One A.I. perspective for relatively modest, long-term investments over the next 5-10 years (UII w/Claude)

sciart0
May 19
23 min read

Good day to you.

In the past investments which were based upon "the fundamentals" presented a strong possibility of reliable returns with minimal risks over extended periods (5-10 years).

However, AI's commercial and societal (destructive?) disruptions, along with other macro factors, seem to potentially generate serious concerns, even emergent flaws, in that premise or paradigm.

Please offer your (and informed other's') perspectives, concerns and predictions please.

Good day.

This is a substantial and genuinely important question — one where the answer depends considerably on a few things I'd like to nail down before diving in:

When you say "fundamentals-based investing," which lens matters most to you — traditional equity valuation (P/E, DCF, moats), macroeconomic fundamentals (interest rates, productivity cycles), or both?
Are you most interested in the threat AI poses to previously "safe" sectors (retail, finance, media, healthcare, etc.), the distortion it creates in valuation models generally, or both?
What's the primary purpose — personal investment decision-making, broader intellectual/strategic analysis, or framing for the 4Chairs/Relations Research context?

All of the above (including your suboptimal, "conclusion-engine" subsets of dichotomies).

Fair enough — and noted on the dichotomies. Your "all of the above" is unambiguous, so I'll launch into a thorough research dive across all three dimensions: the disruption AI poses to traditional valuation frameworks, the sector-specific destruction of formerly durable moats, and the macroeconomic distortions that complicate 5–10 year fundamental assumptions. I'll pull in both bullish and deeply skeptical perspectives:

Lead Researcher

AI disruption of fundamentals-based investing

Research complete•

Writing and citing report...

257 sources

14m 54s

Your report on whether AI is breaking fundamentals-based long-term investing is ready. It examines valuation model disruption through Damodaran's DCF re-ratings, the circular financing architecture connecting Nvidia, OpenAI, Oracle, AMD, and Microsoft, and sector-specific moat destruction across edtech, SaaS, publishing, legal, and financial advice. The analysis weighs bearish perspectives from Grantham, Marks, Burry, and Shiller against bullish cases from Wood and Huang, covers macro productivity dispersion from Acemoglu to ARK, and identifies six fundamentals categories that still work. It concludes with seven staged portfolio recommendations and seven caveats on timing, historical analogies, and contested data points.

AI and Fundamentals-Based Investing: Valuation, Moats, and Risk in the 5-10 Year Horizon

Document (This document is available without charge by e-mail request to me.)

Thanks.

Now summarize please, including guidance for the average investor of $100,000-$1,000,000 who does not require liquidity within 5-10 years.

Synthesized investment guidance for mid-range portfolios with extended time horizons

The Core Situation, Plain and Direct

The premise that "invest in solid fundamentals and wait" reliably produces returns over 5–10 years is not broken — but it has been materially complicated. The complication is not that AI is fake or that markets are irrational. It's that AI has introduced genuine uncertainty into the inputs that fundamentals-based analysis depends on: how long a competitive moat lasts, what a realistic growth rate looks like, whether a company's revenue is really earned from end customers or from a circular chain of other companies investing in each other. The framework still works; the confidence intervals have widened dramatically.

The single most important thing to understand is this: the AI investment world is currently running on a gap between commitments and revenues that has no modern precedent. Goldman Sachs estimates the major hyperscalers — Microsoft, Alphabet, Amazon, Meta, Oracle — spent roughly $400 billion on AI infrastructure in 2025 and are on track for $700 billion in 2026. OpenAI, the most visible beneficiary of all that spending, generated about $13 billion in revenue in 2025.

The companies building the infrastructure are, to a measurable degree, buying from each other. Nvidia has a letter of intent to invest $100 billion in OpenAI; OpenAI has committed $300 billion to Oracle; Oracle's order backlog jumped 359% in one year. Howard Marks at Oaktree Capital estimates roughly 15% of Nvidia's 2026 revenues come from these circular arrangements. This is not fraud — the technology is real — but it means a significant portion of what looks like revenue is actually deferred capex dressed as sales.

