was impacting the stock market, I wrote about the importance of V-E-L-A (an acronym representing our cornerstone investment principles). At the time, I mentioned that many businesses were trading at prices we found attractive relative to our estimates of intrinsic value, and the VELA team was buying accordingly. In hindsight, that certainly was an opportune time to buy! While all the components of our firm’s acronym retain their significance, I want to focus today on the “L,” or “Long-Term Orientation.”  Everything we do at VELA– how we invest in our people, how we analyze and invest in companies, and how we structure investment portfolios – is done through the lens of a five-year (or longer) time horizon. This requires us to think about and analyze businesses over the course of a business cycle, not just the next quarter. We firmly believe that this is the best way to build and compound client wealth in a prudent, methodical, and importantly, sustainable manner. It is on my mind today because, in our view, it is most imperative to think and act with a long-term mindset at a point in time when 1) we believe that many market participants are optimizing for the short term; and 2) market exuberance surrounding the prospects of AI appears to be reaching peak levels.

The world feels increasingly short-term. 

A long-term mindset increasingly feels like a differentiator. By the day, the world itself feels more short term. It is difficult to keep up with the daily barrage of headlines, most of which are disquieting; it has never been easier to gamble on live sports or other aspects of our daily lives; and social media news feeds are designed to capture our attention– often in less than productive ways. Frankly, it has never been easier to be distracted. It has never been more difficult to think long term. At the same time, it has also never been so important. Not surprisingly, market participants’ time horizons and attention spans seem to have shrunk in tandem with our increasingly distracted and short-term world. Similar to the “meme stock” phenomenon in 2021 (prior to the market bust in 2022), retail traders have returned to the stock market in a big way. Many are trading frenetically in zero-day call options (levered bets on single day stock price movements), where volumes are at an all-time high.¹ Investor margin debt (investors borrowing money to buy stocks) now rivals or exceeds both the meme stock and DotCom bubbles.² The enthusiasm has extended beyond our shores to South Korea, where investor margin debt is also at all-time highs and speculators are pulling money out of real estate and life insurance policies to take levered positions on stocks.³  This is not normal behavior, but in our view, speaks to the near-term enthusiasm surrounding AI, the “fear of missing out” on near-term investment performance, and reflects market participants optimizing for near-term results.

This has been a concentrated, momentum-driven market led by the buildout of AI technology.

Momentum-driven markets are characterized by sharp, upward price swings in shares of the perceived ‘winners’ of a given era, as speculators clamor for a share of the spoils. Since 2023, we have observed one of the strongest momentum-driven markets since the DotCom bubble of 1999-2000, led by enthusiasm for stocks connected to the buildout of AI technology.⁴ This has accelerated in the past several months due to strength in semiconductor stocks, which now represent a record percentage of broad market indices.⁵ Five companies (Alphabet, Amazon, Microsoft, Oracle, and Meta), known as the ‘hyperscalers’, have provided the fuel for this momentum market’s fire in the form of more than $500 billion in run rate capital expenditures that are expected to climb to greater than $1 trillion within the next 12 months.⁶ These capital expenditures are earmarked for the buildout of data centers containing the most advanced semiconductors for AI applications for both external customers (ex: Anthropic and OpenAI) and internal business development.  Meta’s capital spending on high-end GPUs, for example, supports Nvidia’s growing cash flow and earnings. These capital expenditures are translating into rapid earnings growth for Nvidia, now the largest market capitalization company in the world at nearly $6 trillion, as well as a host of companies in the semiconductor and surrounding supply chains (let’s call them the “AI buildout players”).  It is also driving the free cash flow generated by the hyperscalers towards zero (if not negative) and requiring them to increasingly raise capital via the debt and equity markets.

