The $1.5 Million Risk In AI-Heavy Portfolios

Part of our AI selloff guide. Start here: what the Nasdaq correction means for your money.

Follow the whole story: every piece in this series is on our AI storyline page.

The Nasdaq 100 closed Monday more than 10% below its Oct 6 peak, its third correction of 2026, before bouncing to about 9% below by Wednesday. Alphabet is down 16% in a month and Nvidia 14%, while Apple and Microsoft are up 5% and 7%. If you own these companies, the question isn’t where the index goes next week. It’s how much a full-blown drop would take from you, in dollars. Here is the math for a $5 million portfolio with $2 million, 40%, in AI megacaps.

Scenario	AI megacaps ($2M)	Everything else ($3M)	Total loss	Portfolio after
Correction only (megacaps −15%, rest −5%)	−$300K	−$150K	−$450K (−9%)	$4.55M
2022-style (megacaps −50%, rest −16%)	−$1.0M	−$480K	−$1.48M (−30%)	$3.52M
2000-style (megacaps −75%, rest −15%)	−$1.5M	−$450K	−$1.95M (−39%)	$3.05M

In a 2022-style drop, the 40% in AI stocks causes two-thirds of the loss.

The same $5 million held entirely in the “everything else” mix would have lost about $800,000 in the 2022-style case and $750,000 in the 2000-style one. So the concentration itself costs an extra $680,000 to $1.2 million. And a 30% loss needs a 43% gain just to get back to even; a 39% loss needs 64%.

See more lines: where the scenario numbers come from

All scenarios are our estimates, not forecasts. “AI megacaps” means an equal mix of Nvidia, Microsoft, Alphabet, Amazon, Meta and Apple. Their peak-to-bottom falls in 2021–22 were about −66%, −37%, −44%, −56%, −77% and −31%, an average near −52%; we use −50% because the peaks and bottoms didn’t all land on the same days. “Everything else” is a 60/40-style mix of stocks and bonds, which lost about 16% in calendar 2022 because bonds fell too. In the 2000 bust the Nasdaq 100 fell about 83%; leaders such as Cisco, Intel and Oracle fell 80% or more, Microsoft about 65%. We use −75% for the megacaps, a bit less than 2000’s leaders fell, because today’s leaders are already highly profitable, and −15% for the rest because bonds rose in 2000–02 while stocks fell. Wednesday’s snapshot: Nvidia 16% below its 52-week high, Alphabet 28%, Meta 19%, Amazon 6%; Microsoft and Apple closed at 52-week highs.

Could These Giants Really Fall That Far?

They already have. From their 2021 peaks to their 2022 lows, Meta lost about 77% and Alphabet about 44%. Those weren’t failing companies. Both were back to record profits by 2024. Their stocks fell as rates rose and investors stopped paying any price for growth. Today’s version of that worry is whether roughly $900 billion of AI spending worldwide this year (JPMorgan’s estimate) earns its money back. Alphabet and Meta are among the ones writing those checks, and they’re down 16% and 13% this month while Apple is up 5%.

Why Do So Many Portfolios End Up At 40%?

Because nobody decided it. A position bought years ago grows into a quarter of the account. Index funds add more of the same names: the biggest AI companies now make up a large share of the S&P 500 and an even larger share of the Nasdaq 100. Add stock compensation for tech employees, and 40% is easy to reach without a single new purchase. The usual reason people don’t trim is the tax bill. If $2 million of stock cost $500,000, selling it all triggers a $1.5 million gain, about $357,000 in federal tax at the 23.8% top rate on long-term gains, before any state tax.

That’s the real trade-off for most concentrated families: a tax bill you can see versus a drawdown you can’t. Many $5 million-plus portfolios would lose more in a 2022-style drop than they’d pay in tax to fix it. A Trefis partner firm can show you how concentrated you really are.

What Do Concentrated Holders Actually Do?

Five common tools, none of which require selling everything at once. Which fits depends on your taxes, age and goals, so treat these as questions to bring to an adviser, not instructions.

Harvest losses before Dec 31. This year’s selloff has left losses in other holdings: Oracle is down 50% over 12 months and CoreWeave 58%. Realized losses can offset gains from trimming a winner. Watch the 30-day wash-sale rule if you plan to buy back.

Sell in stages. Spreading sales over two or more tax years can keep more of the gain in lower brackets. Late December and early January are two tax years, a few weeks apart.

Collar the position. Buying a put and selling a call puts a floor and a ceiling on a stock, often at little upfront cost. Collars that are too tight can count as a sale for tax purposes, so the structure matters.

Exchange funds. You contribute your shares to a pool with other investors’ shares and get a diversified stake back, generally without a taxable sale. They usually require a seven-year hold and are open only to wealthier investors.

Give appreciated shares. Donating stock, directly or through a donor-advised fund, avoids the gain and can bring a deduction. Direct indexing can also help, by building a broad portfolio that leaves out the names you already own too much of.

At 40%, a 2022-style drop would cost a $5 million portfolio about $1.5 million. Roughly $680,000 of that is the price of concentration alone. With losses available to harvest before Dec 31 and several tax-aware ways to diversify, this selloff is a better time than most to bring that number down.

More from the guide: AI chip stocks · Palantir.

What Would A 2022-Style Drop Cost You?

Corrections test concentration, not conviction. If you have $1 million or more invested, see how exposed your portfolio is to the AI trade in a free 15-minute call with a Trefis partner firm.

For questions about this article or underlying research, email help@trefis.com.