There is a particular kind of market anxiety that comes from watching a rally that feels slightly untethered from reality, while also fearing that you might become the person who sold Apple too early and then spent the next decade explaining your "discipline."
For the past year, that anxiety has had a very fashionable name: the AI bubble.
So I did what I always do when I don't trust myself to think clearly about something. I deferred to people who are smarter, more experienced, and less emotionally invested than I am. I read Damodaran, Dalio, Munger, Graham, Marks, Ackman, Bogle. And then I made some decisions.
Here is what I found — and what I'm doing about it.
1. The Basics Are Still the Basics
Before AI, before bubbles, before Nvidia screenshots on Twitter, investing still starts with the same old boring foundation.
And boring foundations are underrated.
Benjamin Graham's core idea remains the anchor: a stock is ownership in a real business. That business has an intrinsic value based on earnings power, balance sheet strength, cash flows, and the price you pay for those cash flows. Mr. Market may become euphoric, depressed, theatrical, or completely deranged. The business is still the business.
That is the entire point of value investing. You are not buying a story. You are buying a claim on future cash flows.
This matters especially in AI, because the story is intoxicating. "AI will change everything" may be true. It also says very little about whether a specific stock is attractive at a specific price.
Damodaran's framework is useful here: every story needs numbers. A good narrative still has to survive valuation. Growth, margins, reinvestment, risk, and discount rates are where the romance meets the spreadsheet.
Dalio adds another uncomfortable reminder: discount rates are not a side detail. They are the whole game. When interest rates rise, future cash flows become worth less today. This hurts long-duration growth stocks most — exactly the kind of stocks that dominate AI enthusiasm.
AI can be structurally real. And AI stocks can still be overpriced. Both things can be true at the same time.
Annoying, but investing is often annoying.
2. The S&P 500 Is Already an AI Bet
My first instinct was to check how exposed I already was through index funds.
The answer: more than I emotionally expected, less than my panic suggested.
The S&P 500 is market-cap weighted. The larger a company becomes, the larger its weight in the index. That means the biggest AI-adjacent companies — Apple, Microsoft, Nvidia, Alphabet, Amazon, Meta, and Tesla — now represent a very large share of the index. I guess John Bogle didn't design index funds with this in mind.
So if you own a plain S&P 500 fund, you already own a serious AI exposure.
This is the part where many people discover that "passive investing" does not mean "neutral investing." It means you accept the market's current weighting of winners.
John Bogle's original case for index funds was beautifully simple: most people should own the market at low cost instead of trying to outsmart it. I still think that is right. But today, owning the U.S. market also means owning a concentrated bet on U.S. mega-cap tech.
But here is Damodaran's counterpoint, and it is genuinely reassuring: in a recent Substack piece on S&P 500 inclusion, he analyzed the price impact of adding major new constituents and found that the bump from passive fund buying has largely disappeared over the past decade. The mechanism that would amplify an AI crash through index funds — the reflexive selling as passive funds rebalance — is weaker than feared. The S&P 500 is not a single fragile thing. It is 500 companies, and the other 493 don't stop existing if Nvidia falls 60%.
That does not make the S&P 500 bad. It means you should know what you actually own.
The practical correction is also very Bogle-like: add global diversification.
A Total World index fund reduces the U.S. tech overweight by spreading exposure across Europe, Japan, emerging markets, financials, industrials, healthcare, energy, and boring companies that do boring things and occasionally make very real money.
Dalio's "Holy Grail" principle makes the same point mathematically: high-quality uncorrelated bets improve the return-to-risk profile of a portfolio. One high-quality bet with a 0.3 return-to-risk ratio becomes 4.3 times better when you hold 15 uncorrelated bets of equal quality. Not 15% better. 4.3 times. This is the mathematical case for diversification, and it has nothing to do with fear or conservatism. It is simply the structure of risk.
3. The 10% Speculative Sleeve Is Permission With Discipline
One of the most useful things I found in Graham is that he was not such a financial puritan.
He allowed room for speculation — within limits. The defensive investor could keep a small speculative allocation, separated from the core portfolio, and sized so that a full loss would not damage the broader plan.
That idea is psychologically brilliant.
Because the usual personal finance advice — "just buy index funds and never think again" — is correct for many people, yet incomplete for actual human beings with curiosity, ambition, ego, fear of missing out, and a Wi-Fi connection.
A small speculative sleeve admits reality. Sometimes you want exposure to a structural shift. Sometimes there is genuine optionality. Sometimes refusing to participate creates its own anxiety.
The key is the order. The core comes first. The speculative sleeve comes second.
For me, this means something like:
- Core portfolio: globally diversified, low-cost index exposure, with bonds and cash calibrated to time horizon and risk tolerance.
- Speculative sleeve: up to 10% for higher-conviction, higher-risk ideas — including AI-adjacent companies — where the business has a real moat, the price offers a margin of safety, and the position size prevents drama.
Most people get this backwards. They run 90% emotion and 10% discipline, then call it strategy.
Graham would probably raise one eyebrow and leave the room.
4. The Hidden Horse: When Value Investing Finds AI
This is where the framework gets exciting rather than defensive.