What Has Actually Been Disrupted

Several categories of investment that looked like "solid fundamentals" five years ago have had their moats genuinely eroded. Understanding which ones matters more than knowing the headline.

Edtech built on information access is already gone. Chegg, once a $14 billion company, is now worth roughly $100 million — a 99% collapse driven entirely by students using AI instead of Chegg's tutoring and homework-help services. The moat was "we have collected the answers." AI made that moat worthless overnight.

Open-web publishing built on Google search traffic is in the early stages of the same collapse. News publishers report search referrals falling 30–40% in a single year as Google's AI-generated summaries replace clicks. This affects any media or content company whose business model depends on advertising supported by organic search discovery.

Traditional per-seat SaaS software — meaning software that charges per employee per month for tasks that are probabilistic and content-driven — is under structural siege. Salesforce dropped 28% and Atlassian dropped 35% in early 2026 as investors re-rated the durability of that model. IDC projects that pure seat-based pricing will be obsolete for 70% of software vendors by 2028.

The Big Law leverage pyramid — in which junior associates bill at high rates to fund partner draws — is being compressed, though not destroyed. AI automates the work that justified hiring large cohorts of young lawyers. Revenue at top law firms actually grew 13% in 2025, but that growth disguises the compression of junior hiring and the beginnings of a structural shift in how legal value is priced.

By contrast, companies with moats rooted in regulation, physical scarcity, brand desire, or deeply embedded deterministic software (the kind where a mistake costs a patient their life or a bank its license) are holding up well and may actually benefit.

The Macro Uncertainty That Should Concern a Long-Term Investor

The honest answer about AI's effect on economic productivity over your 5–10 year horizon is that the experts disagree by an order of magnitude. MIT economist Daron Acemoglu's careful task-based modeling suggests AI will add no more than about 0.66% to total factor productivity over the next decade — a meaningful but modest boost. Goldman Sachs originally projected a 7% GDP uplift. ARK Invest's Cathie Wood projects sustained real GDP growth above 5%. These are not minor differences in emphasis; they are fundamentally incompatible predictions. Any investment model built on a specific productivity assumption is therefore carrying more uncertainty than it acknowledges.

What this means practically is that the classic 5–10 year DCF — where you project cash flows, apply a discount rate, and arrive at a value — still works as a framework, but the output should be treated as a range of scenarios rather than a number. Investors who are not doing that right now are more confident than the data warrants.

What the Sharpest Minds Are Actually Doing

This is worth knowing, not to copy them, but to understand the weight of informed concern.

Aswath Damodaran, perhaps the world's leading expert on equity valuation, has moved money into cash and collectibles for the first time in his career. He cut his Nvidia position in January 2025 after calculating that the stock's then-market price of $123 implied $678 billion in eventual annual revenues at 53% margins — figures he calls "possible but not probable."
Michael Burry, who correctly called the 2008 housing crash, has deregistered his fund to go fully private and is running large short positions against Nvidia and Palantir while drawing an explicit parallel to Cisco in 1999 (Cisco lost 80% of its value after 2000, recovered only 20 years later, even though the internet itself was completely real).
Jeremy Grantham calls the probability that this AI bubble does not eventually burst "slim to none." Howard Marks — the most measured of them — puts the odds that "this time is actually different" at roughly 20%.

On the other side,

Cathie Wood argues AI will create $13 trillion in software opportunity alone and that hyperscaler capital expenditure is rational.
Jensen Huang argues Nvidia has reinvented computing for the first time in sixty years. The IMF notes — and this is an important counterpoint — that AI capex is mostly equity-funded by cash-rich companies rather than debt-funded as in the telecom bubble of 1999, which means a correction would be painful but probably not systemically destabilizing in the way the housing crisis was.

Practical Guidance for Your Scenario

Given a 5–10 year horizon, $100,000–$1,000,000, and no liquidity need, here is what the weight of the evidence suggests. These are not certainties; they are probability-weighted positions.