The AI buildout players are a relatively narrow group of companies who are perceived to benefit, at least in the near term, from the buildout of AI technologies.  They have lifted markets to historic valuation levels and now represent a very significant percentage of nearly all US market indices, both small and large.⁷ By our estimates, nearly 50% of the U.S. stock market’s value is now comprised of a relatively narrow list of ~50 of the AI buildout players. Collectively, the AI buildout players trade at ~15x trailing twelve-month sales, ~40x forward earnings and account for greater than $35 trillion in market capitalization.⁸ Regarding the future stock return prospects of this group of companies, I asked Claude via Microsoft Copilot the following question:

Query: In the history of the Russell 3000 or S&P 500, what are the odds of companies trading at 15x sales outperforming the broader market index over a 5-to-10-year period?

Claude via Copilot: The Short Answer – Very Low. The historical base rates strongly suggest that stocks trading at 15x sales have a low probability (likely 10-20%) of outperforming the broader index over a 5–10-year period.

The great investor Sir John Templeton famously said, “The four most dangerous words in investing are: ‘this time is different.’” Yet, this time will have to be different for long-term investment results to be substantially more positive than history for this cohort of businesses. Though not impossible, from a risk-adjusted framework, we like our odds against this cohort in both (1) protecting client worth and (2) building long-term wealth.

Taking the long view, the AI buildout is a traditional capital cycle. 

Taking a step back to see the longer-term picture, we view the immense buildout of AI data centers as part of a traditional capital cycle, with both peak (as we are accelerating into now) and eventual trough and normalization phases. The limitation: we’ve never seen a full cycle in AI, and as with most technology-driven cycles, it is tough for anyone to pinpoint definitively where in the cycle we are. Right now, the prevailing narrative is that demand for computing power exceeds supply due to advancements in areas like Claude Code from Anthropic and agentic applications that are feeding an insatiable demand for computing power.  This is true. However, the laws of economics tell us that supply and demand will come into balance, eventually. With $1 trillion+ in hyperscaler capital expenditures planned for 2027, potentially exceeding national defense spending in the United States, one thing is certain: supply is coming, and in a big way over the next several years.

If you have an investment horizon of 12 months or less, as many market participants seem to have these days, the idea of the AI buildout slowing down or supply meeting demand seems too far in the distant future to matter.  For now, “party on!” is the prevailing mantra.

What could cause the cycle (and/or market sentiment) to turn?

Market participants appear to be treating the upward trajectory in hyperscaler capital expenditures as something that is unlikely to change anytime soon, as evidenced by the high valuation of the AI buildout players. When or how the cycle peaks is anyone’s best guess, but we believe 2026 is likely the peak year for hyperscaler capital expenditures growth (~100%) and 2027 could decelerate meaningfully (analysts estimate ~20 – 40% capital expenditures growth in 2027). 2028 consensus estimates for hyperscaler capital expenditures assume a plateauing around 2027 levels.⁹ Thus, it’s fair to say that 2026-27 will likely be peak earnings growth years for the AI buildout players. We believe the majority of market participants today are underestimating the capital cycle and associated risks. These same participants are playing with a dangerous cocktail of 1) paying peak multiples of sales for the AI buildout players; 2) at peak margins; and 3) likely closer to peak-of-the cycle earnings.

We pose a series of questions intended to illuminate potential risks associated with the AI buildout cycle:

Hyperscaler-Related Supply / Demand Risks:

