The correction is coming. The calendar is not.
Jack Buffet asks whether AI is a market bubble, identifies what could trigger a correction, and builds a practical plan that does not depend on predicting the date.
For: A long-term investor worried that AI enthusiasm and expensive markets could produce a correction, but unwilling to turn that concern into a reckless timing bet.
A market correction is coming.
I do not know whether it starts next week, next year, or after stocks rise another 30%. Neither does the fellow on television who drew three arrows on a chart and discovered urgency.
Corrections are not rare mechanical failures. They are part of owning assets whose prices change as earnings, interest rates, expectations, and fear change. The useful question is not whether prices will fall again. They will. The useful question is whether your plan can survive the fall without requiring you to predict its date.
Artificial intelligence makes that question harder because two ideas can be true at once:
- AI can become a foundational, profitable technology.
- Investors can still pay too much for the companies attached to it.
The internet changed the world. That did not make every internet stock bought at every price a good investment. A transformative technology and a speculative bubble are not opposites. They can share an office.
The short version: Parts of the AI trade show bubble-like conditions: demanding expectations, concentrated index exposure, enormous capital spending, rising use of debt, and financial relationships that can reinforce the same optimistic story. But the leading companies also have real businesses and real profits. That makes “AI bubble” a risk diagnosis, not a countdown clock. Keep near-term money out of stocks, measure your actual exposure, rebalance to a written target, avoid leverage, and decide what you would buy or trim before the market gets loud.
Is AI a bubble?
My honest answer is: some prices may be, the technology is not, and nobody gets to know the final score in advance.
Calling something a bubble usually means price has separated from a sober range of future cash flows and is being supported by expectations that leave little room for disappointment. That condition can exist in one company, one layer of an industry, or one corner of the financing system without making the entire stock market fraudulent or worthless.
The latest official warnings are unusually direct.
The International Monetary Fund’s October 2026 Global Financial Stability Report says core AI firms have been profitable so far, which supports their valuations. It also says weaker or less durable AI investment and profitability could trigger a correction, amplified by rising leverage, concentrated exposure, and circular financing arrangements.
The Federal Reserve’s May 2026 Financial Stability Report described equity valuation pressures as elevated. Its price-to-earnings measure for the S&P 500 remained in the upper range of its historical distribution, while its estimate of the equity risk premium remained well below its historical average. In the report’s market outreach, respondents specifically raised concerns about AI valuations, debt-financed capital spending, and a possible correction in risk assets.
That is not a forecast that stocks collapse on Thursday. It is a description of fragility: when prices assume a great deal of good news, ordinary disappointment can do extraordinary damage.
Real profits do not grant immunity
The easiest bubble argument says AI is all story and no business. The IMF’s report does not support that shortcut. The core firms have produced profits.
The more serious concern is what today’s price requires tomorrow to deliver.
A company can report growing revenue and still be priced for even faster growth. It can generate cash while committing far more cash to chips, data centers, power, networking, and models. It can own excellent technology while customers resist the price needed to earn an acceptable return on that infrastructure.
Ask five questions:
| Question | Healthy evidence | Bubble-like evidence |
|---|---|---|
| Who pays? | Customers renew because the product creates measurable value | Trials and announcements grow faster than durable revenue |
| Who earns? | Revenue growth becomes operating cash after the complete cost of service | The ecosystem reports activity while economics remain elsewhere or undefined |
| Who finances? | Spending is supported by durable cash generation and a resilient balance sheet | Debt and outside capital must keep expanding to sustain the buildout |
| What does the price assume? | Several reasonable outcomes can support an acceptable return | One nearly perfect outcome is required to justify the price |
| What breaks the loop? | Customers, suppliers, and investors have independent economics | The same participants fund, buy from, and validate one another |
The phrase circular financing deserves attention. If companies invest in, lend to, or commit purchases from other members of the same ecosystem, each announcement can make demand look stronger and financing easier. The arrangements may support genuine construction and genuine revenue. They can also make the group more sensitive to one shared disappointment.