The first priority is reducing your passive exposure to the concentration problem. If your core portfolio is an S&P 500 index fund, roughly 35% of every dollar you invest is currently going into seven companies, five of which are deeply entangled in AI capital commitments. This is the highest index concentration since the dot-com peak. A reasonable adjustment is to shift a meaningful portion of your domestic equity exposure to equal-weighted S&P 500 funds, mid-cap indices (S&P 400), or international developed markets, which have outperformed U.S. large-cap by about 10% in the past year and trade at far lower valuations. This is not market timing; it is restoring the diversification that a cap-weighted index no longer provides.

The second priority is auditing your existing holdings for the three patterns of AI moat destruction described above: per-seat SaaS with probabilistic cores, content businesses dependent on search traffic, and any leverage model built on entry-level cognitive labor. If you hold individual stocks or sector ETFs in these categories, the structural case for trimming is stronger than it was two years ago, independent of what the stock has done recently.

The third priority is identifying where genuine scarcity is being created by AI rather than destroyed. Power generation and transmission infrastructure — particularly utilities with data-center contracts or nuclear power agreements — faces a decade-scale demand surge that no AI model can substitute. Advanced semiconductor manufacturing (TSMC specifically, not just Nvidia) holds a physical bottleneck that cannot be easily replicated. Companies with large proprietary datasets that are legally cleared and continuously refreshed are increasingly valuable as foundation AI models commoditize. These categories have long-duration moats that are strengthened by the AI buildout regardless of whether the AI bubble inflates or pops.

The fourth and most conceptually important priority is building a margin of safety into your 5–10 year plan that accounts for a Grantham-style drawdown scenario. The historical pattern — railroads in the 1860s, radio in the 1920s, internet in the 1990s — is that transformative technologies deliver on their promise but that leading companies draw down 50–90% before the durable winners emerge. Amazon fell 92% after 2000 and then became one of the greatest wealth generators in history. If you own AI-exposed equities, the question is not whether the technology works but whether you can hold through a 60% drawdown without making a forced decision. A cash or short-duration bond allocation sized to your psychological tolerance for that scenario is not pessimism — it is optionality on the most likely outcome.

The one thing you should not do is make a binary bet in either direction — either "AI will transform everything so I'm all-in on Nvidia and the Mag 7" or "this is a bubble so I'm moving to cash." Both positions require a timing precision that no one, including the sharpest contrarians, has demonstrated.

The synthesis position that the evidence most strongly supports is: the technology is real, some companies will generate extraordinary value, the current valuations of the leading names price in an unrealistically narrow set of outcomes, and diversifying away from that concentration while maintaining long-duration exposure to physical AI infrastructure and AI-resistant moats is the most defensible posture for your horizon.

Have you baked-in potential societal disruptions (such as not only AI effects to lives, labor, thus marketplaces, and geopolitical factors, but to applicable advancements of tech, climate and black swans)?

If not, please do so.

Synthesized multifaceted disruptions across labor, geopolitics, technology, climate, and contingencies

The short answer is: not fully.

The prior analysis was weighted heavily toward AI's direct financial and sectoral effects. What follows bakes in the broader disruption landscape — and the honest conclusion is that doing so makes the investment environment more uncertain, not less, while also revealing some underappreciated durable opportunities.

The most useful way to think about this is in layers.

AI is the most visible disruptor right now, but it is operating inside a larger system that is itself being destabilized along several axes simultaneously. When multiple destabilizing forces interact, the emergent outcomes are genuinely harder to predict than any single-variable model suggests.

That is not a reason for paralysis — it is a reason for a different kind of portfolio architecture than the one most long-term investors currently hold.

The Societal Disruption Layer AI Models Usually Miss

The direct labor-market effects of AI are already partially priced. What is not priced is the political economy feedbackthat follows from those effects. History is instructive here. The first Industrial Revolution genuinely increased aggregate wealth, but it also produced forty years of social dislocation, child labor, urban immiseration, and eventually the labor movement, antitrust law, progressive taxation, and the regulatory state. Each of those responses materially altered the investment landscape in ways that would have been nearly impossible to model from the vantage point of 1820.