• What if supply catches up to demand because hyperscalers discover new ways to improve their return on investment on their enormous spend? What happens to Nvidia as the hyperscalers, its largest customers, increasingly prioritize their own custom chips that are a fraction of the cost and increasingly compelling on a relative performance basis?
• What happens to semiconductor firm average selling prices (“ASPs”) and margins that currently reflect shortages when supply eventually meets demand? This is especially relevant in more commoditized areas like memory – a matter of “when,” not “if.”
• What if we end up with an over-supply of semiconductors in the United States due to permitting issues, societal pushback, or lack of power infrastructure? A significant number of data center projects have been delayed or cancelled due to these issues. Given the material time lag from the time of cash expenditure to the physical implementation/monetization of this spend (ranging from quarters to years), to what extent does this exacerbate the peak of the earnings cycle?
• What happens if depreciation charges from these increasingly high capital expenditure levels materially eat into hyperscaler earnings power (beyond what they can offset from laying off employees)? On a related note, what happens if investors in hyperscalers lose patience with them outspending their cash flows and levering their balance sheets for returns that remain in the distant future? What will this digestion period look like if they pull back on capital expenditures?
• As hyperscalers increasingly look to the debt markets to finance their capital expenditures, what happens if the appetite amongst debt investors dries up? What if future venture capital fundraising, which has been a source of growth for tech companies, tightens or freezes up?
• What if there has been a coordinated attempt amongst the tech giants and their venture capital customers to engage in circular financing arrangements, leading to artificially high growth rates?
• What if we find that OpenAI has serious going concern risks and a low probability of a long-term path to profitability?
• Similar to Super Micro Computer allegations of illicit chip smuggling into Chinese entities via Southeast Asia, what if we discover larger scale fraudulent activity at the upper end of the AI chip market?

Higher-Level Market Risks:

• What happens in a momentum-driven market when earnings growth decelerates?
• What if we find out that S&P 500 margins are peak of the cycle due to over-earnings in the tech sector? S&P 500 operating margins peaked in 2000 and did not regain those highs until 2018.
• A surge of IPO activity usually shortly precedes market peaks. What happens if the record setting IPOs of SpaceX, Anthropic, and OpenAI IPOs are timed at or near the peak of the cycle? What happens if these same IPOs raise money at unsustainably high valuations (Space X is intending to raise capital at ~100x trailing twelve month sales)?
• With indices, hedge funds, levered retail investors, and others “all-in” on the AI buildout trade, what happens when growth slows and/or there are signs of supply catching up to demand?
• When the AI cycle turns, what businesses and investment managers will be caught flat footed for managing to near term performance and taking undue amounts of risk?
• Where can investors find diversification from risks of an AI overbuild when nearly all broader market indices are now exposed to this risk in a material way?

These scenarios are worth contemplating when the bull case for AI is viewed as virtually infallible, tech earnings growth rates are at or near peak of the cycle, venture capitalists are selling enormous amounts of shares to the investing public, and hope regarding the prospects of AI springs eternal.  We believe that these risks, collectively, are under-appreciated today based on prevailing valuations of AI buildout companies and their concentrations in the broader market indices. As a result, our strategies increasingly look and behave differently than the broader market indices. This is a feature, not a bug in our process.

Similar to prior ground-breaking technologies, we believe that AI will enhance productivity.

AI is undoubtedly a tool that will lead to productivity gains.  The advances with Anthropic’s Claude, in particular, have been very impressive. We have witnessed firsthand the “magic” of AI in our workflows and have our eyes wide open regarding current and future benefits of this technology.  We view many of these enhancements as productivity-enhancing, similar to the advent of the internet and other ground-breaking advancements over the course of human history.  We have and will invest in companies where AI is a material part of the business model, but are only willing to invest when the businesses are priced at a discount to our estimates of intrinsic value. Alphabet, Inc. (Google), a long-term holding at VELA, is a great example of our investment process at work. Shortly after the release of ChatGPT in late 2022, this company was perceived as an “AI loser.”

The prevailing narrative at the time was that Google Search was dead and potential antitrust headwinds weighed on the stock. It was the view of our excellent tech analyst, Chris Brinich, at the time that no company was in a better position to navigate AI and no company had invested more in AI than Alphabet over the preceding decade. Alphabet is now seen as one of the clearest beneficiaries of AI. We expect these sharp narrative swings to prevail over the next few years as the market searches for the clearest, first order winners from AI. Similar to our experience with Alphabet, we suspect that many of the eventual ‘AI winners’ are not the perceived winners of today, while many perceived ‘AI losers’ will adapt to new realities better than the market currently anticipates.

Positioning our client portfolios for long term performance.

