Concentrated Stock Position: What Each Exit Actually Costs

Key takeaways

• Over the five years to 4 September 2026, NVIDIA weekly closes gave 47.87% annualised volatility against 16.29% for an S&P 500 tracker, a ratio of 2.94 to one.
• Selling half that holding into the S&P 500 tracker cut the blend to 30.21% volatility. Selling half into a US bond tracker cut it further, to 24.53%.
• The S&P 500 and Nasdaq 100 trackers correlated at 0.94 on weekly returns, so moving from one large-cap into a technology index shifts less risk than it appears to.
• On an £800,000 gain at the 24% rate in force from 6 April 2026, spreading the sale over five tax years saves £2,880 of Capital Gains Tax, not a fortune.
• Exchange Place: All Cap III LP reported $1,432,209,044 sold to 336 investors in its August 2026 Form D/A, an average of about $4,262,527 each.

You own one stock, it's most of what you have, and selling it means a capital gains bill you can see from space. What are the routes out, and what does each one actually cost?

There are four, and they aren't equally available. You can sell in stages. You can swap the shares into an exchange fund, if you're a US taxpayer. You can hedge with options. Or you can hold. Modelled against five years of real weekly returns and the 2026 to 2027 UK tax rules, the picture is unflattering to three of them. Staged selling saves much less tax than people expect. Exchange funds lock money up for seven years and are sold a few hundred investors at a time. Hedging is priciest exactly where the position is wildest. Doing nothing is the only one that costs nothing up front, which is why it wins by default rather than on merit.

How much risk a concentrated stock position actually carries

Start with the size of the problem. Take weekly closing prices from 3 September 2021 to 4 September 2026, 261 weekly returns in all. NVIDIA (NVDA) came in at 47.87% annualised volatility. The SPDR S&P 500 ETF Trust (SPY) came in at 16.29%. That's a ratio of 2.94 to one, and it is the whole case for diversification in two numbers.

Volatility understates how that feels. Over the same window NVIDIA's deepest fall from a weekly closing peak was 66.0%. The tracker's was 24.8%. A reader who wants the difference between those two measures set out properly can find it in our note on volatility vs drawdown.

NVIDIA is an extreme, and picking the extreme would be cheating. So the same calculation was run on eight large US listings across sectors. Coca-Cola (KO) came in at 16.56% annualised volatility, barely above the index itself. Correlation with the S&P 500 tracker ranged from 0.17 for Exxon Mobil (XOM) up to 0.70 for NVIDIA. Single-stock risk is not one number. It's a range, and where your holding sits inside that range decides how much any exit is worth.

Selling into the index diversifies less than the arithmetic suggests

Here's the part most sell-down plans skip. What you buy with the proceeds matters as much as what you sell. Weekly return correlations over the same five years:

The 0.94 between the two equity trackers is the line to look at. Selling a mega-cap technology name and buying the Nasdaq 100 keeps you in almost the same weather. NVIDIA's correlation with the Nasdaq 100 tracker was 0.79, against 0.70 with the broad index. The same problem shows up inside broad funds too, which is the subject of our piece on fund overlap and mega-cap concentration.

Run the blends and the ranking is not the intuitive one. Holding 75% NVIDIA and 25% of the S&P 500 tracker gave 38.87% annualised volatility. Half and half gave 30.21%. But half NVIDIA and half the iShares Core U.S. Aggregate Bond ETF (AGG) gave 24.53%, lower still, because bonds correlated at 0.15 with the stock and carried far less volatility of their own. Selling three quarters into the equity index took the blend to 22.31%. Only the full sale reached the index's own 16.29%.

So the first finding is that a concentrated stock position sold halfway into an equity index is still a volatile portfolio. The second is that the destination does more work than the fraction sold.

Staged selling saves less UK tax than it feels like it should

Now the tax. Take a £1,000,000 holding with a £200,000 cost, so an £800,000 gain, held by a higher rate taxpayer. From 6 April 2026 the Capital Gains Tax rate for higher and additional rate payers is 24%, and the annual exempt amount for the 2026 to 2027 tax year is £3,000.

Sell the lot in one year and the bill is £191,280. Sell a fifth a year for five years and it's £188,400. The saving is £2,880, or 0.29% of the position. That's the entire tax benefit of staging for a higher rate payer, because the only thing you gain is four extra £3,000 allowances at 24%.

It's better for someone with a low income. GOV.UK's own worked example puts £20,000 of taxable income against a £37,700 basic rate band, leaving £17,700 of headroom taxed at 18% rather than 24%. Claim that headroom five times instead of once and the saving rises to £7,128, or 0.71% of the position. Real, but still under 1%.

Set that against the volatility numbers above and the ordering becomes clear. Staging changes the tax bill by tenths of a percent. It changes the risk profile by tens of percentage points, because for most of the period you are still holding the stock. Anyone weighing a multi-year sell-down may also want our work on employer stock concentration, which covers the same decision from the equity compensation side.

An exchange fund swaps the tax bill for a seven-year lock

US taxpayers have a route that doesn't exist in the UK. You contribute your shares to a partnership alongside other people contributing theirs, and take back a share of the pooled portfolio. Section 721(a) of the Internal Revenue Code makes that contribution tax-free. Section 721(b) then removes the exemption where the partnership "would be treated as an investment company (within the meaning of section 351) if the partnership were incorporated".