We are seeing early-stage versions of the same dynamic. Yale School of Management research published in early 2026 documents a 16% decline in early-career employment in AI-exposed occupations since late 2022, with software developer entry-level postings down roughly 35% since January 2023. Anthropic's Dario Amodei has publicly projected that AI could eliminate approximately 50% of entry-level white-collar jobs within five years. These are not fringe estimates — they represent mainstream projections from people who build the technology.

The investment-relevant consequence is not primarily the direct labor-market disruption itself. It is the political response to it. When significant portions of the workforce — particularly younger, educated workers who vote, organize, and form cultural narratives — experience downward economic mobility, the historical pattern is regulatory retaliation, redistributive taxation, antitrust action, and in some cases protectionist trade policy. Any one of these can materially alter the return assumptions for the companies currently priced as permanent monopolists.

The Magnificent Seven are already under antitrust scrutiny in the U.S. and Europe; that scrutiny intensifies as the labor argument becomes more politically salient. A structural windfall profits tax on AI companies, forced licensing of model weights, or mandatory revenue-sharing with displaced workers — none of these are currently in most 5–10 year investment models, and all have historical precedent.

For the investor with your horizon, this means that the political durability of AI's profit margins is a genuine unknown. Companies that depend on regulatory permissiveness — that is, most of the large AI labs — are carrying political risk that their current valuations do not reflect.

Geopolitical Fracturing as a Portfolio Variable

The prior analysis treated the AI investment landscape as essentially a U.S.-centric story with some mention of international diversification. That framing is increasingly inadequate.

The most significant geopolitical variable is the U.S.-China semiconductor and AI competition, which has moved well beyond trade policy into the architecture of the global technology supply chain. U.S. export controls on advanced chips to China — specifically Nvidia's H100 and H20 series — have bifurcated the global AI market in ways that are still playing out.

China is investing massively in domestic semiconductor manufacturing through SMIC and others, and while current Chinese chips remain one to two generations behind TSMC, the gap is narrowing. The relevant investment implication is that TSMC's current bottleneck monopoly on advanced fabrication is a time-limited moat, not a permanent one. Its durability depends on U.S. political will to maintain export controls, Taiwan's continued political stability, and TSMC's own ability to execute on its Arizona expansion — all of which involve genuine uncertainty.

More broadly, the world is in the early stages of a supply-chain deglobalization that began with COVID, accelerated through the Ukraine war, and is now being institutionalized through industrial policy on multiple continents. The Inflation Reduction Act, the CHIPS Act, the EU's AI Act, and China's Made in China 2025 successor policies are all expressions of the same underlying dynamic: governments are willing to sacrifice economic efficiency for supply-chain resilience and technological sovereignty. For a long-term investor, this means that the era of maximally efficient global supply chains — which underpinned the profit margins of many S&P 500 companies for three decades — is not returning. Any investment model that projects pre-2020 margin structures for globally integrated manufacturers is using a broken input.

The Taiwan question deserves explicit mention rather than being treated as a tail risk. Taiwan produces roughly 92% of the world's most advanced semiconductors. A Chinese military action against Taiwan — whether invasion, blockade, or coercive disruption — would not merely be a humanitarian catastrophe; it would instantaneously destroy the physical substrate on which the entire AI buildout depends. This is the most severe black swan in the investment universe right now, and it is underpriced in virtually every equity model because markets systematically underprice low-probability, high-consequence events until they are imminent. The investor who holds significant AI infrastructure exposure without a hedge against this scenario is, whether knowingly or not, taking a large geopolitical bet.

Climate as a Long-Duration Investment Variable, Not a Political One

Climate disruption is frequently treated as a political topic, which causes investors to either over-index or under-index on it based on their priors rather than the evidence. The relevant framing for a 5–10 year investor is not "do I believe in climate change" but rather "what physical and regulatory changes are already locked in, and how do they affect the assets I hold."

