Speculation, Leverage, and the AI Build-Out
Try this fun experiment. Go to whatever AI tool you use most and ask it a simple question: based on how much I use you, what would you have to charge me to make money?
I did. The answer came back at roughly ten times what I pay today, and that was with an average tech profit margin baked in. Sit with that for a moment, because almost everything I think I know about how badly the world wants AI, I've learned at a price that doesn't come close to covering its cost.
Picture an auto plant where every car costs me $100,000 to build and I sell it for $10,000. Every adult in North America would be lined up around the block, and I couldn't build them fast enough. Demand would look infinite, and it would tell me almost nothing about what the car is worth.
I'll start with the build-out, because it's the part we can actually measure. The four largest hyperscalers plan to spend on the order of $725 billion on capital expenditures in 2026, up about 77 percent from 2025's record $410 billion (with Big Tech's combined spend expected to clear $1 trillion in 2027).1
It's fair to ask whether $725 billion is even a big number. I think the chart says it is. Measured against the size of the economy, the AI build-out already runs about 1.2 percent of U.S. GDP. It has quietly passed the telecom-and-fiber spree of the early 2000s (near 1.0 percent), and ranks among the largest infrastructure waves the country has ever undertaken. On current guidance it climbs toward 2 percent in 2026 and higher still in 2027.2 Only the railroads truly dwarf it, at roughly 5-6 percent of GDP in the 1880s, though that was a narrow, undiversified economy (and the rails they laid kept earning for the better part of a century).
The other great build-out of that era is the more sobering precedent. Electrification went up fast, but it took decades for households and factories to reorganize themselves around the new power before the payoff arrived. That long gap between building the thing and actually using it is one I'll come back to.3 And that gap is sharpened by what today's dollars actually buy: the chips wear out in five years, not fifty. So the honest analogy is probably not the durable railroad. It's the early-2000s telecom gear that went obsolete on a much faster clock.
The fiber comparison is the one to watch, because it's exactly how the dotcom build-out became a graveyard of infrastructure companies. In the late 1990s, carriers buried enormous quantities of fiber on the belief that internet traffic was doubling every three months. It wasn't. Real backbone traffic was doubling closer to once a year, while the technology for pushing light through each strand improved faster still: dense wavelength-division multiplexing let a single fiber carry dozens of signals at once (don't ask me to explain that). So as demand doubled, effective supply expanded tenfold or more.4
The same thing is happening in AI data centers today. Since GPT-4 launched in 2023, the cost of achieving GPT-4-level performance has fallen by an order of magnitude or more, depending on the benchmark and token mix. On some benchmarks, estimates suggest a roughly fifty- to sixtyfold decline (a steeper drop than personal computing or even dotcom-era bandwidth).5
That collapsing cost is the same kind of multiplier: every efficiency gain lets a given pile of GPUs and data centers serve far more demand than its price tag implies. Wonderful if demand is exploding, ruinous the moment supply runs past it.
And here is the trap. The hyperscalers are locked in a prisoner's dilemma. Each treats falling behind in AI as an existential threat, so nobody can afford to be the first to ease off, even as the spending now outruns the cash these businesses generate.6 Nobody wants to be the CEO who blinked. So who, in a race like that, puts on the brakes early enough to avoid the cliff? And who in this whole arrangement has an honest estimate of what the demand is?
The first thing you notice when you go looking for demand is how it behaves when the price is zero. Uber's engineers, spending the company's money rather than their own, consumed the entire 2026 AI-coding budget in four months. As its CTO admitted, "the budget I thought I would need is blown away already."7 That’s our $10,000 car reenacted inside a single company: when the price to the person using it is zero, consumption looks infinite and tells you nothing. The demand signal isn't the blowout; it's the cap. When Uber responded by putting the spend on a meter rather than writing a blank check, it revealed where its willingness to pay actually stops.