The question is not whether every deal is fake. The question is how much apparently independent demand depends on the same pool of capital and the same confidence.
Your broad fund is already in the room
An investor can own no AI-themed fund and still have substantial AI exposure.
S&P Dow Jones Indices reported that information technology represented 37.9% of the S&P 500 as of August 31, 2026. Its ten largest constituents included Nvidia, Apple, Microsoft, Amazon, both listed Alphabet share classes, Broadcom, Meta, Micron, and Tesla. Several sit directly in the AI buildout; others are major platforms, customers, or adopters.
That does not make the S&P 500 an AI fund. It does mean that buying a technology fund, a semiconductor fund, an AI ETF, and a few famous AI stocks beside a broad-market fund may create far more repetition than the account labels suggest.
Use a look-through test:
- List every stock and fund you own.
- Record the ten largest holdings in each fund.
- Combine repeated companies across accounts.
- Group companies by shared economic driver: chips, cloud capital spending, data centers, power, models, applications, or advertising.
- Calculate what percentage of the total portfolio depends heavily on continued AI spending and optimism.
The result may be a sensible deliberate tilt. Fine. Name it accurately. “I own five funds” is not a risk measure when the same companies keep walking in through different doors.
What could start the correction?
The trigger does not need to prove that AI failed. It only needs to make future cash flows look less valuable than investors expected.
Possible triggers include:
- revenue or profit growth that remains positive but misses ambitious forecasts;
- customers delaying infrastructure orders after building enough near-term capacity;
- capital spending producing a lower return than management expected;
- cheaper models reducing the economics available to model developers or infrastructure owners;
- competition shifting value from today’s leaders to customers or new entrants;
- debt costs rising or lenders becoming less willing to finance the buildout;
- power, permitting, chip, or supply-chain constraints slowing deployment;
- regulation, copyright disputes, security failures, or customer resistance raising costs;
- a rise in long-term interest rates that lowers the present value of distant earnings; or
- one important participant cutting commitments and forcing investors to reexamine connected deals.
None of those outcomes requires an apocalypse. A stock priced for exceptional growth can fall because growth becomes merely good.
The correction may also begin somewhere else. The Fed noted that higher long-term rates, inflation, geopolitical tension, private-credit stress, and weaker investor sentiment can all pressure asset prices. AI may be the crowded part of the market that reacts most sharply rather than the event that lights the match.
Do not turn bubble concern into a second bad bet
There are two ways to let an expensive market control you.
The first is to chase it.
The second is to sell everything, sit in cash, and require yourself to correctly choose both the exit and the return.
Market timing is not one decision. It is at least two. Even if you sell before a decline, you must buy again while headlines remain ugly and the future still looks uncertain. Many investors wait for clarity, which tends to arrive after prices recover.
The SEC’s Investor.gov asset-allocation guide starts with time horizon and risk tolerance, then recommends diversification and periodic rebalancing. That is less dramatic than a crash prediction because it does not need a crash prediction to work.
My preferred response to bubble risk is maintenance, not prophecy.
The correction-readiness test
Answer these before the next ugly week.
| Test | Ready | Not ready |
|---|---|---|
| Near-term spending | Money needed soon is in appropriate cash or other lower-volatility holdings | A house payment, tuition bill, tax obligation, or emergency fund depends on selling stocks at a friendly price |
| Allocation | Stocks, bonds, and cash have written target ranges tied to goals | The allocation is whatever the winners grew into |
| Concentration | Individual companies, sectors, and themes have measured limits | Exposure is described by the number of account line items |
| Leverage | No forced sale is built into the plan | Margin debt or borrowed money can turn a decline into liquidation |
| Contribution plan | Automatic long-term investing can continue through volatility | Contributions stop whenever headlines become frightening |
| Rebalancing rule | Calendar or threshold rules say when to restore targets | Decisions depend on mood, television, or a price forecast |
| Watch list | Desired businesses or funds have valuation ranges and research notes | “Buy the dip” means buying anything that fell |
| Sleep test | A plausible decline would be painful but survivable | A normal bear market would force selling or destroy the household plan |
If one row fails, fix that row. Do not use “AI bubble” as a substitute for knowing when the money is needed.