Several consequences are now essentially certain within your investment horizon regardless of political outcomes. Insurance repricing is already underway. Major insurers — State Farm, Allstate, and others — have withdrawn from or drastically curtailed coverage in California, Florida, and parts of the Gulf Coast. This is not a future risk; it is a current one that is repricing real estate, affecting commercial lending, and beginning to impair the collateral values that back a significant portion of bank balance sheets. The Federal Reserve has explicitly incorporated climate-related financial risk into its bank stress-testing framework. An investor holding bank stocks or real-estate-heavy REITs in high-climate-exposure geographies is carrying a risk that was not in the standard model five years ago.

Water scarcity is also already repricing agricultural land, affecting supply chains for semiconductor manufacturing (which is extraordinarily water-intensive — TSMC's Arizona plant is facing water permit challenges), and creating operational risk for data centers in the American Southwest and Southeast. The AI buildout, ironically, is itself a significant climate stress: the IEA projects that AI data centers will consume roughly 1,000 terawatt-hours of electricity globally by 2026, equivalent to Germany's entire consumption.

That demand is running into a power grid that was not designed for it. Microsoft has disclosed that its data-center construction is constrained by power availability, not capital. Google and Amazon have made large nuclear power purchase agreements specifically because utility-scale renewables cannot be deployed quickly enough. For the investor, this creates a legitimate long-duration opportunity in energy infrastructure — particularly nuclear, grid modernization, and electrical transmission — that is genuinely durable regardless of which AI companies ultimately win the model race.

Finally, climate-driven physical disruption is beginning to impair supply chains in ways that affect corporate earnings in sectors far removed from what most people think of as "climate stocks." The 2024 Panama Canal drought reduced shipping capacity by roughly 30% for several months, affecting everything from consumer goods to chemicals to agricultural commodities. Events like this — which will become more frequent — introduce supply-chain volatility that standard investment models treat as random noise but which is increasingly patterned. Companies with resilient, regionalized supply chains and strong balance sheets to absorb disruption without distress will carry a structural advantage over the coming decade that is not yet reflected in most valuations.

Technological Acceleration Beyond AI

AI is the most visible technology disruption, but it is running alongside several others that will interact with it in ways that are difficult to model independently but important to hold as a cluster.

Biotechnology is in the early stages of a productivity revolution driven by AI-accelerated drug discovery, protein folding (AlphaFold's impact on biology is roughly analogous to what the internet did to information access), and increasingly by CRISPR-based therapeutics moving from lab to clinic. The FDA approved its first CRISPR therapy in late 2023; by your 5–10 year horizon, gene editing for common diseases including some cancers, sickle cell disease, and potentially cardiovascular conditions will likely be clinically routine. The investment implications cut in multiple directions: biotech companies with AI-accelerated pipelines represent a genuine long-duration opportunity; pharmaceutical companies whose competitive advantage rests on traditional drug discovery timelines face structural compression of their moats; and health insurance companies face significant uncertainty about how dramatically their actuarial models may need to be revised as treatment efficacy improves.

Energy technology — specifically the trajectory of battery storage, grid-scale renewables, and the nascent hydrogen economy — is moving faster than most infrastructure investment models assume. The cost of lithium-ion battery storage fell roughly 90% between 2010 and 2023; a similar trajectory over the next decade makes utility-scale storage economically viable in ways that would fundamentally alter the business model of fossil fuel peaker plants and eventually base load. This does not mean oil and gas are uninvestable in your horizon — demand will remain significant through 2030 — but it does mean that the terminal value assumptions in traditional energy company DCF models deserve explicit scrutiny.

Quantum computing remains earlier-stage but warrants mention because its eventual arrival — and the current best estimates cluster around the early 2030s for cryptographically relevant quantum capability — would simultaneously break most of the encryption infrastructure that underlies modern financial transactions, communications, and data security. This is not a 5-year risk for most investors, but it is inside the extended tail of a 10-year horizon, and governments and financial institutions are already investing in post-quantum cryptography standards. Any investment in a company whose competitive advantage depends on proprietary encrypted data or cryptographic security should be stress-tested against this.