Microsoft ran the same arithmetic and simply switched, canceling most of its internal Claude Code licenses by June and steering engineers to a cheaper tool it already owned.8 And here is a test we can grade later: watch Uber's 2027 budget and the fate of that cap. If these tools are producing the returns the boom assumes, next year's number should be dramatically bigger. If the cap survives, that is the sound of demand being honestly priced for the first time.
The deeper problem for most industries is that AI capabilities are advancing faster than organizations can absorb them. Companies must integrate the technology into existing workflows, preserve institutional knowledge, overcome user resistance, handle the weird edge cases, and put sensible controls around all of it. Letting an employee use a chatbot to polish an email is trivial. Handing an autonomous agent the authority to act with no human in the loop is a different order of risk, and companies know it.
Fully self-driving cars were going to make human drivers obsolete a decade ago; but the real world is complex and messy.
None of this means that AI as a technology is weak. Adoption is already widespread, and workers plainly find it useful. The hard part is converting individual productivity into measurable business value.
MIT's Project NANDA, which reviewed more than 300 enterprise AI initiatives in 2025, found that only about 5 percent had produced measurable financial impact or productivity gains.9 The constraint wasn't model capability; it was integration, workflow redesign, and plain organizational learning. The capability is there, but the demand must be built function by function: that's a multiyear project, not a switch to flip.
Klarna is the cautionary tale. After freezing hiring, cutting its workforce from roughly 5,500 to 3,400, and boasting that an AI chatbot was doing the work of 700 customer-service agents, the company watched service quality slide and started putting humans back into support. Its CEO conceded that an excessive focus on cost had produced "lower quality."10 And Klarna may not be an outlier: Gartner expects that by 2027, half of the companies that cut customer-service staff for AI will be rehiring for the same roles, presumably with a good deal less fanfare than the layoffs got.11
This matters financially as much as operationally. If productivity gains or labor savings arrive more slowly than expected (or must be reversed), AI spending becomes an added cost rather than a self-funding investment. Corporate technology budgets are finite. Even the largest companies have to split spending among AI tools, data preparation, cybersecurity, cloud migration, and clean up the legacy systems and technical debt they've been ignoring for years. If AI lifts productivity but doesn't yet remove meaningful costs, adoption adds expense before it subtracts any. Uber, case in point: it underestimated what its engineers' AI tools would cost, found the year's budget gone in four months, topped it back up, and then put the spending on a meter.7 More money, yes, but managed as a cost line, not celebrated as an investment with obvious returns.
There's a second, subtler vulnerability in the demand: A large share of today's model demand is coming from venture capital. AI startups took roughly half of all the venture money raised in 2025, about $210 billion of the year's $425 billion total.12 Then the pace picked up. In the first quarter of 2026, global venture funding hit an all-time record near $300 billion, and AI took $242 billion of it, roughly 80 percent, up from 55 percent a year earlier. That was more in a single quarter than the sector raised in all of 2025.13 Now ask where that money goes. A meaningful share flows right back out the door to rent compute and buy tokens from the model labs. The demand the labs report is subsidized twice. The labs sell below cost, and many of their buyers are paying with investor money rather than earned revenue. Back to our $10,000 car: now imagine that half the people in line are spending other people’s money.
Even with the subsidized demand, the frontier model economics look thin at best. OpenAI's revenue tripled in 2025, from $3.7 billion to $13 billion, one of the fastest ramps in corporate history, and it got more efficient as it grew: expenses per dollar of revenue fell from roughly $3.40 to $2.60. Yet operating losses more than doubled, from roughly $9 billion to about $21 billion, because research and development (overwhelmingly the compute to train the next model), went from $7.8 billion to $19.2 billion, more than the company's entire revenue.
The first quarter of 2026 implies the hole is still deepening. This is the treadmill. Each model generation costs multiples of the last, and no lab can step off without ceding the frontier, because customers defect to whoever ships the best model. I've switched my own favorite model four times. It is the supply side's prisoner's dilemma arriving on the income statement.
