Key takeaways
- Residential property bought inside a self-directed pension triggers a 40% unauthorised payments charge, a scheme sanction charge of generally 15%, and a 15% surcharge past the threshold.
- A scheme can borrow 50% of the net fund value before the purchase, so HMRC's own example of a £200,000 fund carrying £50,000 of debt leaves £25,000 of headroom.
- Stamp duty on a £275,000 freehold commercial purchase is £3,250: nothing on the first £150,000, then £2,000, then £1,250 on the last slice.
- One published Property SIPP schedule runs to £1,500 a year before VAT, or £1,800 with it, for a borrowed, self-managed, VAT-registered single property.
- In the Investment Property Forum's data for the decade to 2013, a one-property portfolio averaged 16.7% standard deviation against 12.8% for twenty.
A pension can own the premises your business trades from, and residential is the line it cannot cross
Can your pension buy the building your business trades from? Yes, provided the building is commercial. What counts as commercial is set out in HMRC's Pensions Tax Manual, and it's more specific than most people expect.
A self-invested personal pension is what the legislation calls an investment-regulated pension scheme. HMRC defines one as a scheme where "the member is able (whether directly or indirectly) to direct or influence the manner of investments the scheme makes." A small self-administered scheme qualifies too, on a separate test: at least one member who can direct the investments, and "fewer than 50 members". The label matters. Investment-regulated schemes are the only ones taxed for holding what the Finance Act 2004 calls taxable property, and taxable property means residential property plus most tangible moveable things: art, antiques, classic cars, plant and machinery.
Buy one of those inside the scheme and the bill is punitive by design. The member pays an unauthorised payments charge at a flat 40% of what was spent. The scheme administrator pays a scheme sanction charge, "generally an amount of 15% of the value". If the payment reaches the surcharge threshold, broadly 25% or more of the member's pension rights inside a 12 month period, a further 15% lands on the member. HMRC puts the member's own total at 55%. The two liabilities fall on different people, the member and the administrator, and together they reach 70% of the price paid.
The boundary between commercial and residential gets drawn building by building, and it's worth reading the guidance literally. "If a building includes a shop with a wholly separate flat above it is treated as two separate buildings. The flat is a residential property and the shop is a commercial one." Wholly separate means a separate entrance and no interconnection. Where the two parts share a common entrance and you can move between them without crossing a communal area, HMRC treats the whole building as residential. One doorway decides the tax treatment of the entire asset.
Some exceptions run the other way. A hotel owned in its entirety is treated as commercial property; own one room of it on a long lease that carries a right to stay, and you're in timeshare territory, which is residential. A commercial building being converted to flats stays non-residential while the work runs, because it isn't yet suitable for use as a dwelling. It becomes residential when the works are substantially completed.
None of this makes a pension a workaround for residential property. The buy-to-let Section 24 finance cost restriction bites on property held personally, and the taxable property rules are what stop a scheme being used to step around it.
Commercial property in a SIPP buys rent that arrives untaxed
Here's the mechanism people are actually paying for. Section 186 of the Finance Act 2004 exempts income from investments held for a registered pension scheme from income tax. Section 271 of the Taxation of Chargeable Gains Act 1992 exempts the gains from capital gains tax. Rent paid into the scheme arrives gross. A later sale of the building triggers no capital gains tax at all.
Outside a pension the same rent is ordinary taxable income. GOV.UK states it in one line: "When you rent out property you may have to pay tax." Nothing in the pension version is a loophole. It's the standard treatment of a registered scheme applied to a building instead of a fund.
That gap is the entire commercial case for commercial property in a SIPP, and it's the same asset location question that decides which holdings belong in a wrapper and which don't. It's also why the taxable property rules exist. Without them the identical exemption would apply to a holiday cottage.
A second attraction has nothing to do with tax. If your company is the tenant, the rent it pays leaves the company and lands in your pension rather than a third-party landlord's account. The company keeps its premises. The scheme gets an income stream. Both are the same transaction seen from opposite ends.
The rent has to be a market rent, even when you are the tenant
This is where the arrangement most often goes wrong, and HMRC's position on it is short. A sponsoring employer or member renting a scheme-owned property "would have to pay the commercial rent due on the property. Failure to pay the commercial amount of rent for the property would lead to an unauthorised payment tax charge on an amount equivalent to the shortfall."
Underpaying, in other words, is not a saving. It converts the shortfall into an unauthorised payment taxed at 40%, with the administrator's sanction charge behind it. The same logic governs the purchase. Buy from a connected party above market value, or sell to one below it, and HMRC treats the difference as an unauthorised payment. Connected takes its meaning from section 993 of the Income Tax Act 2007, which reaches spouses, civil partners, relatives, relatives of a spouse, and companies you control.