Black Swans: The Ones Worth Naming

The term "black swan" is overused to mean "bad surprise," but Nassim Taleb's original meaning is more specific: events that are outside normal expectation, carry extreme impact, and are retrospectively rationalized as predictable. The truly unknowable ones cannot be listed by definition. But there are several events that are currently underpriced relative to their objective probability.

A global pandemic on the scale of COVID or worse is, according to epidemiologists, not a tail event but a periodic feature of a densely connected world with significant human-wildlife interface. COVID revealed that standard investment models had essentially zero pandemic risk priced into them despite a century of evidence that pandemics occur. The next one could interact with AI in complex ways — potentially accelerating vaccine development (a genuine positive) while also accelerating the spread of misinformation and compressing the time window for coordinated response.

A major cyberattack on financial infrastructure — specifically, coordinated attacks on payment settlement systems, cloud providers, or the power grid — has been explicitly flagged by the Bank of England, the Federal Reserve, and CISA as a top-tier systemic risk. The concentration of financial infrastructure on a small number of cloud providers (AWS, Azure, Google Cloud host roughly 65% of global enterprise workloads) means that a sophisticated state actor could cause more economic disruption through a cyberattack than through any conventional military action short of nuclear. This risk is rising, not falling, as AI lowers the barrier to entry for sophisticated cyber offense.

Social and political fragmentation in major democracies — driven partly by economic inequality, partly by algorithmic radicalization, and partly by the labor displacement described above — represents a genuine risk to the institutional stability that long-term investment returns depend on. Property rights, contract enforcement, currency stability, and the rule of law are not self-maintaining systems; they require political legitimacy. That legitimacy is under more strain in more countries simultaneously than at any point since the 1930s. This is not a prediction of collapse, but it is a reason to hold more of your portfolio in assets that retain value across political regime changes — specifically hard assets, internationally diversified equity, and short-duration instruments.

What This Changes in the Practical Guidance

Layering all of this onto the AI-specific analysis from before, the practical guidance for someone in your position shifts in five specific ways:

First, geographic diversification is more important than it has been at any point in recent memory. The prior guidance to tilt toward international developed markets becomes more urgent when you factor in geopolitical fragmentation, Taiwan risk, and the political backlash dynamics. International developed markets — Europe, Japan, South Korea, Australia — trade at roughly half the valuation multiple of U.S. large-cap equities, carry meaningfully different geopolitical risk profiles, and give you exposure to economies that are earlier in their AI adoption curve (which may mean more upside than downside in your specific time window). Emerging markets require more selectivity, but India specifically represents a demographic and technological trajectory that is increasingly decoupled from China risk and represents one of the more defensible long-duration equity stories.

Second, real assets — broadly defined — deserve a larger allocation than conventional modern portfolio theory suggests. This means energy infrastructure, water infrastructure, agricultural land or related equities, gold as a currency-risk hedge, and potentially commodities tied to AI's physical requirements (copper for electrification, uranium for nuclear, rare earths for advanced magnets in wind turbines and EV motors). These are not exciting investments. They are durable ones. They retain value through inflation, political disruption, and technology cycles in ways that financial assets do not.

Third, your cash or short-duration position should be thought of not as "waiting for a better opportunity" but as optionality insurance. The scenario where multiple disruptions interact — a geopolitical shock compressing AI valuations at the same time a climate event impairs supply chains, for example — would create buying opportunities in genuinely excellent businesses at dramatically lower prices. Having 15–25% in cash or short-duration Treasuries means you can act as a buyer rather than a forced seller in that scenario.

Fourth, the specific AI-resilient categories identified earlier — physical power infrastructure, rights-cleared proprietary data, regulatory-moated businesses — gain additional conviction when you layer in the non-AI disruption factors. Power infrastructure benefits from both AI demand and climate-driven grid modernization. Businesses with genuine pricing power protected by regulation — utilities, essential pharma, payment networks — are resilient across all the disruption scenarios described here, not just the AI-specific ones. They are boring. They compound.