Consider how much has to go right from here. Investors must keep funding losses that are still growing. Demand, price, and overall revenue must all keep expanding at a very fast pace, or profitability drifts too far into the future to underwrite.
And the treadmill is only one jaw of the vise. The other is commoditization. It's fair to ask whether a frontier model is protected the way a cutting-edge chip fab is protected. It is next to impossible for a startup to build a TSMC-equivalent plant; the know-how is locked in decades of process engineering, supplier relationships, yield learning, and machines that cannot simply be ordered and assembled into a moat. Frontier models have walls too, but they are far more porous. Once a model is exposed to users, some of its behavior leaks into the world one answer at a time. Fast followers can distill pieces of the leader’s capability into smaller, cheaper models, often running on older hardware, with each new generation standing on the shoulders of the last. The copy does not have to match the frontier model to damage its pricing power; it only has to be close enough, cheap enough, and available soon enough for most customers’ actual use cases. So the lab spends compounding billions to buy an edge that may fade in quarters, then has to spend again. Costs compound while the price per unit of intelligence keeps falling. Scale helps, but it may not be enough if every advance is rapidly copied, compressed, and repriced. The race does not need to end for the technology to succeed; it may need to end for the business model to earn attractive returns.
OpenAI doesn't expect to make money before 2029, and collects a subscription from a grand total of five percent of its users.14 The lone counterexample only sharpens the point. Anthropic, the one frontier lab claiming a profit, says it makes money on every enterprise account from the first dollar. But its headline figure, a projected $559 million of operating profit for the quarter just ended, leans on cheaply rented cloud and is struck before stock-based compensation, to say nothing of what it costs to train the next model.15 Venture money is subsidizing the startups, the model labs are subsidizing every customer, and almost nobody is making a profit, even in a flood of capital. And that is what makes the arrangement fragile. Venture funding is highly cyclical, and we are likely at or near a peak. Before it turns, the real demand must show up. And even if it does, the harder question remains: can the model layer ever get off the commodity treadmill?
None of this means the infrastructure won't be worth having. Like the railroads and the dark fiber before it, the hardware will outlive the companies that built it. In every build-out of this magnitude, the asset survives and most of the capital does not. With more than half the world's venture money chasing a business model exactly one company even claims to run at a profit, the numbers point not to a soft landing but to a washout: a culling of most investors who bought the lottery ticket hoping to be holding the next great winner.
The supply side also comes with a governor the spreadsheets tend to ignore. The whole build-out assumes American communities are on board, and increasingly they are not. In the first quarter of 2026 alone, local opposition blocked or delayed roughly $130 billion of data-center projects (about as much as in all of 2025), and a Gallup poll found that 71% of Americans don't want one built anywhere near them, which may be the closest thing to bipartisan consensus the country has produced in years.16
The opposition is driven by power bills and water, and it's been inflamed by how clumsily the companies have treated their own people: Cloudflare's CEO sorting staff into "builders, sellers, and measurers" and then cut the measurers17; Klarna gutting its support team for a chatbot only to immediately rehire.10
The public mood has turned sharply, and the next step is political. If executives aren't more careful, the coming round of elections and legislation may end up deciding how fast any of this gets built. That, along with the challenges with electricity supply and TSMC's chip output, may be the self-correcting mechanism that keeps the whole thing from getting too crazy. It's one reason the tidy extrapolation to 2.6 percent of GDP by 20272 is anything but a given, and a reason to turn, at last, to the froth all this confidence is already producing.
It's worth remembering that a big share of this capex lands straight in the earnings of the companies selling the picks and shovels, and that's where the cyclicality is easy to overlook. Take Micron, which makes the memory inside the AI server. Its adjusted earnings went from $1.91 a share to $25.11 in a single year, a thirteenfold jump, lifting quarterly profit to a record near $28 billion.18 And Micron isn't alone. Add Nvidia, Broadcom, and the data-center infrastructure and power-equipment names (the Vertivs and Emersons of the world), and you have a group that, by a Goldman Sachs Research estimate, accounts for roughly half of the S&P 500's expected earnings growth this year.19 Those earnings are real, but they aren't necessarily recurring. They're the output of a build-out that, by definition, eventually ends, or at least decelerates, and when it does, the earnings of the companies selling into it can revert just as fast.