The practical effect is that a SIPP property purchase from your own business has to stand up as an arm's length transaction. A price a stranger would have paid, a lease on ordinary terms, and rent that keeps being paid in the year the business would rather not pay it. The pension is a landlord with no discretion to grant a rent holiday.
Borrowing stops at 50% of the fund, not 50% of the building
Pension fund borrowing is permitted and capped. A registered scheme "may borrow an amount up to the equivalent of 50% of the net value of the fund prior to the borrowing taking place", and the asset being bought with the loan doesn't count towards that value. Go past the cap and the excess becomes a scheme chargeable payment carrying a 40% scheme sanction charge, reportable to HMRC.
HMRC's worked example shows how tight the arithmetic gets. A scheme holding £200,000 of assets with £50,000 of existing borrowing can borrow "£200,000 less £50,000 x 50% = £75,000, that is, further borrowing allowed of £25,000". Run the same test on a clean £200,000 fund and the ceiling on a purchase is £300,000, before stamp duty and fees are added.
That ceiling is the first concentration problem, and it's structural rather than behavioural. A fund just large enough to buy a modest industrial unit is usually a fund for which that unit becomes most of the portfolio. Borrowing lifts the purchase price without adding a second asset. It adds leverage to one building.
Stamp duty, VAT and roughly £1,800 a year before the building is let
The costs of commercial property in a SIPP arrive in three layers, and only the first is familiar.
Stamp duty land tax applies to purchases in England and Northern Ireland, and a commercial building uses the non-residential scale. Take HMRC's own example of a £275,000 freehold commercial property: "0% on the first £150,000 = £0", then "2% on the next £100,000 = £2,000", then "5% on the final £25,000 = £1,250", giving a "Total SDLT = £3,250". A new lease carries a second charge, calculated on the net present value of the rent.
VAT is the layer that surprises people. Supplies of land and buildings are normally exempt, but an owner can opt to tax. HMRC's notice is clear about what follows: "Once you have opted to tax all the supplies you make of your interest in the land or buildings will normally be standard-rated, and you will normally be able to recover any VAT you incur in making those supplies." The standard rate has been 20% since January 2011. If the seller has opted, the purchase is standard-rated, and recovering that VAT means the scheme registers, opts to tax in turn, and then charges VAT on the rent it collects. A tenant who cannot recover VAT feels that as a real rent increase.
Then the administration, which is where a property scheme separates from an ordinary one. iPensions Group publishes a fee schedule for its Property SIPP, and the version on its site in September 2026 gives a fair picture of the shape. Property purchase, £800. Property administration, £400 per annum. Trustee and administration, £550 per annum while the fund is uncrystallised. Borrowing arrangement, £200, then £150 per annum to administer the loan. VAT registration, £150, with £200 per annum for quarterly returns. A self-managed property carries a £200 per annum risk premium. Set-up is £250.
Add the recurring lines and a single borrowed, VAT-registered, self-managed property costs £1,500 a year before VAT, or £1,800 with VAT at 20%. The one-off lines come to £1,400 before VAT. Measured against the £275,000 building in the stamp duty example, £1,800 a year is 0.65% of the asset, payable whether or not the tenant pays.
That last figure is the one worth holding onto, because it isn't a percentage fee. It falls as the property gets larger and bites hardest on a small one. On a £150,000 unit the same £1,800 is 1.2% a year.
One building carried 16.7% volatility, twenty carried 12.8%
The concentration argument usually gets made loosely. The Investment Property Forum measured it. Its 2015 study Individual Property Risk analysed the performance records of over 1,000 commercial properties held over the period 2002-2013, then built simulated portfolios of different sizes from that sample.
The result is quotable: "A single asset portfolio (excluding ground rent investments) on average has a standard deviation of 16.7%; adding a second property reduces the portfolio's standard deviation to 14.8% whereas a portfolio of 20 assets on average has a standard deviation of 12.8%." Over the same decade the IPD All Property index itself ran at 12.7%.
Read the gap rather than the levels. About 3.9 percentage points of annual volatility separate one building from twenty, and nearly all of what's left at twenty is market risk that owning more buildings cannot remove. The IPF says adding properties beyond that point "results only in marginal reductions in portfolio risk". So the diversifiable portion of the risk in a single commercial property is real but bounded. It's the distance from 16.7% to roughly 12.8%, not the distance from 16.7% to zero.
The distribution matters more than the average. Most individual properties in the sample had a 10-year standard deviation below 16%, but "3% of the sample properties had standard deviations in excess of 30%", and the average individual property sat "in the region of 17%". A pension holding one building holds one draw from that distribution. The tail is populated by properties with short unexpired lease terms, few tenants and heavy capital expenditure needs.