Fifth and finally: the 5–10 year horizon you have is actually an advantage in this environment, not a burden. Most institutional investors are constrained by quarterly performance reporting and career risk, which pushes them toward consensus positions that are already priced. The scenario that the evidence most supports — genuine technology transformation, significant interim valuation correction, eventual emergence of durable winners at much lower prices — is precisely the scenario that rewards the patient, diversified, margin-of-safety investor who does not need to be right about timing. Your job is not to predict which disruptions materialize. It is to hold a portfolio that survives all of them and captures the upside of whichever ones create lasting value.

The synthesis position, with all of this baked in, is this: the world is not ending, but the era of simple index-fund-and-wait investing producing predictable 7–10% annual returns with low variance is likely behind us for at least this cycle. The investor who understands the disruption landscape, diversifies genuinely across geographies and asset classes, holds real assets as a hedge against political and climate risk, maintains liquidity as optionality, and avoids the specific categories where moats have been structurally broken will, on a 10-year view, almost certainly outperform the investor who does none of those things. What you cannot do is eliminate uncertainty. What you can do is build a portfolio that is not destroyed by it.

So, in sum, your valence is for directing attention to bias weighted to defensive posturing, versus offensive actions, for the forthcoming decade?

If so, or in not, would mapping this to equalling inflation be too conservative as an objective; or if not, what would you as an optimal annual ROI target summation metric?

The characterization is fair but needs one important refinement. The guidance is not weighted toward defense instead ofoffense — it is weighted toward building the platform from which offense becomes viable. Those are structurally different postures, and conflating them is the most common mistake investors make when environments become uncertain. Pure defensiveness — matching inflation and nothing more — is its own form of risk, because it forfeits the compounding advantage that your 5–10 year horizon specifically grants you. A genuinely patient investor with no liquidity requirement is holding one of the rarest assets in the market: time. Squandering that on pure capital preservation is a strategic error.

So the more precise framing is this: the guidance is for sequenced offense — building the defensive architecture first (genuine diversification, real assets, liquidity optionality, exit from structurally broken moats) specifically so that you have the capacity to take concentrated offensive positions when the disruption scenarios described create the buying opportunities that patient capital is uniquely positioned to exploit. Marks, Damodaran, and Grantham are not saying "go to cash and stay there." They are saying "build the capacity to be a buyer when others are forced sellers." That is offense, not defense. It just requires discipline in the setup phase.

Now to your ROI question, which deserves a more rigorous treatment than most advisors give it.

Targeting inflation-matching as your primary objective would indeed be too conservative for your situation, and here is precisely why. Inflation matching — currently running around 3–4% in the U.S. depending on the measure — can be achieved almost entirely through short-duration Treasury Inflation-Protected Securities (TIPS) or I-bonds, with essentially zero equity risk. If that is your target, you have no reason to hold equities at all, and certainly no reason to engage with the complexity of the disruption landscape we have been mapping. The very fact that you have a 5–10 year horizon with no liquidity requirement means that accepting equity volatility in exchange for real return above inflation is not just reasonable — it is the rational expression of the advantage your situation provides.

The more useful question is what a realistic and defensible target looks like given the specific environment described, as opposed to the targets that were reasonable in the 2010–2020 period of low rates, stable globalization, and mean-reverting markets.

The conventional long-term equity return assumption — roughly 7% nominal, 4–5% real after inflation — was built on a specific set of conditions: U.S. equity starting valuations in the moderate range (CAPE of 15–20), globally integrated supply chains expanding profit margins, a multi-decade tailwind from falling interest rates, and relatively predictable geopolitical conditions. Most of those tailwinds have either reversed or are no longer reliable.

Starting from a CAPE above 40 on U.S. large-cap equities, the historical evidence is unambiguous: 10-year forward real returns from that starting point have averaged roughly 0–2% annually, not 4–5%. This is not a prediction; it is an empirical regularity across every instance of similar starting valuations in recorded market history.