That is the constraint the whole structure is built around. The regulation defines an investment company as one where "more than 80 percent of the value of whose assets" are "held for investment and are readily marketable stocks or securities". Stay above 80% and the swap is taxable. So exchange funds hold something else. Belcrest Capital Fund LLC, an Eaton Vance vehicle with net assets of about $633.0 million at the end of 2011, told the SEC it "normally invests at least 20% of its assets as so determined in certain real estate investments". A fifth of the fund is there for the tax code, not because you wanted property.

Then the lock. Section 704 treats the contributing partner as recognising gain if the contributed property is distributed to another partner "within 7 years of being contributed". Section 737 mirrors it from the other direction, using property "contributed to the partnership by the distributee partner within 7 years of the distribution". Congress lengthened that window from five years to seven in 1997. Leave early and the deferral you bought unwinds.

The fees are not trivial either. Belcrest's underlying portfolio charged an investment adviser fee of 0.46% of average daily net assets for 2011, on top of a 0.20% annual servicing fee at fund level and a 0.60% annual management fee on the real estate subsidiary's gross investment assets. That is a stack of charges running for seven years against a deferral, not a forgiveness, of the gain.

These are private placements, and the Form D filings show the scale. Exchange Place: All Cap III LP reported $1,432,209,044 sold to 336 investors on 20 August 2026, an average of about $4,262,527 each, with $45,000,000 of estimated sales commissions disclosed. Its sibling, Exchange Place (Equity Only): All Cap LP, reported $1,313,299,247 from 199 investors in July 2026. Roughly $6.6 million a head. This is not a retail product, and nothing equivalent exists for a UK investor.

What a collar can and cannot do for a concentrated stock position

The third route is to keep the shares and buy protection: a put below the price, usually funded by selling a call above it. The appeal is obvious. You cap the downside without a disposal, so no tax event.

Except the tax code watches for exactly that. Section 1259(c)(1) treats you as having sold an appreciated position if you enter a short sale, an offsetting notional principal contract, or "a futures or forward contract to deliver the same or substantially identical property". A fifth catch-all covers other transactions with substantially the same effect, but only "to the extent prescribed by the Secretary in regulations". Where a given collar sits in that scheme is a question for a tax adviser, not a blog. The statute also gives a narrow escape for positions closed "on or before the 30th day after the close of such taxable year", which tells you how carefully the timing is policed.

The cost problem is simpler and doesn't depend on jurisdiction. Option premiums scale with implied volatility, and implied volatility scales with the thing you are insuring. Over the five years to September 2026 the Cboe Volatility Index averaged 19.14 and the Nasdaq-100 version averaged 23.79. Those are index-level numbers. A stock realising 47.87% volatility is a different order of expense again. The insurance costs most precisely where the concentration hurts most, which is the same reason nobody sells cheap flood cover on a flood plain.

The UK routes that move shares without a disposal

UK holders have narrower options, and they're mostly about wrappers. Shares coming out of a Save As You Earn scheme escape Capital Gains Tax if they go into an ISA "within 90 days of taking them out of the scheme". Share Incentive Plan shares get the same 90-day treatment. That is a genuine gap in the tax net, and it closes fast.

The size of the gap is the catch. The ISA subscription limit is £20,000 for the 2026 to 2027 tax year. Against a £1,000,000 concentrated holding that's 2% a year, and 50 years to shelter the lot. Useful at the margin, irrelevant at scale. Where the sheltered assets should sit once they're inside is covered in our piece on asset location.

The strongest case against selling

Here is the objection, and it's a good one. Over the same five years used for every figure above, NVIDIA's price rose 908.4%. Anyone who ran a disciplined staged sale in 2021 gave up most of that. The critics of diversification schedules are not making a subtle point: for the specific holding, in the specific window, doing nothing won by a distance.

That's a real objection and it should be stated at full strength. It's also survivorship in its purest form. The 47.87% volatility and the 66.0% drawdown were the price of admission for the 908.4%, and they were visible before the return was. A framework that only tells you which stock to hold after the fact isn't a framework. The honest version of the counter-case is narrower: deferral has value, tax paid is gone, and an investor who genuinely understands the business may be pricing it better than the market. Whether that describes you is not something a correlation matrix can answer.

What this data cannot tell you

Every number above comes from one five-year sample, from 3 September 2021 to 4 September 2026, which spans a bear market and a very large technology rally. Correlations move, and they move most when it matters, as our note on correlation instability sets out. The 0.15 between NVIDIA and bonds is not a constant.

Two more limits. These are price returns, so dividends and bond coupons are excluded, which is fine for volatility and correlation and wrong for total return. And a backtest is not a forecast: the blends above show what those weights did, not what any weighting does next. The tax arithmetic is a worked example under 2026 to 2027 UK rules with no other gains, losses or reliefs in the year, and the US material describes a structure rather than any particular fund on sale today.

What would change the conclusion

Three things. If the correlation between your holding and the index you'd buy is nearer 0.17 than 0.70, the sale moves far more risk than the model above shows, and the case for it strengthens. If the annual exempt amount or the 24% rate moves, the staging arithmetic changes with it, though it would take a large move to make tenths of a percent matter next to tens. And if a UK equivalent of the section 721 swap ever appeared, the whole ordering would need redoing, because the binding constraint here is not risk measurement but the fact that a British holder's only real exit is a taxable one.

What none of this decides is how much concentration you're carrying in the first place. That's a position-level question, and it's the one worth answering before any of the exits above are priced.