The most counterintuitive finding from recent Morningstar research: of the companies most exposed to AI as a future revenue driver, 80–100% are currently screening as undervalued. The market is so concentrated on the names everyone already knows — Nvidia, Microsoft, Meta — that entire categories of AI beneficiaries are trading at discounts to intrinsic value.
The market often overpays for the cleanest story and underprices the less glamorous infrastructure behind it. In a gold rush, the romantic hero may be the miner, but the durable business may be selling the tools, logistics, and bottleneck machinery.
That is why companies like ASML, TSMC, AMD, Oracle, and other infrastructure-layer players are worth studying carefully.
Take ASML as an example. It is one of the most important companies in the semiconductor supply chain because it produces extreme ultraviolet lithography machines — the equipment needed to manufacture the most advanced chips. No advanced AI hardware stack exists without companies like this in the background.
When ASML sold off after weaker guidance, the story around AI had not disappeared. The price had changed. That is the kind of moment value investors care about.
The useful question is not "Do I believe in AI?" The useful question is: at this price, with these cash flows, this moat, this balance sheet — am I being compensated?
That question is less sexy than "AI will change the world." It is also much more useful.
5. What the Thought Leaders Actually Seem to Be Doing
What I found most interesting is that the serious investors are not all screaming the same thing.
They disagree. They hedge. They resize. They change their minds. They separate the technology from the valuation.
Ray Dalio holds gold, non-U.S. equities, commodities, and bonds alongside equities — his All-Weather portfolio is designed to perform across all economic environments, not to predict which one comes next. His specific 2026 advice: diversify away from USD-denominated assets, add hard assets, and treat Chinese AI commoditization as a genuine threat to U.S. AI pricing power.
Damodaran trimmed two Magnificent Seven holdings but did not exit AI entirely. His portfolio adjustments were made on price discipline — when the story stopped justifying the price, he reduced. He remains invested in businesses where the AI story is being validated by actual earnings.
Howard Marks had a literal change of mind after a hands-on tutorial with Claude — and updated his memo accordingly. He now believes AI is structural, not a fad, but still insists that "no one should be all-in." The transformation is real; the appropriateness of current prices at current rates is not.
Bill Ackman moved $2.09 billion from Alphabet into Microsoft in Q1 2026. This was not a retreat from AI — it was a platform bet within AI. His view: mega-cap AI names are undervalued given the scale of the transformation.
Charlie Munger, in his last public comments before his death in November 2023, was characteristically blunt about the AI frenzy: show him the incentive, and he'd show the outcome. Wall Street gets paid to sell AI stocks. His inversion — what would destroy me here? — remains the most useful single question to ask before any speculative allocation.
Respect the technology. Distrust the crowd. Watch the price. Size the risk.
6. My Practical Framework
After going through all this, my current framework is simple.
The core: 90%
A globally diversified, low-cost portfolio. For me, this means the core should not depend on one country, one sector, one currency, or one theme. The S&P 500 can still be a major component, but a U.S.-only portfolio now carries meaningful concentration in mega-cap tech. Global diversification is the correction. This is the Bogle/Dalio part of the portfolio: low cost, broad exposure, humility built into the structure.
The speculative sleeve: up to 10%
AI-adjacent investments can belong here, but only when three conditions are met:
- The business has a real moat. Patents, network effects, switching costs, scale, distribution, supply-chain control, or unique technical capability. A press release with "AI" in it does not count.
- The price offers a margin of safety. "Cheaper than Nvidia" is not a valuation argument. The company has to be attractive relative to its own intrinsic value.
- The position size keeps me sane. If a position can damage my financial life or sleep, it is too large. Excitement is not a risk-management system.
The discipline: always
Rebalance. Avoid adding aggressively when the position is already working and ego starts giving investment advice. Avoid abandoning the core because it looks boring next to whatever stock is currently behaving like a meme with a Bloomberg terminal.
And above all: separate thesis from price. AI may transform the world. That still does not mean every AI stock deserves every valuation.
7. The Conclusion I Actually Believe
The AI bubble is probably real in parts of the market. The AI transformation is also probably real. That combination is what makes the whole thing so difficult.
If AI were obviously fake, the answer would be easy. If AI were obviously cheap, the answer would also be easy. Instead, we are dealing with a powerful technology, massive capital flows, concentrated market leadership, genuine productivity potential, speculative excess, and very uncertain timing.
So the answer is: do not panic.
Own the market, but understand your concentration. Diversify globally. Keep a speculative sleeve if you want one. Use Graham's margin of safety, Bogle's low-cost discipline, Dalio's diversification logic, Damodaran's valuation framework, Marks's cycle awareness, and Munger's brutal incentive filter.
The goal is not to predict the exact top of the AI cycle. The goal is to build a portfolio that lets me participate if the transformation continues — and survive if the market discovers, once again, that trees do not grow to the sky.
That may sound less glamorous than catching the next Nvidia. Fine.
I would rather be solvent, diversified, and occasionally lucky than brilliant, concentrated, and one rate cycle away from a personality crisis.