The hardest stock to trim is the one that made you look smart.
That is the NVIDIA problem for a lot of serious investors now. You bought it early, or added during the AI panic, or simply held through a move that turned a normal position into the loudest line item in the account. The company kept executing, the estimates kept moving higher, and what used to be a growth allocation quietly became a portfolio-defining bet.
The uncomfortable answer is that NVIDIA can still be a remarkable business and still be too large for your portfolio. The risk may no longer be that NVDA is obviously expensive. The risk is that its success changed your portfolio without asking you.
How NVDA becomes a concentration risk
Most concentration problems do not begin with reckless behavior. They begin with a position that worked.
NVIDIA is the cleanest example in the current market. A stock that compounds several hundred percent over a few years does not need new buying to become oversized. A 4% position can become 10%. A 7% position can become 20%. If the rest of the portfolio rises more slowly, the winner takes control of the risk budget simply by outrunning everything else.
That is different from making a fresh decision to put 20% of your money into one stock today. Most investors would hesitate before doing that. But they will tolerate the same exposure if it arrived through gains, because the position carries emotional proof: it worked. Selling feels like punishing success. Holding feels like loyalty to a thesis that was right.
There is also the hidden index problem. Even if your direct NVDA position looks manageable, NVIDIA is likely sitting inside your S&P 500 fund, your total-market fund, your growth ETF, your QQQ position, and possibly your target-date fund. The direct shares are only the visible part. The look-through exposure across funds is where the real number lives.
That is why the right question is not, “Do I like NVIDIA?” It is, “How much NVIDIA do I actually own across every account, and what happens to my plan if the stock has a normal drawdown?” We wrote about the broader issue in Is Your Portfolio Too Concentrated?, but NVDA is the version that tests people because the company quality is real.
What the numbers say right now
NVIDIA is not a story stock floating on vibes. The financials have been extraordinary.
For fiscal 2026, NVIDIA reported revenue of $215.9 billion, up 65% from the prior year. In the first quarter of fiscal 2027, revenue reached $81.6 billion, up 20% sequentially and 85% year over year. Gross margins remain in a territory most hardware businesses never touch: roughly 75% on both a GAAP and non-GAAP basis in the latest quarter.
That matters because a lot of lazy NVIDIA commentary still treats the stock as if the price moved while the business stood still. That is not what happened. The share price exploded, but earnings power exploded too. Depending on the source and estimate set used, NVIDIA recently traded around the high-teens to mid-20s forward earnings range, with one recent data provider showing a trailing P/E near 29.5 and forward P/E near 24.0 as of late June 2026. Barron's recently cited an even lower forward multiple, around 19, after the stock pulled back below $200.
That is the twist. NVIDIA may not be the cartoonishly expensive stock many investors assume it is. The market has repriced the business, but the earnings base has caught up faster than almost anyone expected. Analysts still lean heavily positive, with one public consensus feed showing a Strong Buy rating from 38 analysts and an average 2026 price target near $299.
None of that makes the stock safe. It changes the risk. The bear case is less about “this company has no earnings” and more about whether the current earnings power is durable, whether AI infrastructure spending stays aggressive, whether hyperscalers eventually push more work to custom silicon, and whether margins this high can survive competition. Recent market concern has focused exactly there: AI chip competition, custom chips from large customers, and whether AI capex can keep expanding at the current pace.
The point is not that the bulls are wrong. The point is that the stock already reflects a company that has executed almost perfectly. When a holding becomes enormous, you do not need the thesis to break for the position to hurt you. You only need expectations to stop rising.
The real risk is not the company — it is the portfolio math
A great company can still be a bad-sized position.
That sentence sounds obvious until the stock in question is your best winner. Investors are usually comfortable saying a speculative company should be capped at a small weight. They are much less comfortable applying the same discipline to a dominant, profitable, cash-generating leader. NVIDIA feels different because the business is different. It has become the infrastructure supplier to the AI buildout, not a small-cap lottery ticket.
But portfolio math does not care why a position is large. It only cares what happens if it falls.
If NVDA is 5% of your total portfolio, a 30% drawdown costs you 1.5% at the portfolio level. Annoying, but manageable. If it is 20%, the same drawdown costs you 6%. If it is 30%, a normal single-stock correction becomes a portfolio event. And NVIDIA has shown repeatedly that even great companies with strong fundamentals can move violently when expectations shift.
This is where serious investors need to separate business quality from risk contribution. NVIDIA may deserve a premium. It may continue compounding. It may remain the dominant AI platform company for years. None of that answers whether it should be 18% of a portfolio also loaded with QQQ, VOO, MSFT, AAPL, and other AI-adjacent names.
