Microsoft Concentration Risk: How Big Has One Winner Grown in Your Portfolio?

Open your brokerage statement and find the allocation page. If you have owned Microsoft through the past three months, the percentage next to it is larger than the last time you looked — and you did not buy a single additional share to make that happen. That line moved on its own. Understanding exactly how it moved, and what it now commits you to, is a different exercise from deciding whether the company is doing well.

The prevailing reading of a quarter like this is that it validates the holding. The stock worked, the thesis worked, and the portfolio is worth more. All of that can be true at the same time as a less comfortable fact: the same price move that made you money also rewrote your risk budget without asking. Microsoft concentration risk is not a judgement about Microsoft. It is arithmetic about your portfolio.

📈 One quarter, and 28 years of context

MarketWatch reported that Microsoft’s stock was on track to post its biggest quarterly gain in 28 years. Two words in that sentence deserve attention before anything else.

The first is “on track.” The calendar third quarter closes today, September 30, 2026, so until the final print settles, the record is provisional rather than confirmed. The second is “28 years,” which places the last comparable quarter in the late 1990s. That is the useful part of the headline: not the magnitude, which you can read off your own account, but the rarity. A gain that has no peer in nearly three decades of trading history is, by construction, an outlier — and outliers are precisely the events that break a position-sizing plan, because plans are usually built around ordinary quarters.

I am deliberately not putting a percentage return in this article. MarketWatch is a secondary summary, and the figures that matter for your account are the ones on your own statement, not a rounded market number. What follows is the mechanism, sized with arithmetic you can check against your own holdings.

Step one: how drift compounds a weight

Portfolio weight is a ratio, which means a single position’s gain moves both the numerator and the denominator. The numerator moves fully; the denominator moves only by the position’s share of the total. That asymmetry is the whole engine, and it is why weights climb faster than most investors intuit.

Take a $100,000 portfolio with an $8,000 Microsoft position — an 8% weight, which is already an active bet rather than an index-like exposure. Hold everything else flat, and watch what a range of outcomes does to that 8%.

Notice that a doubling does not double the weight. It takes 8% to roughly 15%, because the portfolio itself grew 8%. That dampening is the reason drift feels manageable from a small base — and the reason it stops feeling manageable once the base is large. Run the identical exercise from a 20% starting weight and the same doubling lands you at 33.3%, because the denominator now absorbs far more of the move. Concentration accelerates concentration.

Where the number stops being a number

A 15% single-name weight means something specific and testable: a 40% drawdown in that one holding costs you 6% of total portfolio value, before anything else moves. At a 33% weight, the same 40% decline costs 13.2%. You can run this on your own statement in ten seconds, and the answer is the honest measure of what you are exposed to. It is a far better question than whether the company’s business is strong, because strong businesses have delivered 40% drawdowns many times without the business itself failing.

There is a second, less visible layer. If you also own broad index funds, you own Microsoft inside them, and your true exposure is the direct position plus the look-through weight of every fund you hold. I cannot size that for you — it depends on which funds you own and their current published holdings — but the effect only ever runs in one direction: your real exposure is higher than the line on the allocation page, never lower. Fund providers publish holdings monthly, and that is where the number lives.

Rebalancing bands, and what each one costs

“Rebalance when it gets too big” is not a rule, because “too big” is not defined. A band is a rule: you specify a target weight and a tolerance, and you act only when the tolerance is breached. The width you choose is the entire decision, and it trades two costs against each other — the cost of letting risk drift versus the cost of transacting.

Work the +75% row from the first table against a ±2pp band. The position sits at $14,000 in a $106,000 portfolio; 8% of $106,000 is $8,480; the trim is $5,520, or roughly 39% of the position. That is the number that stops most people, and it is worth saying plainly: a band that has not fired in years demands a large, uncomfortable trade the first time it fires. The discomfort is not a sign the rule is wrong. It is the accumulated drift you postponed, arriving all at once.

The bottom row is the one that describes most real portfolios. Absent a rule, you have not chosen a 15% weight in anything — the market chose it for you, and it will keep choosing, in whichever direction prices happen to run.

⚠️ Where this arithmetic stops holding

The mechanism above is clean because it ignores everything that makes trimming a real decision. Four things break it.

• Tax drag on a large gain. In a taxable account, a trim on a position that has appreciated sharply realises that gain, and the tax is paid in cash now against a risk reduction whose benefit is probabilistic and may never be needed. This is the single strongest argument against mechanical trimming, and it is not resolvable in the abstract — it depends on your rate, holding period, and available losses.
• Account location changes the answer. The same trade in a tax-advantaged retirement account carries no immediate tax consequence. Where a household holds multiple account types, rebalancing inside the sheltered account first can reduce a total-portfolio weight without triggering a realisation event. The constraint is capacity: the sheltered account may simply be too small to move the aggregate weight.
• Employer overlap. If your income, equity compensation, and largest holding all trace to the same company, portfolio weight understates your exposure, because a bad outcome would hit salary, unvested awards, and investments together. The drift table cannot price that correlation. Nothing can, precisely.
• The band is symmetric, and that is the point. A rule that only ever trims winners is not a rule; it is a preference. The same 8% target with a 6% lower bound obliges you to add after a decline, which is far harder in practice than trimming after a gain. If you would not honour the lower bound, you do not have a band — you have a ceiling, and you should say so.

One more limit deserves naming. A rare quarterly gain tells you what the price did; it does not settle whether the operating results underneath it justify the move, and this article makes no attempt to argue either side of that. Position sizing is worth doing precisely because it does not require you to win that argument.

What would change the answer

The case for leaving a large winner alone rests on one defensible claim: that the expected after-tax cost of trimming exceeds the expected benefit of the risk you remove. That claim can hold. It holds most clearly for a highly appreciated taxable position in a portfolio whose owner has a long horizon, stable income unconnected to the company, and the genuine capacity to sit through a deep drawdown without selling. Those conditions are specific, and they are checkable against your own circumstances rather than against a market narrative.

Where they fail, the arithmetic reasserts itself, because the drawdown math is indifferent to how good the quarter was. Two pieces of evidence would settle it for your account, and neither requires a view on the stock: the look-through weight once you add the position inside your funds to the position you hold directly, and the dollar loss your household could absorb if that combined exposure fell by half. If the second number is smaller than the first implies, the weight is too large regardless of what the next quarter brings — and if it is comfortably larger, the record quarter changes nothing you need to do.