Memory has always been one of the most cyclical businesses in technology; its profits move in waves rather than plateaus: the good years give way to years where revenue is down and profits can swing to outright losses. Notice that this is the model layer's problem wearing different clothes. The natural escape from the commodity treadmill is to own the picks and shovels instead: if the models all end up looking the same, buy the companies selling the hardware they run on.
But the hardware is a commodity too, just on a slower clock.
Nothing attracts competition like a fat profit margin. When Micron earns $25 a share in a quarter, rivals break ground on new factories; a couple of years later the new supply shows up, and the price of memory does what the price of every commodity does when supply shows up. So the model builders lose their pricing power to cheap copies, the chip makers lose theirs to the next wave of factories, and somewhere in this stack someone is supposed to durably earn back the largest capital build-out of our lifetimes.
Yet the market is capitalizing this peak as though it were permanent or, more likely, most investors are secure that the wave is still building today and they can surf off when the time comes.
That's the trouble buried in today's index: a full multiple paid on earnings that may themselves be sitting at the top of a cycle. If the capex wave crests, the effect intensifies, as earnings fall and the multiple investors will pay compresses at the same time. Paying a premium for a durable earnings stream is one thing. Paying a premium for the peak of a cycle is another entirely. And as we're about to see, some investors are borrowing to do exactly that.
If you want the clearest evidence that this has become more than a story about the level of infrastructure spend, look at what American investors are doing with borrowed money.
Margin debt, the cash investors borrow against their portfolios to buy still more stock, is flashing at an extreme on two independent measures at once. The first chart shows the level. Margin debt, which FINRA reports monthly, now stands at about 6.0 percent of disposable personal income, a record, and above the peaks of the dot-com top (4.2 percent), the 2007 top (4.0 percent), and 2021 (5.0 percent).20 Income is the right yardstick precisely because it can't be inflated by the bubble itself.
The second chart shows the pace. Margin balances are up roughly 54 percent over the past year, inside the +40 percent zone that has preceded every major top of the last quarter-century: 2000, 2007, and 2021. Record leverage, piled up at a speed that historically shows up only near cycle peaks. Borrowed money cuts both ways, which is exactly what makes it the most telling gauge of speculative fervor we have. Right now it's both stretched and still accelerating. The same impulse is on display in overseas brokerage accounts, in stranger forms still, and that's where we go next.
The leveraged enthusiasm on display in American margin accounts is visible across Asia too, in forms that wouldn't look out of place in 1929. In South Korea, margin debt has climbed to all-time records as retail investors borrow to pile into the very memory-chip names at the heart of the AI trade.21 This spring the country cleared its first leveraged single-stock products, funds engineered to move at twice the daily pace of a single stock. The debut, tied to Samsung and SK Hynix, drew a record 2.4 trillion won, about $1.7 billion. It was the largest fund launch in the country's history.22
Within weeks of listing, the funds had grown big enough to move the very market they track. Their forced daily rebalancing began swinging billions of dollars of Samsung and SK Hynix shares in single sessions. The leverage was no longer just riding the rally; it was driving it. By June, the Bank of Korea was warning that these products were destabilizing trading, and regulators were weighing curbs.23
In Hong Kong, the loudest signal may be the CSOP SK Hynix Daily (2x) Leveraged Product, a fund built from derivatives to deliver twice the daily move of one Korean chipmaker. Within months it swelled to about $13 billion, making it the largest single-stock ETF in the world.24
These are not investments in any ordinary sense. A daily-reset, double-leveraged bet on one stock is designed to be held for a day, not a decade. Regulators warn it's unsuitable for buy-and-hold; its entire purpose is to magnify a short-term wager. When a market's worth of capital pours into instruments like these, the speculative temperature is no longer a matter of opinion.