The IPF is explicit about what drives that tail. Specific risk "is predominantly driven by lease events (particularly tenant default and lease expiry) and by asset management." A single-let unit occupied by your own company sits at the far end of that description: one tenant, one lease, one covenant. It's the same property concentration risk that appears when a house dominates a household balance sheet, with one difference: the tenant and the owner share a payroll.
The strongest case against the concentration argument
Here's the objection, and it's a serious one. Volatility measured on annual valuations isn't the risk that matters to someone who owns their premises and intends to occupy them for twenty years. If you never have to sell in a bad market, a valuation that moves is an accounting event and nothing more.
Three things push back on that. The scheme has to value the property anyway, for the borrowing test and for "annual returns" and benefit events, so the number becomes real at exactly the moments benefits are drawn. The correlation that the concentration argument cares about isn't price, it's income: if the business fails, the tenant and the covenant fail together, and the pension loses its rent in the same week its member loses their trade. No amount of valuation smoothing hides that one. And a scheme can't sell a third of a building to pay a pension.
The measurement itself cuts the other way too. A separate IPF study of index smoothing found that the valuation process "is therefore likely to result in individual property valuations which vary less than market prices", and put its central estimate of true property risk at "13% to 15%, or 1.3 to 1.5 times that observed in the valuation index". Over the period 1971 to 2005 a desmoothed standard deviation of 15% sat against an observed 10%. The 16.7% above is a valuation-based figure. On that evidence it understates the swing in a single building rather than exaggerating it, which is the same effect that makes REITs vs direct property look like different asset classes when they hold the same buildings.
There's a regulatory asymmetry worth knowing about too. Occupational schemes are capped at "five per cent" of their resources in employer-related investments by regulation 12 of the Occupational Pension Schemes (Investment) Regulations 2005. The cap doesn't apply to small schemes, defined as "a scheme with fewer than 12 members" where all the members are trustees and decisions are unanimous. A small self-administered scheme holding its sponsoring employer's premises is therefore doing something a larger scheme could only do within a five per cent cap. The exemption removes the cap. It doesn't remove the exposure the cap was written about.
What this evidence cannot tell you
Several limitations belong on the page rather than in a footnote.
The IPF sample is institutional. "Eight investors" supplied it, covering insurance company funds, managed property funds, pension funds and listed property companies, which skews larger, better let and better managed than the small industrial units and offices that actually end up as commercial property in a SIPP. If anything that biases the single-asset volatility figure downwards for this use case, not upwards.
The measurement window closes in 2013 and contains the 2008 crash, so it's a high-volatility decade by construction. The IPF says as much: the downturn "led to a tripling in the standard deviation (to 12.7%) in the overall IPD All Property return compared to the previous period." Its earlier study, which "covered the period 1995-2004", found the average standard deviation for a portfolio of 50 properties "was 4.9% compared to 12.7% now". The shape of the curve is the durable finding. The levels are period-specific.
The fee schedule is one provider's, published for one product. Others structure charges differently, and even this one reserves fees "payable on a time cost basis" for work it doesn't list, which makes a total harder to predict rather than lower.
And none of it measures what a business owner is usually optimising for, which is control of the premises and freedom from a landlord. The data cannot tell you what that's worth to you.
What would change the conclusion
If the fund is large relative to the building, most of this dissolves. The concentration case describes a scheme where the property is the bulk of the assets, and the 50% borrowing cap means a first purchase usually starts there. A second property bought into a much larger fund is a weight, not a concentration.
If the tenant isn't connected to the member, the link between the pension's income and the member's livelihood breaks, and the worst version of the risk goes with it. A building let to an unrelated business is a different asset from the same building let to your own.
If the taxable property rules changed, the whole shape changes. The 40% charge is what keeps residential property out of self-directed pensions, and residential is where most individual investors' property instincts point. That's a statutory choice, not an economic law, and it has been revisited before.
If what you want is the exposure rather than the building, HMRC already describes the route. An interest in a UK REIT is not treated as an indirect holding of taxable property, provided the scheme's interest stays under 10% of the share capital, voting rights or income, and isn't held to let a member occupy the property. LedgerTouch tracks a single illiquid holding as a weight like any other, which is mostly useful for seeing how large it has quietly become.
The thing to watch isn't the valuation letter that arrives once a year. It's the unexpired term on the lease and the number of tenants standing behind the rent, because those are what the IPF found drives the tail of the distribution. A building with eight years to run and three tenants is a different holding from the same building with eighteen months to run and one.