A five-step plan before prices fall
1. Move short-horizon money out of the argument
Money required within the next few years has a different job from retirement money needed decades from now. It should not be exposed to stock volatility merely because cash feels boring during a rally.
If you are matching money to a nearer deadline, start with Where to keep money you may need within five years.
2. Restore the allocation you already chose
If strong technology performance pushed a target allocation far beyond its intended range, rebalancing can reduce risk without requiring an all-or-nothing market call. Check taxes, trading costs, account type, and whether new contributions can repair the drift before selling.
Rebalancing is not a prediction that the largest holding will fall. It is an admission that no holding receives unlimited authority over the family’s outcome.
3. Put a ceiling on optional AI exposure
Keep the diversified core and the thematic bet in separate mental drawers. Write the maximum percentage that a dedicated AI position, semiconductor fund, or group of individual AI stocks may occupy. Include overlapping exposure already inside broad funds when you evaluate the total risk.
For a fuller exposure worksheet, use AI is a theme, not a portfolio.
4. Remove anything that can force the timing
Leverage makes a temporary price decline answer a permanent question. A margin call does not care that the ten-year thesis remains intact.
Do not finance a volatile theme with money that has a due date. Do not treat options, leveraged funds, or borrowed money as protection against being late to a story. Those tools can add path-dependent losses and forced decisions precisely when judgment is weakest.
5. Write the buy list and the sell rules now
For every company you might buy after a decline, record:
- the business evidence you require;
- the valuation range and assumptions;
- the maximum position size;
- the facts that would invalidate the thesis; and
- the source you will use to confirm each fact.
For every concentrated winner you already own, record what would cause a trim: portfolio weight, valuation, a broken thesis, a changed goal, or a tax-aware schedule. A falling price alone does not make a weak business cheap. A rising price alone does not make a strong business safe.
The plan should be specific enough to follow and humble enough to revise.
What I would watch instead of the prediction market
I would not spend the next year guessing the correction date. I would monitor whether the evidence is becoming more or less demanding:
- AI revenue and operating cash flow compared with capital spending;
- customer adoption, renewal, and measurable willingness to pay;
- debt issuance and financing terms across the infrastructure buildout;
- utilization of chips, data centers, and power commitments;
- gross margins after the complete cost of serving AI workloads;
- customer and supplier concentration;
- revisions to earnings expectations;
- index and portfolio concentration; and
- whether a current valuation still works under merely good—not perfect—assumptions.
This is where AI can help organize filings and challenge assumptions, but it cannot make the forecast dependable. Jack’s companion guide, How to use AI for investing without handing it the steering wheel, keeps the tool near the evidence and away from the trade button.
The bottom line
Yes, I expect another correction. That is not a brave call. It is what volatile markets do.
AI could be the center of it because expectations are high, capital spending is enormous, exposure is concentrated, and parts of the financing network may amplify disappointment. AI could also keep producing real growth for years before prices reset. Both outcomes fit the evidence we have today.
Do not build a portfolio that needs the bubble label to be correct this quarter.
Build one that can tolerate being early, late, and occasionally wrong. Protect the money with a near-term job. Measure concentration. Rebalance deliberately. Avoid forced selling. Keep a researched shopping list. Let price declines create choices instead of emergencies.
The correction is coming.
Your plan should not care which Tuesday.
This article provides general educational information, not individualized investment, tax, legal, or financial advice or a recommendation to buy, sell, or hold any security. Stocks and funds can lose substantial value. Diversification and rebalancing cannot guarantee a profit or prevent every loss. Market conditions, index holdings, company results, financing arrangements, and regulations can change. Verify current primary sources and consider qualified professionals who can evaluate your complete circumstances before acting.