So, are we in a bubble? I think so. But no one can say for certain; we only ever know in hindsight. If we are in one, I can't tell you how high it goes or when it turns. What I can say is that a genuinely historic wave of capital is being committed to AI infrastructure, while the level of demand stays uncertain until we get a clean pricing signal (and that demand looks like it could undershoot the trajectory of supply).
If supply outruns demand, earnings across the AI complex contract (and not just for whoever showed up last). Once capacity outruns demand, returns compress for every infrastructure player still on the field, early movers and their shareholders included, and the structure underneath isn't built to absorb an air pocket.
Valuations are full; the earnings beneath them are cyclical peaks dressed up as plateaus; sentiment is euphoric; and the whole arrangement is laced with leverage (from record margin debt, to options wagging the cash market, to daily-reset funds that decay rapidly and don’t keep pace with a flat stock).
Every one of those layers is an amplifier. They make the climb feel effortless and the reversal feel violent, because that's what leverage does: turbo charge the way up, brutal on the way down.
Too many people have gone all in on a perfectly orchestrated build-out. Perfection is often improbable and has always been a fragile thing to wager on.
I'm watching for evidence that would change my mind. Four things would do it:
Line up a few of those and I'd conclude the demand is real and arriving on schedule, and the build-out is justified. I'm watching.
None of this is a call to short the top or guess the date; bubbles run further and longer than any skeptic expects, and I have no interest in being early and broke.
It's a case for pulling in the sails.
Capital this concentrated always leaves a great deal behind; sound, unglamorous businesses get ignored, sold, or actively liquidated to fund the mania. That's where the next decade's returns are subtly being set, while others rent 2x exposure to a Korean memory stock.
The move now isn't to play the game better, or to hunt some third-tier beneficiary further out the supply chain before the crowd finds it. It's to step off the field, keep some powder dry, and go looking where no one else is bothering.
Sources
1. Big Tech capital-expenditure forecasts: CNBC, “AI boom: Big Tech capital expenditures now seen topping $1 trillion in 2027,” April 30, 2026. https://www.cnbc.com/2026/04/30/ai-boom-big-tech-capital-expenditures-now-seen-topping-1-trillion-in-2027-.html
2. AI, telecom, and railroad capital spending as a share of U.S. GDP: Michael Allison, Investment Research Partners, “AI Is the New Railroad (Sort of),” Aug. 2025 — drawing on Paul Kedrosky (“Honey, AI Capex Is Eating the Economy”) and the U.S. Bureau of Economic Analysis; early-2000s telecom figure consistent with Federal Reserve Bank of Richmond (Wolman, 2003); 1880s railroad estimate per Ulmer (1960), NBER. https://www.investmentresearchpartners.com/post/chart-of-the-week-8-3-2025
3. Electrification adoption timeline: Brian Potter, “The Grid, Part II: The Golden Age of the Power Industry,” Construction Physics; and “The U.S. Economy in the 1920s,” EH.net. https://www.construction-physics.com/p/the-grid-part-ii-the-golden-age-of
4. Dot-com fiber overbuild and wavelength-division multiplexing: “Dark Fiber — An Archaeology of the Dot-Com Bubble,” Technostatecraft. https://www.technostatecraft.com/p/dark-fiberan-archaeology-of-the-dot
5. Decline in the cost of LLM inference: Guido Appenzeller, “Welcome to LLMflation,” Andreessen Horowitz (a16z), Nov. 2024. https://a16z.com/llmflation-llm-inference-cost/
6. AI data-center spending versus the dot-com fiber buildout, and capex outrunning cash flow: IEEE ComSoc Technology Blog, Sept. 27, 2025. https://techblog.comsoc.org/2025/09/27/big-tech-spending-on-ai-data-centers-and-infrastructure-vs-the-fiber-optic-buildout-during-the-dot-com-boom-bust/