That does not mean 0–2% real is what you should target or expect from a thoughtfully constructed portfolio. It means that a cap-weighted S&P 500 index fund held passively for 10 years from today's valuations is more likely to deliver in that range than in the historical average range. A more intelligently constructed portfolio — one that applies the framework we have developed — should be able to do meaningfully better.

Here is how to think about what is achievable and at what level of risk, working from the ground up rather than from a top-down target.

The real-asset and infrastructure layer — power infrastructure, grid modernization, nuclear, physical AI bottlenecks — has historically delivered 6–8% nominal total return with significantly lower volatility than equities and meaningful inflation protection, because these assets have pricing tied to regulated rate cases, long-term contracts, or commodity prices that move with inflation. In the current environment, with a decade-scale demand surge from AI and climate-driven grid investment, there is a credible case for this layer outperforming its historical average. Call this your base layer: targeting 6–8% nominal, 3–5% real, with moderate volatility.

The international developed equity layer — Europe, Japan, South Korea — trades at CAPE ratios of roughly 15–20, close to historical averages, with dividend yields of 3–4% already embedded. Starting valuations at these levels have historically produced 7–9% nominal 10-year returns without needing heroic assumptions about growth. This layer adds both return potential and genuine diversification against U.S.-specific political and valuation risk.

The opportunistic offense layer is where your 5–10 year horizon creates genuine alpha potential that shorter-horizon investors cannot access. This is the allocation — perhaps 15–25% of the portfolio — held in cash or short-duration instruments now, not as a permanent allocation but as deployed capital waiting for the correction scenario that the weight of evidence suggests is more probable than not. If AI valuations correct 40–60% from current levels over the next 2–4 years — as Grantham and Damodaran both consider the base case — and you are a net buyer at those levels into the companies with genuine durable moats (the Amazons of the next cycle, not the Ciscos), the return on that deployed capital over the remaining horizon is historically very high. Amazon from its 2002 low to 2010 returned roughly 2,000%. You do not need that kind of outcome to justify the strategy; even a modest version of it materially improves your portfolio-level return.

Putting these layers together, a reasonable target — meaning an outcome that is achievable without requiring either the bull case or the bear case to fully materialize — is a real return of 4–6% annually above inflation, which translates to roughly 7–9% nominal if inflation stabilizes around 3%. This is modestly below the naive historical S&P 500 average but meaningfully above both inflation-matching and the likely forward return of passive cap-weighted indexing from current valuations. Importantly, it is achievable across a range of disruption scenarios rather than dependent on any single one resolving favorably.

Where this target becomes vulnerable — and intellectual honesty requires naming it — is in two tail scenarios.

The first is Cathie Wood's scenario: AI delivers the full productivity transformation, valuations hold or expand, and passive index investing continues to compound at 12–15% annually. In that case, the defensive architecture described here would underperform a simple index fund, potentially by several percentage points per year. The cost of the insurance would have been real.
The second is Grantham's full bear case plus geopolitical shock: a 50–70% drawdown in U.S. equities, extended recovery period, dollar weakness, and sustained inflation. In that case, even the 4–6% real target would require active management of the opportunistic layer to achieve.

The honest answer to your question about optimal ROI target is therefore this: 4–6% real annually is the right planning target for a thoughtfully constructed, genuinely diversified portfolio over your horizon — not as a floor to merely match, and not as a ceiling to timidly stay within, but as the central estimate around which you build your architecture with the explicit understanding that upside capture in the correction scenario could push it meaningfully higher. Targeting inflation-matching alone would be abandoning the advantage your situation provides. Targeting 10%+ real annually would require assumptions about the future that the disruption landscape we have mapped does not support as the base case.

The most important reframe, though, is to resist letting the ROI number become the organizing principle of the decision-making. The number is an output of good architecture, not an input to it. Investors who start with a target return and work backward to justify the risk required to achieve it — a pattern Marks specifically warns against — are the ones who end up holding the concentrated AI positions that are most vulnerable to the correction. The investor who starts with the disruption landscape, builds genuine resilience into the portfolio, maintains the capacity for opportunistic offense, and then measures the result will almost certainly find that 4–6% real is a conservative description of what they actually achieve.