The common mistake is to evaluate NVDA as a standalone stock while ignoring the correlated exposure around it. A portfolio with NVIDIA, Microsoft, Amazon, Broadcom, Taiwan Semiconductor, QQQ, and an S&P 500 fund is not seven unrelated bets. It is heavily exposed to the same AI infrastructure cycle, the same mega-cap growth factor, the same valuation regime, and the same investor crowding. When the cycle is working, that stack feels brilliant. When the market questions the cycle, all of it can move together.
That is why we think about NVDA as a position-sizing question before we think about it as a buy-or-sell question. The stock can remain a core holding. It just does not get to become the portfolio by accident. If you need a broader framework, How Much of Your Portfolio Should Be in One Stock? is the right companion piece.
How to think about trimming
Trimming a winner is not a confession that you were wrong. It is often the opposite. You were right, and now the position has grown past the size that made sense when you bought it.
The first step is to define the number. Not the number in your brokerage account. The total number. Add direct NVIDIA shares, the NVIDIA weight inside each ETF or mutual fund, any employer-plan exposure, and any correlated AI or semiconductor names that behave like cousins of the same trade. If that total makes you uncomfortable, the answer is not necessarily to sell everything down immediately. It is to decide what exposure you would choose today if you were rebuilding the portfolio from cash.
That mental reset is useful because it strips out the anchoring. The fact that your cost basis is low does not tell you the right position size. The fact that the position is up does not tell you the right risk budget. The question is simpler: given your goals, time horizon, taxes, and existing holdings, what percentage of the portfolio should depend on NVIDIA continuing to execute?
For some investors, the answer may still be high. If you have a long horizon, strong conviction, low need for near-term liquidity, and a portfolio intentionally built around secular growth, a large NVDA position may be a chosen risk. That is different from an accidental one.
For others, the right move is gradual trimming. Sell a portion, not the whole position. Use price strength or predetermined bands rather than emotion. Redirect proceeds into areas that genuinely diversify the risk: value, international, small-cap, bonds, cash for known goals, or simply less correlated equity exposure. If the position is in a taxable account, tax lots matter. Selling the highest-basis lots first may reduce the tax hit. Pairing gains with losses elsewhere can help. Spreading trims across tax years may be better than doing one dramatic rebalance.
This is where investors often freeze. They know the position is too large, but the embedded gain creates a tax bill, and the tax bill becomes a reason to do nothing. That is understandable, but it is not always rational. A tax cost is visible and immediate. Concentration risk is invisible until it arrives all at once. We covered that trap in Rebalancing Without Triggering a Tax Bomb, and NVDA is exactly the kind of winner where the framework matters.
There is also a behavioral benefit to trimming before you are forced to. When you reduce an oversized position on your own terms, you preserve conviction in the remaining shares. When you wait until the stock is already falling hard, every decision becomes emotionally contaminated. You are no longer managing risk. You are negotiating with regret.
What a sensible NVDA plan looks like
A sensible plan does not start with “sell” or “hold.” It starts with a range.
For example: “I am comfortable with NVIDIA at 8% to 10% of my total portfolio, including fund overlap. If it rises above 12%, I trim back. If it falls below 6% and the thesis remains intact, I may add.” That kind of rule will never be perfect, but it is better than vibes. It turns a stock that can dominate your emotions into a position with boundaries.
The range should reflect the role NVDA plays. Is it a long-term compounder? A high-conviction AI infrastructure bet? A tactical winner you are letting run? A legacy position with a huge tax gain? Each answer implies a different tolerance for size.
What does not work is pretending the stock is diversified because NVIDIA itself sells into many markets. Company diversification is not portfolio diversification. A business can have data center, gaming, software, networking, automotive, and sovereign AI exposure, and still be one stock with one multiple, one investor base, one management team, and one cycle of expectations. If the market decides AI infrastructure spending is overearning, the stock will trade like one stock.
That is the discipline the NVIDIA moment demands. Respect the business. Respect the execution. Respect the possibility that the bull case continues. But also respect the portfolio math. The investors who get hurt by winners are not usually the ones who liked bad companies. They are the ones who let a good company become too much of the plan.
The take
NVIDIA may still be one of the highest-quality growth companies in the market. The latest financials support why investors remain excited: massive revenue growth, extraordinary margins, and a consensus that still leans strongly bullish. This is not an obvious junk valuation story.
That makes the position-sizing question more important, not less. When a stock is both fundamentally excellent and already a major winner, investors become very good at justifying whatever size it has become. But your portfolio does not need admiration. It needs balance.
If NVDA is still a right-sized holding, keep your thesis and monitor it. If it has quietly become the position that determines whether your plan works, trim with discipline, manage the tax bill intelligently, and let the remaining shares be a holding you can live with through a real drawdown.
See your true NVIDIA exposure across direct shares, ETFs, retirement accounts, and overlapping AI holdings in Aurvus — then decide whether the position is still sized on purpose. Check your real concentration →
Aurvus provides portfolio analysis for informational purposes and is not a registered investment advisor. Consult a qualified professional about your specific situation before making investment decisions.