7. Uber’s 2026 AI budget: TechCrunch, “Uber caps employee AI spending after blowing through budget in four months,” June 2, 2026. https://techcrunch.com/2026/06/02/uber-caps-employee-ai-spending-after-blowing-through-budget-in-four-months/
8. Microsoft’s internal Claude Code licenses: Forbes, “Microsoft Ends Claude Code Licenses As It Shifts Developers To Copilot,” June 1, 2026. https://www.forbes.com/sites/jonmarkman/2026/06/01/microsoft-ends-claude-code-licenses-as-it-pushes-copilot-cli/
9. Enterprise AI outcomes: MIT Project NANDA, “The GenAI Divide: State of AI in Business 2025,” 2025. https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf
10. Klarna’s AI reversal and rehiring: Entrepreneur, “Klarna CEO Reverses Course By Hiring More Humans, Not AI,” 2025. https://www.entrepreneur.com/business-news/klarna-ceo-reverses-course-by-hiring-more-humans-not-ai/491396
11. Gartner rehiring prediction: Gartner, “Gartner Predicts Half of Companies That Cut Customer Service Staff Due to AI Will Rehire by 2027,” Feb. 3, 2026. https://www.gartner.com/en/newsroom/press-releases/2026-02-03-gartner-predicts-half-of-companies-that-cut-customer-service-staff-due-to-ai-will-rehire-by-2027
12. 2025 global venture funding and AI’s share: Crunchbase News, “Global Venture Funding In 2025 Surged As Startup Deals And Valuations Set All-Time Records,” 2026. https://news.crunchbase.com/venture/funding-data-third-largest-year-2025/
13. First-quarter 2026 venture funding and AI’s ~80% share: Crunchbase News, “Q1 2026 Shatters Venture Funding Records As AI Boom Pushes Startup Investment To $300B,” 2026. https://news.crunchbase.com/venture/record-breaking-funding-ai-global-q1-2026/
14. OpenAI’s losses and paid-user share: Fortune, “OpenAI’s financials have leaked, showing $21 billion in losses against $13 billion in revenue,” June 16, 2026. https://fortune.com/2026/06/16/openai-financials-leaked-losses-revenue-profit/
15. Anthropic’s first “profitable” quarter (operating profit struck before stock-based compensation): TechCrunch, “Anthropic says it’s about to have its first profitable quarter,” May 20, 2026; and Forbes, “Anthropic And OpenAI Are Taking Opposite Paths To AI Profitability,” May 21, 2026. https://techcrunch.com/2026/05/20/anthropic-says-its-about-to-have-its-first-profitable-quarter/
16. Data-center project cancellations and public opposition: NBC News, “Data center opposition sharply rising in 2026, study finds,” 2026; and Fortune, “A grassroots NIMBY revolt is turning voters in Republican strongholds against the AI data-center boom,” Dec. 16, 2025. https://www.nbcnews.com/tech/tech-news/data-center-opposition-sharply-rising-2026-study-finds-rcna349728
17. Cloudflare layoffs and the “measurers” memo: Fortune, “Cloudflare posted record revenue, then cut 20% of its workforce,” May 21, 2026. https://fortune.com/2026/05/21/cloudflare-ceo-matthew-prince-layoffs-ai-automation-measurers/
18. Micron third-quarter fiscal 2026 results: Micron Technology, Inc., press release (Form 8-K), 2026. https://www.sec.gov/Archives/edgar/data/0000723125/000072312526000013/a2026q3ex991-pressrelease.htm
19. AI beneficiaries’ share of S&P 500 earnings growth: Goldman Sachs Research (Ben Snider), “The S&P 500 Is Forecast to Climb as Earnings Growth Powers Stocks Higher,” May 28, 2026. https://www.goldmansachs.com/insights/articles/s-and-p-500-forecast-to-climb-as-earnings-growth-powers-stocks-higher
20. U.S. margin debt: FINRA, Margin Statistics (customer debit balances), monthly, measured against U.S. disposable personal income (Bureau of Economic Analysis, via FRED). Historical peak ratios use FINRA’s combined all-firm series (NYSE plus NASD member firms, per FINRA’s methodology note), consistent with the current data; this runs above the NYSE-only figures often quoted for the 2000 and 2007 peaks (roughly $300 billion versus $278.5 billion in March 2000, and about $416 billion versus $381.4 billion in July 2007). https://www.finra.org/rules-guidance/key-topics/margin-accounts/margin-statistics