In summary, momentum investing has rarely been as much in style as it is today – buy the winners, ignore everything not tied at the hip of the AI buildout cycle. This has led the S&P 500 to elevated valuations across nearly any measure.¹⁰ At this time, the incentive for investment teams with a shorter time horizon (think 12 months or less) is to join the crowd. Optimizing for short-term performance (often at the expense of long-term performance) will never be our game. Nobody knows when the cycle will turn, and our strategies will undoubtedly lag the market when the momentum trade reaches “fever pitch” levels, as it has recently. However, it is our job as valuation centric, long-term investors to think over the course of a business cycle and to structure our investment portfolios for the best risk-adjusted expected returns over a 5-year plus time horizon.

The VELA investment team continues to work tirelessly to structure our portfolios to reflect our highest conviction investment ideas. Our portfolios increasingly look different from broader market indices.  A growing number of the companies we own trade less than 15x free cash flow, not sales. Many of the companies we own, especially those at the lower end of the market cap spectrum, are trading at the most attractive valuations we have observed in 10+ years while the broader market trades at all-time highs. Importantly, these businesses trade at attractive discounts to what we believe they are worth. We believe the odds are in our favor when we pay attractive prices for businesses with strong competitive advantages, generating strong free cash flows, run by owner-oriented management teams, and buttressed by strong balance sheets. These companies are not standing still, and we expect many will leverage AI to defend their competitive advantages over the next five years. We believe that our investment philosophy is a time-tested, enduring strategy that will benefit our client portfolios in the years and decades to come.

We thank you for your trust in us and truly appreciate your partnership over the long term.

Disclosures:

The views expressed are those of VELA Investment Management, LLC as of 6/4/26 and are subject to change. These opinions are not intended to be a forecast of future events, a guarantee of future results, or investment advice. Third-party information in this report has been obtained from sources believed to be accurate; however, VELA makes no guarantee as to the accuracy or completeness of the information.

VELA Investment Management, LLC is a registered investment adviser. Information presented is for educational purposes only and does not intend to make an offer or solicitation for the sale or purchase of any specific securities, investments, or investment strategies. Investments involve risk and unless otherwise stated, are not guaranteed. Be sure to first consult with a qualified financial adviser and/or tax professional before implementing any strategy discussed herein. Past performance is not indicative of future performance.

As of time of publication, the following securities referenced above were held in one or more VELA strategies: Alphabet, GOOGL (All Cap Concentrated, Large Cap Plus, Large Cap, Income Opportunities); Amazon, AMZN (All Cap Concentrated, Large Cap Plus, Large Cap); Facebook, META (Large Cap Plus, Large Cap); Microsoft, MSFT (Large Cap Plus, Large Cap, Income Opportunities). As of time of publication, no VELA strategies held Oracle, ORCL. The holding(s) identified above do not represent all of the securities purchased, sold, or recommended for the VELA strategies. A complete list of holdings is available upon request. Holdings are subject to change at the portfolio managers’ discretion and without notice. Holdings discussed above are intended to be illustrative in nature and do not constitute a recommendation to buy or sell any particular security.

Intrinsic Value is a measure of what an asset is worth, arrived at by means of an objective calculation or complex financial model. Intrinsic value is different from the current market price of an asset. However, comparing it to that current price can give investors an idea of whether the asset is undervalued or overvalued.

Capital Expenditure (CapEx) is the money a company spends to buy, upgrade, or maintain long-term physical assets—such as property, buildings, equipment, or technology—to expand its operational capacity.

Free Cash Flow (FCF) is the cash a company generates after covering its operating expenses and capital expenditures. It represents the true discretionary cash available to pay debt, issue dividends, or reinvest in growth.

The S&P 500 (Standard & Poor’s 500 Index) is a stock market index that tracks the performance of 500 of the largest publicly traded companies in the United States. It is widely considered one of the best representations of the overall U.S. stock market and the broader American economy.

^1-10 Source: Bloomberg