21. South Korean margin-loan records: Seoul Economic Daily, “Korea’s Margin Loans Hit Record 38 Trillion Won as KOSPI Nears 9,000,” 2026. https://en.sedaily.com/markets/2026/06/01/koreas-margin-loans-hit-record-38-trillion-won-as-kospi
22. Korea’s first single-stock leveraged products: The Korea Herald, “Samsung Asset draws record W2.4tr for Korea’s first single-stock leveraged products,” 2026 (funds listed May 27, 2026, per KED Global). https://www.koreaherald.com/article/10756350
23. Market impact and regulatory response to the Korean leveraged ETFs: Bloomberg, “Korea Weighs Curbs on Leveraged Samsung, SK Hynix ETFs as Risks Rise,” June 22, 2026; and “Bank of Korea warns single-stock leveraged ETFs are rattling markets,” Cryptobriefing, 2026. https://www.bloomberg.com/news/articles/2026-06-22/korea-mulls-steps-to-rein-in-leveraged-samsung-sk-hynix-etfs
24. CSOP SK Hynix Daily (2x) Leveraged Product: KuCoin, “Hong Kong SK Hynix 2x Leveraged ETF Becomes Fastest-Growing in Asia, Fourth Fastest Globally,” 2026.
IMPORTANT DISCLOSURES
The information and statistics contained in this communication have been obtained from sources we believe to be reliable but cannot be guaranteed. Opinions and statements of financial market trends that are based on market conditions constitute our judgment and are subject to change without notice. Any projections, market outlooks or forecasts discussed herein are forward-looking statements and are based upon certain assumptions. Other events that were not taken into account may occur and may significantly affect the returns or performance of these investments. Any projections, outlooks or assumptions should not be construed to be indicative of the actual events which will occur. Please remember that past performance may not be indicative of future results.
The objective of the City Different SMID Cap Core strategy is to generate strong risk-adjusted returns in a focused portfolio of small and mid-cap stocks invested across the business life cycle. Portfolios invest primarily in U.S. stocks with greater than $100 million in market cap. Portfolios typically hold between 20-35 positions, with no position representing more than 10% of the portfolio at time of purchase. Not less than 80% of portfolio assets will be invested in companies whose market cap is within the market cap range of the Russell 2500 Index. The strategy may take on international exposure up to 25% of portfolio assets. An investment in the strategy is subject to certain risks. The value of an investment may fall as well as rise and is not guaranteed. Equities may decline in value due to both real and perceived general market, economic and industry conditions.
The Russell 2500 Index measures the performance of the small to midcap segment of the US equity universe, commonly referred to as "smid" cap. The Russell 2500™ Index is a subset of the Russell 3000® Index. It includes approximately 2500 of the smallest securities based on a combination of their market cap and current index membership.
Different types of investments involve varying degrees of risk, and there can be no assurance that the future performance of any specific investment, investment strategy, or product, or any non-investment related content, made reference to directly or indirectly in this communication will be profitable, equal any corresponding indicated historical performance level(s), be suitable for your portfolio or individual situation or prove successful. Due to various factors, including changing market conditions and/or applicable laws, the content may no longer be reflective of current opinions or positions. No discussion or information contained herein serves as the provision of, or as a substitute for, personalized investment advice. To the extent that a reader has any questions regarding the applicability above to his/her individual situation of any specific issue discussed, he/she is encouraged to consult with the professional advisor of his/her choosing. City Different Investments is neither a law firm nor a certified public accounting firm and no portion of this content should be construed as legal, tax, or accounting advice.